No posts in this digest period.
\n"},{"id":"digest:digest-hn:354","title":"Digest: Hacker News: Aug 12 - Aug 13, 2026","link":"https://rssglue.subdavis.com/feed/digest-hn/rss","author":"System","published_at":"2026-08-13T10:00:00+00:00","content":"\n\nNo posts in this digest period.
\n"},{"id":"digest:digest-hn:351","title":"Digest: Hacker News: Aug 11 - Aug 12, 2026","link":"https://rssglue.subdavis.com/feed/digest-hn/rss","author":"System","published_at":"2026-08-12T10:00:00+00:00","content":"\n\n\nStealing Reasoning Traces from Proprietary LLM APIs
\n\n\n\n464 points | 199 comments
\n\n\n\n\"Stealing\" something you already paid for (tokens), but that you can't have access to(!). And trained on the sum of human knowledge.\nTraining on other model outputs ought to be business as usual, stop using morally charged terms made up by future monopolists: https://thomasdullien.github.io/posts/2026-06-15-rl-economic...\n— Aissen\n
England set to be one of the first countries to eliminate hepatitis C
\n\n\n\n304 points | 215 comments
\n\n\n\nI am glad they’re screening for it. I was born to someone who had the virus, but didn’t know I had Hep C myself until I submitted myself to an exceptionally thorough STI testing panel. I say exceptional, because the standard ones I’d done throughout college and early adulthood do not happen to include that particular test. So I feel a bit lucky to have been able to get diagnosed and treated in my mid-20s, before the disease could do much damage.\n— thex10\n\n
456 points | 191 comments
\n\n\n\nThis is the thesis behind the \"Information Theory, Inference, and Learning Algorithms\" course that was taught at Cambridge University.\n> Why unify information theory and machine learning? Because they are\ntwo sides of the same coin. In the 1960s, a single field, cybernetics, was\npopulated by information theorists, computer scientists, and neuroscientists,\nall studying common problems. Information theory and machine learning still\nbelong together. Brains are the ultimate compression and communication\nsystems. And the state-of-the-art algorithms for both data compression and\nerror-correcting codes use the same tools as machine learning.
Book (creative commons): https://www.inference.org.uk/mackay/itila/book.html
Lectures: https://m.youtube.com/playlist?list=PLruBu5BI5n4aFpG32iMbdWo...\n— farfatched\n
OpenAI’s head of ethics leaves less than a year after joining
\nhttps://archive.ph/W48UVhttps://aimagazine.com/news/why-did-openai-head-of-ethics-ch...\n\n\n
250 points | 323 comments
\n\n\n\nIn five years, I'd love to read a book about the history of AI ethics. I suspect it will read like Voltaire.\nMy sense is that it is radically shifting from a fluffy marketing arm to a department expected to contribute meaningfully to development and justify its impact. I would expect an ethics team to build frameworks that can help train/eval the model that the company spend millions of dollars and months training is going to be aligned to the ethical stances the company chooses. If they can't, the waste to time and money is huge if the model requires retraining for ethics reasons. That's a different job than pondering roko's basilisk or whether AI is alive.
The people in AI ethics who spent years thinking their job was marketing or publishing thinkpiece papers may be having to adapt quickly or get out of the field.\n— madrox\n
258 points | 117 comments
\n\n\n\nI feel like this language would really benefit from some sort of 1-pager overview.\nI just spent a fair bit of time on the official site, and I still don't think I have a very good grasp of what problem this language aims to solve, or why I would select it over other similar languages\n— swiftcoder\n
Go is an ideal language for AI-assisted software engineering
\n\n\n\n335 points | 395 comments
\n\n\n\nIt doesn’t matter to me how good the LLM is at writing Go if the compiler can’t stop it from accidentally leaving another part of the software with invalid state as a result of a change the LLM is making.\nWhat am I talking about? Nil and partially constructed structs are impossible to prevent the creation of in Go.
Sure, if you’ve got a small program with limited scope, that’s probably fine if you look through squinted eyes. But the teams I work with are working on sprawling, evolving software where the compiler saying “hey, that’s not a valid Widget” would be extremely useful and save much heartache.
An LLM does a good job of “checking” for other uses and “checking” if everything is going to work correctly, but - supposedly we’ve committed the concept to code so that the compiler can actually verify it - and Go intentionally permits invalid states of structs. This makes Go a fundamentally problematic language choice for the kind of software I work with teams on, LLM or not.\n— beaker52\n
284 points | 130 comments
\n\n\n\nNvidia's biggest advantage in AI has never been only their hardware performance but how entrenched their software is in ML research that flowed down stream. However, if you've actually used CUDA C/C++, it's pretty one of the worst software development ecosystem imaginable: you get all the footgun of regular C++, plus GPU compute pretending to be C++ and but doesn't actually behave like C++ because CPU and GPU compute are fundamentally different, and the only reason people put up with it is because Vulkan and HIP C/C++ are even worse.\nGoogle's limitation is that they still don't offer TPUs in a PCI-E card/dev board that people can plug in to their PC for local development and sane low level API to develop against, instead you have to go through their cloud and their full software stack which greatly limits ecosystem growth. The minute that Google figures that out, that's when Nvidia's dominance would be challenged.\n— YuechenLi\n
Apple Silicon and macOS VMs: Faster LLM Inference with llama.cpp
\n\n\n\n278 points | 43 comments
\n\n\n\nIt looks to me like this won't speed up llama.cpp for everyone, just for users running it in this particular kind of Virtualization.framework VM.\nThe fix here works around a problem where the VM was causing llama.cpp to select the wrong kernels.\n— simonw\n
244 points | 198 comments
\n\n\n\nI've been using it for the last month or so. IMO in the same way we went from tab complete -> prompts -> agents, this feels like a next step on that evolution. I highly suspect others will be following suit. I was surprised with how much it felt natural to interact with agents in this way.\nBiggest advantage is each one owns its own routines, context, and domain, and they can communicate between each other. Similar to hermes they build out their own skills, but by keeping the bots separated by domains, you end up getting better results out of them.
Additionally though each one has their own computer, which means async work feels like it actually works. I haven't had to juggle worktrees for the last month.
Biggest downsides are token expenditure. I've used more tokens this month than not this month. That's not a typo - I've used less tokens in the last 5 years prior to this month than I have this month. Always on perpetual agents use a LOT of tokens. IMO this is building for the future state where tokens are vastly cheaper, ie in a post-ASIC world.
The coolest thing I had it do for me was sourcing fabric for swag: https://image.non.io/d83664c1-5807-4a18-abe4-41928c198410.we...
I wanted to make something that didn't feel like just my logo on a shirt, so I had one of my bots reach out to ~40 fabric suppliers in vietnam, negotiate prices, lock one in, and get samples made. First samples should be finished today. It's been something I've wanted to do for ages, so it was cool seeing it actually happen. The fabric supplier bot worked with one of my prototyper bots to create a randomly generated pattern using my logo, which it then sent as a .ai file to the supplier.\n— jjcm\n
Woman pulled over twice after Flock-linked software connected her to homicide
\nhttps://local12.com/news/nation-world/flock-camera-error-lea...\n\n\n246 points | 222 comments
\n\n\n\nWe shouldn't be wasting time figuring out how to make these tools usable. A healthy society requires privacy and a healthy society leads to long term safety trends. Mass surveillance as a tool, fundamentally, even if perfectly implemented and used, is fundamentally flawed and unfixable since it destroys the long term health of society. That is why I am against Flock and similar technologies.\n— jmward01\n\n
Meta Muse Glimmer – Open weights 30B local coding model
\n\n\n\n685 points | 384 comments
\n\n\n\nI lament the comments saying this in any way redeems Meta (the company).\nThe researchers releasing this stuff have almost nothing to do with Meta other than being bankrolled by the slaughterhouse.
You aren't the customer, you are the pawn in big tech's game of thrones. Your good will is a commodity to be traded, almost literally. It will be used against you the moment it's convenient. This is open weights because Meta couldn't monetize it in any other way than to cloud developer's judgement of their reputation.
But I guess most people just don't care.
I'm glad it's open. It does not make me think any better of Meta.\n— avaer\n
Over 181,000 AI meeting recordings left wide open in note taking app
\n\n\n\n329 points | 107 comments
\n\n\n\nIt's hilarious how these companies handle security breaches.\nI once reported superadmin user/pass committed to github at a major YC backed background check company I worked at and everyone tried to make it seem like it was my fault.
I had just started working there and found it in the first week.
Anyway, had to show that it was committed by their main Staff engineer 2 years before I even worked there. For 2 years everyone's background check data in the United States that went through this thing - millions per year - thousands of Uber drivers, DoorDash, etc. all were viewable with no clearance. Anyone including overseas contractors, new hires, etc. could just login and check anyone's criminal history.
Reporting it was a disaster. They all tried to cover their asses, this huge drama and hand waving started. They tried to blame anyone and everyone. Eventually it was just AWS fault somehow (it wasn't, the Staff engineer was a dumbass, he committed it to a ruby seed file).
-----
I digress, the CTO didn't respond because he was more worried about how it would make him look. This industry is dead - the wrong people work in it.\n— purplemoonx\n
Mark Zuckerberg attacks 'closed' AI rivals as Meta returns to open models
\nhttps://archive.is/20LOJhttps://www.meta.com/thefutureisforeveryone/\n\n\n
335 points | 356 comments
\n\n\n\nI think something that doesn't get said enough is Meta did, albeit intentionally kick off the origin of the open source race back in 2023 with the release of llama.\nI'm not a big fan of meta in general, but they've done enough good, and it's possible that it was intentional as well. I don't know, I wasn't in the rooms, and I think it's worth giving them some reasonable doubt.
No one is purely good, and no one is purely evil. This is net good regardless.\n— bushido\n
The UK's war on anonymity has come to America
\n\n\n\n432 points | 322 comments
\n\n\n\n> A joint statement by Buffy Wicks, the lead author of AB 2273, Jordan Cunningham, another AB 2273 author, and 5Rights Foundation states, “The Bill is practicable and realistic, drawing as it does on the UK’s Age Appropriate Design Code (AADC).”\nHow is Buffy Wicks so damn gullible with tech legislation? Her AB 1043 and AB 1856 (Digital Age Assurance Act), intended to protect children online and regulate Big Tech (but unintentionally criminalizing open source), was sponsored by a dark money influence operation from Meta.
https://www.reddit.com/r/linux/comments/1rshc1f/i_traced_2_b...
She is actually well-respected with housing and transportation bills but I don't know what's going on with her tech bills.\n— tangotaylor\n
Mars Bar from 1991 found – and it's 20g bigger than today's
\n\n\n\n289 points | 446 comments
\n\n\n\nShrinkflation is all over. Burger buns at fast food places were more dense 20 years ago. The small burgers had 1/8th pound patties instead of the current 1/10th pound. It's hard to find a product that hasn't gotten worse or more expensive, even adjusted for inflation.\n— smolder\n\n
Show HN: Needle2: 14MB agentic LLM for phones, wearables, smart home and robots
\nHey HN,Henry from Cactus here!
We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2.
The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges 300-700 on sub-$200 phones such as the Samsung A-Series.
On the tool call and mobile device use benchmarks, Needle 2 trades wins with closest small models like LFM2.5 230M and Apple Foundation Model, at 5x to 70x smaller, both at f16 vs Needle 2 at 2bit. Needle is based on Simple Attention Networks from our paper (https://arxiv.org/abs/2607.18363).
Edge AI has lately meant Macs and PCs, but that is just 1.5 billion of over 21 billion connected IoT devices in the world today, and in emerging markets most phones ship under $200, no NPU, cheap GPUs. These include budget phones, Raspberry Pis, microcontrollers, wearables, small robots like Reachy Mini, and connected home devices.
A conventional transformer of Needle's width and depth spends 164 MFLOPs per token, and even one squeezed down to Needle's parameter count spends 87, Needle spends 70. Even on a high-end phone, an always-on assistant lives inside a power budget; every MFLOP is milliwatt-hours, and Needle spends 7x to 85x fewer of them per token than the smallest performant LLMs. More about the architecture in the link.
When we structure intelligence for consumer devices as functions with typed parameters, the only hard part is mapping a messy sentence onto them; which function, with which values. Our research found that when framed that way, the problem needs no world knowledge and no open-ended prose, which is why 45M parameters suffice.
Needle 2 expands to structured extraction where the schema can be passed in-place of tools and the model returns structured output. You can use Needle as a text-classification model with an enum field, as a summarization model by providing a schema that extracts key fields, everything but free-range decode.
Every product has its own tool vocabulary and fine-tuning needle helps it achieve frontier-level performance on custom tasks, so using the python package (https://github.com/cactus-compute/needle), Needle can be fine-tuned Needle on a Mac/PC in minutes to a few hours, with automated data-generation pipeline, just pass a couple samples.
Nonetheless, every response carries a learned confidence score based our Cactus Hybrid technique. If above your threshold, act, below it, escalate to the cloud or bigger model. Combining Needle 2 with a private DeepSeek-v4-Flash deployment works particularly well for enterprise-level tasks at barely any cost, we can help with this setup.
We have put a lot of thoughts into Needle 2 but might still be missing quite a lot, please use the playground in the provided link to test Needle and share your thoughts, always appreciated!\n\n\n
288 points | 106 comments
\n\n\n\nThis is cool. I definitely think the \"micro\" sized LLM space is underappreciated, so it's always good to see work like this. I foresee a paradigm in some contexts where you have a hierarchy of LLMs, with more competent models actively training smaller models to solve specific tasks very efficiently, and something like this could be the smallest layer in that stack.\nWith that being said, the web demo is not particularly impressive. It really doesn't like anything I throw at it. I'm fine with accepting that fine-tuning is the solution to this, but I wonder if there's anything to gain from a bigger model? I know it's completely counter to the whole point of this, but a 14MB binary using 28MB of RAM seems unnecessarily small and pretty arbitrary.
Like, what does a 28MB binary get you? Or a 140MB binary? Or a 1.4MB binary? I'm guessing the choice of 14MB came from minimizing the size as much as possible while meeting certain requirements/performance expectations, but even a Pi 5 has plenty more room to spare. Curious if there's a good explanation for this (which I may have missed in my skim of the post).\n— nater5000\n
Illinois Just Passed a Law That Puts Linux on the Hook for Age Verification
\n\n\n\n270 points | 337 comments
\n\n\n\nLinux distro founder here (stagex)\nI will never be compelled to implement this, and would never merge it.
Every release requires quorum signatures by an international maintainer team, and the distro is designed to work offline-first, with some variants not even supporting network drivers in the kernel, so Illinois legislators can eat shit.\n— lrvick\n
As AI eats the web, the internet’s collective memory is disappearing
\n\n\n\n299 points | 291 comments
\n\n\n\nAI will kill the internet because it is killing the incentive to make it. It is an industrial-strength example of why we don’t allow stealing.\n— ChiMan\n\n
H3-metal – Native MiniMax-H3 inference for Apple Silicon
\n\n\n\n195 points | 25 comments
\n\n\n\nI've been using MiniMax H3 on my M5 Pro 64GB MacBook Pro through ComfyUI. It works extremely well.\nI had to modify the default ComfyUI workflows to use a GGUF quant (city96's ComfyUI-GGUF custom node, UnetLoaderGGUF in place of the stock loader) [0].
I use the model labeled Q5_K_M. There is Q8_0 available as well, which is 34GB and fits fine in 64GB unified memory if you keep resolution modest.
The main issue is speed, a ~9-second 480x864 clip at 20 steps takes me a bit over an hour. So this will be cool to try for the speed up alone.
There's a lot of great information and workflows available to follow on the r/StableDiffusion subreddit.
[0] https://huggingface.co/Abiray/MiniMax-H3-GGUF/tree/main/unet\n— Meleagris\n
210 points | 109 comments
\n\n\n\nJust like learning Lisp will make you rethink programming languages, and learning Erlang will make you understand the true power of concurrency: learning Smalltalk will make you understand what \"object oriented\" actually means.\nI haven't used Squeak since college but I'm glad it was part of the cirriculum.
Btw, almost all of Javascript's good parts come from Smalltalk.\n— avaer\n
How I use LLMs to learn complex topics
\n\n\n\n325 points | 185 comments
\n\n\n\nI thought LLMs were a great tool for learning new topics - perhaps even complex ones. But overtime, I have had several frustrations with this. First, I get exhausted reading LLM prose. I really don't want to read anything generated by something like Opus 5 at this point. Second, as I dive deeper, I need a way to organize the information in a useful way as I begin to branch out in many different directions. I have tried to use the LLM to fix this by having it generate a web page with diagrams and organized information flow. It's an improvement, but I still run into the issues I described in my first pint - LLM prose is annoyingly dense, and the useful information gets lost in a bunch of noise. You can direct it do something like \"use plain English and avoid LLM prose - provide only as much information as necessary to demonstrate the point\", but it is once again only a marginal improvement.\nAnd then I begin to think to myself that I should just read a book on the topic written by a trusted source who put a lot of effort into teaching the topic properly and presenting the information in a thoughtful way. So, I am back to books and mostly try to use LLMs to clarify certain questions or ideas I have.\n— 8bitsout\n
287 points | 130 comments
\n\n\n\nOn the previous HN posts I got this impression:\nThe developer's astrology app containing tarot reading features was rejected from Apple's App Store that prohibits astrology apps.
He then replaced the app's content with a clone of, if not the identical open source astronomy app \"Dark Hours\", copying even the name. The original is here: https://darkhours.app
The developer was then in contact with John Gruber, who wrote an article claiming Apple's reviewers confuse a nice, scientific astronomy app with astrology. The developer reviewed the article before publication.
It then turned out that the developer forgot to mention that originally he did upload an astrology app for review. Gruber felt duped, and noted that it's unclear whether an updated build containing the new astronomy app was ever submitted to Apple for review: https://daringfireball.net/2026/08/retraction_app_store_reje...
Now the developer seems to claim that he merely accidentally cloned that open source app when working with Claude.
I am not buying it.\n— user43928\n
Windows 11's built-in Weather app wastes more than 1 GB of RAM
\n\n\n\n333 points | 275 comments
\n\n\n\nEasy workaround for this:\n1. Install uBlock Origin in Edge.
2. Start Edge, browse to MSN Weather.
3. Click the \"Add an Application\" button in the address bar to get a Start Menu icon for the page.
4. Delete the in-box Weather app icon.
Now you get the same Weather app in about 130MB of RAM, with no ads. It's not as nice as a native app, of course, but it's 1000% better than the useless ads and MSN feed that you can't block from the built-in Weather icon.
(Also, go into Widget settings and turn off \"Discover / Microsoft Start feed\". Same crap, different surface. Get rid of it.)\n— SBArbeit\n
Everything you do is being recorded
\nhttps://archive.is/JxaZ7\n\n\n183 points | 143 comments
\n\n\n\nThe original research project this company looks based on was very cool and way early: https://sandlab.cs.uchicago.edu/jammer/\n— jrexilius\n\n
Taxi drivers rarely die of Alzheimer's
\n\n\n\n262 points | 184 comments
\n\n\n\nTaxi Drivers: The mean age at death is approximately 67.8 years.\nGeneral Population: in the same dataset, life expectancy averaged 74 years.
The average age at which patients are typically diagnosed with Alzheimer's disease is 79.
People in these jobs don't live long enough on average to get diagnosed, at the same rate. The same effect will happen in any job that lowers your life expectancy.\n— cortic\n
Silicon Valley misreads science fiction and undermines democracy
\n\n\n\n233 points | 181 comments
\n\n\n\nI don’t even think most of it is deeply philosophical.\nThey just got uberrich and now worry more about the concerns of the uberrich than the concerns they had when they were normies.
Rule of law, functioning government providing services, functioning public commons, free press, freedom of speech for all, any social safety net, etc don’t matter when they are so far removed from the public they can just go buy yourself an island/citizenship/president.
Suddenly small differences in marginal tax brackets, antitrust enforcement & limits on the ability to simply buy the government they want are the biggest encroachments on freedom they feel.\n— steveBK123\n
Tom Stanton's supersonic trebuchet breaks sound barrier with gravity alone
\n\n\n\n202 points | 67 comments
\n\n\n\nHere's a direct link to the video, to avoid the blogspam cruft...\nhttps://www.youtube.com/watch?v=Co57SfcT-h0\n— sdenton4\n
229 points | 787 comments
\n\n\n\nI'm building the easiest way to start your own high-quality blog: https://github.com/BryanHogan/astro-starter-template\nAstro is a framework that uses no JavaScript by default. I also use just HTML and CSS, so no bloated additional frameworks or styling libraries.
All blog content is written as Markdown or .mdx files, so it's easy to write and move to any other tool if you wish to do so.
You can host it for free using any major provider since it's just a static website (e.g., GitHub Pages, Cloudflare, etc.).
Making it similar to my own website which is on: https://bryanhogan.com/
(Repo: https://github.com/BryanHogan/bryanhogan )\n— bryanhogan\n
224 points | 54 comments
\n\n\n\nWhat's funny to me is that Scot Hanselman, an MS employee, wrote about it for a - not so- long time ago as well:\n- https://www.hanselman.com/blog/urls-are-ui\n- https://www.hanselman.com/blog/dont-ever-break-a-url-if-you-...\nThere are people in MS who knows this is an issue. But knowledge is not equal to behaviour.\n— feldrim\n
231 points | 103 comments
\n\n\n\n> To the companies: You don’t need HackerOne anymore. The tokens to build your own in-house platform cost less than single year of HackerOne.\nYou know, the biggest thing that HackerOne delivers is a universal payments system that requires absolutely no efforts from companies. Have you tried to manually pay hackers from around the world? It is a laborious process involving trying to find what providers are compatible and what forms of money go where. It is extremely taxing to handle this. HackerOne provides real, tangible value in not making people think about how precisely to pay a hacker and in what currency. No amount of tokens solve the accounting problem, and it is foolish to imply otherwise.\n— Shank\n
“Code was never the hard part” is an insult to all programmers
\n\n\n\n524 points | 344 comments
\n\n\n\nI believe there are some programming jobs in which the code is absolutely the easier part. Not all of us work in signal processing, integrated systems or have to push upstream to Linux kernel because the company we work for really needs a memory allocation optimization for its data centers.\nNavigating customer requirements and building something that satisfies both market's needs and company strategy can be an incredibly difficult and frustrating problem to solve. Especially if you need to also oversee the execution of the strategy. So not only you have to predict what they want or know the domain deeply enough to understand what they say they want is not what they really want, you also have to come up with a plan for executing your solution in a corporate environment.
There is a reason that books like \"the staff engineer's path\" cover topics such as local maximums, communication, establishing support for executing a plan or creating alignment on big efforts. In large corporate environments with multiple international customers, code is most of the time not the hardest problem.\n— prinny_\n
Denmark Requires Oral Defenses for Students' Written Work to Counter AI Cheating
\n\n\n\n478 points | 227 comments
\n\n\n\nThis is already the case for Master's degrees and above in Denmark. I've been the opponent examining students. We ask Master's students to give US instruction on a topic (drawn randomly on a card in front of them, from a list they know a few days in advance). They do a chalk and talk on a topic for about fifteen minutes with three to five professors being \"dumb students\". It works fantastically and at the end it's very apparent to both student and teachers how well a topic is understood. Marks are rarely contested and in my experience most students think the process is fair.\nOf course, woe betide those with social phobias....\n— azalemeth\n
Fastmail offers EU data region
\n\n\n\n299 points | 135 comments
\n\n\n\nEU data regions are a reflexive action by companies that try to hold on to their EU customers (and more and more are leaving, surprisingly the larger ones seem to be leading here). Realize that as long as you are still hosted on US owned infrastructure or that if there are US (or: five-eyes) owned companies anywhere in the stack your data can still be forcibly pulled and often without you being aware that this happened. There are only very few such stacks that are 100% owned by EU entities.\n— jacquesm\n\n
A domain can now say it is for sale, in DNS
\n\n\n\n327 points | 129 comments
\n\n\n\nSomething I've wondered... if you publicly say that a domain is for sale and someone has a trademark for it would you automatically lose in arbitration?\nAround 1998 I registered a domain. Sony registered a trademark with the same name a few years after that. Someone on a Gmail account asked if he could buy it - I later found out he worked for Sony. I told him no - it's for a game I've been working on. I went on a vacation for a month and when I got back there was a fedex package filled with documents from Sony saying I'm violating their trademark and they'll take the domain.
I got a lawyer and he told me that I shouldn't offer it for sale as that would show the arbitration board that I don't need the domain. But he also told me that in order to fight the trademark it would probably cost a couple of hundred thousand dollars. So I could keep the domain but not use it for commerce...
In the end I ended up selling it to Sony but through my lawyer - I never stated it was for sale. But this was early on for domains and I wonder what the process is like now.\n— comrade1234\n
Timeline of the OpenAI accidental attack against Hugging Face
\n\n\n\n312 points | 319 comments
\n\n\n\nNorbert Wiener in 1960:\n\"As is now generally admitted, over a limited range of operation, machines act far more rapidly than human beings and are far more precise in performing the details of their operations. This being the case, even when machines do not in any way transcend man's intelligence, they very well may, and often do, transcend man in the performance of tasks. An intelligent understanding of their mode of performance may be delayed until long after the task which they have been set has been completed. This means that though machines are theoretically subject to human criticism, such criticism may be ineffective until long after it is relevant. To be effective in warding off disastrous consequences, our understanding of our man-made machines should in general develop _pari passu_ with the performance of the machine. By the very slowness of our human actions, our effective control of our machines may be nullified. By the time we are able to react to information conveyed by our senses and stop the car we are driving, it may already have run head on into a wall.\"
\"In neurophysiological language, ataxia can be quite as much of a deprivation as paralysis. A patient with locomotor ataxia may not suffer from any defect of his muscles or motor nerves, but if his muscles and tendons and organs do not tell him exactly what position he is in, and whether the tensions to which his organs are subjected will or will not lead to his falling, he will be unable to stand up. Similarly, when a machine constructed by us is capable of operating on its incoming data at a pace which we cannot keep, we may not know, until too late, when to turn it off.\"
Source: https://www.cs.umd.edu/users/gasarch/BLOGPAPERS/moral.pdf\n— RGS1811\n
256 points | 92 comments
\n\n\n\nSide comment about the difference between “My server is a phone now” and “My phone is a server now.”\nIn linguistics, the terms “theme” and “rheme” refer to the status of information in a sentence [1]. The theme of a sentence is something that is known or can be assumed from the context (that is, old information), while the rheme is information that is newly introduced in the sentence, often as a comment on the theme. The tendency in many (most? all?) languages is for the theme to precede the rheme in connected discourse.
Without context, a reader is therefore likely to assume from “my X is a Y now” that X is something the writer had previously and that Y is something new. Because, these days, most people have a phone but few have their own servers, I was initially confused by the title “My server is a phone now.” I had to read the article to realize that the word order is in fact appropriate: for the writer, the server is old information (something he has had for a while) while the phone (as a server, at least) is new.
In English, theme and rheme might be inferred not only from the position in a sentence but also from how a noun is marked. Possessives like “my” are generally used before nouns referring to things that are previously known or assumed; the definite article “the” is similar. In contrast, the indefinite article “a” (or “an”) usually marks information being newly introduced, so it is more common with rhemes.
I worked as a Japanese-to-English translator for many years. Because the grammatical structure of Japanese sentences is very different from that of English, sentence-by-sentence translation often ends up with paragraphs where the information flow is awkward and unnatural. I gradually learned that I could produce smoother translations by reorganizing my English sentences into an old-to-new information sequence.
[1] https://en.wikipedia.org/wiki/Topic_and_comment\n— tkgally\n
US Military's cyber command unit grapples with cluster of deaths by suicide
\nhttps://archive.ph/W2tpE\n\n\n215 points | 355 comments
\n\n\n\nI suspect the Cold War of cyber warfare that is happening is orders of magnitude bigger than the public knows.\nImagine knowing this and not being able to get emotional support from your friends and family.\n— YesThatTom2\n
Amazon Is Creating the Biggest Pollution Source in the Country
\n\n\n\n220 points | 145 comments
\n\n\n\nLet's add as well \"SpaceX’s Terafab will rely on natural gas power plants, not Tesla solar panels\" to the mix :o(\nhttps://techcrunch.com/2026/08/07/spacexs-terafab-will-rely-...\n— zepearl\n
New Amazon Data Center Is Set to Have the Most Polluting Power Plant in the U.S.
\n\n\n\n216 points | 276 comments
\n\n\n\nThose in power seem to think that a compute advantage is necessary to maintain economic advantage. Despite this, the voters universally hate data centers, so much so that both NY and TX have data center construction moratoriums.\nDespite this, those in power refuse any environmentally aware energy policy nor water policy. Despite this, companies have been trying to be environmentally aware for either moral or marketing purposes to get social permission to build DCs. Despite this, companies are “forced” to build their own “dirty” power sources since political power is disrupting green energy build out. Despite this, demand is still high so that compute costs are continuing to rise. Despite this, our economy is increasingly predicated on a few companies using more compute than geopolitical rivals with cheaper electricity. Despite this, those in power are, again, unwilling to create a compromising policy (aka clean energy) that voters actually would demand before tolerating more DCs.
What’s going to lose first? Voters/environment, politicians/the oil lobby, big tech, or the overall american economy?\n— vineyardmike\n
US strikes $1.2B deal to pay German firm to halt offshore wind projects
\n\n\n\n777 points | 744 comments
\n\n\n\n> Interior Secretary Doug Burgum said in a statement posted on X that Americans deserve an energy system built on common sense and not one dependent on \"costly subsidies\".\nI agree. This is why we are building offshore wind and not subsidizing fossil fuels, right?\n— saagarjha\n
What happens if an entire class of workers loses faith in their careers
\n\n\n\n359 points | 495 comments
\n\n\n\nAddressing the question in the title, \"What happens if an entire class of workers loses faith in their careers\", look at what happened to printers. That was a good skilled trade for centuries. Then it just went away. Phototypesetting took the high volume end. Desktop publishing took the low end. Computer-controlled presses took press operators. And then the newspaper business died.\nI've heard homeless former printers talking about the good days at the Burger King near the cable car turntable at Powell and Market. They thought they had careers.\nNow they're struggling to survive.
Knitting is not the answer. This is a survival issue. Somehow I suspect the author inherited money.\n— Animats\n
404 points | 243 comments
\n\n\n\nI've been using it extensively since the release and the best summary I can give is that it's good enough to use it for (almost) everything and cheap enough that the cost are irrelevant. I'm running it in Oh My Pi with a second instance running as \"advisor\" and even with 5-6 active sessions (effectively 12 streams) I'm struggling to spend more than 5 bucks per day.\nOpenCode Go even has double limits temporarily so for 10 USD you effectively get 140 USD of tokens to spend. It would impress me if someone could burn that amount with \"normal\" usage. Even when running multiple sessions.
I have a Claude Max subscription but I've barely touched it, it just feels like a step back to have to think about limits and usage even though the models are stronger.
The beauty of intelligence at this cost (even if it's not SOTA) is that it opens a whole bunch of new use cases. Test failure in CI? Have the bot automatically propose a fix, its cheap enough that you can discard it w/h issues. Test coverage too low? Auto generate tests on CI for every pull-requests! Monitoring server logs, continuous security audits and investigating every received exception now becomes possible.
I'm thinking about having it automatically filter and re-rank my social media feeds so I can steer the algorithm instead of the other way around.
Perhaps other people (with enormous budgets) were already doing all of the above but for us this is a really exciting release!\n— LaurensBER\n
Oracle bans AI-generated code from OpenJDK
\n\n\n\n356 points | 243 comments
\n\n\n\nOracle, the law firm with a tech business attached, probably wants to retain the option to sue other people for AI-washing their proprietary code, and that doesn't work if they're also publicly accepting AI contributions to their code with no apparent concern for the provenance of it. The tech business would probably tell them this isn't going to be a needle they are going to be able to successfully thread, but the law firm is in the driver's seat.\n— jerf\n\n
A year of fighting scrapers on my 1.5 million-page website
\n\n\n\n358 points | 340 comments
\n\n\n\nThe worrying thing here is that so many people accepted outsourcing the decision on who can see their website to a large company (Cloudflare). If the company decides that a certain user should not see the website, the user will not see the website, and no one will know about it, and the user will have no recourse. That is not the open web that I would like to see.\nA second side effect of a knee-jerk reaction to bots crawling websites is that if you try to fight all bots, you also end up hurting real users that use \"bots\". If I run a local script or an LLM that needs to fetch a web page from your site and you block that script or LLM as being a bot, you hurt me, the user. I wanted the information and I have no way of getting at it without investing my own time and effort in going to your website. That might or might not be what you expected, but it's worth taking into account.
And finally, something worth noting is that there are so many websites whose owners complain about bots, but the real problem is that the website is poorly built and should be improved anyway. Bot traffic is not necessarily bad.\n— jwr\n
226 points | 48 comments
\n\n\n\nRelated, and linked in the readme: https://github.com/xoreaxeaxeax/smiiiiiiiiiiiiiiii (using the slow instructions to break SMI)\n— Retr0id\n\n
U.S. economy lost 23,000 jobs in July, a sudden reversal
\n\n\n\n314 points | 305 comments
\n\n\n\nAs a data scientist it is very frustrating to see a consistent lack of error bars on these numbers or any discussion about the magnitude of uncertainty around them by all news reports every time these numbers get released\n— astro1234\n\n
App Store Rejection of the Week: Dark Hours
\n\n\n\n254 points | 151 comments
\n\n\n\n> But that’s not what happened at all. Godier proceeded through a series of escalations up to the App Review Board and the Review Board responded that they determined the original rejection was valid because, I shit you not, “We understand that the app includes a live tarot reading feature.”\n> It has no tarot function, no horoscopes, and nothing that I, or anyone else I’ve asked, would associate with astrology.
I have dealt with the Byzantine App Store bureaucracy before. What you do here, is you submit a new version, pretty much identical to the previous one, with the comment “tarot function removed”.\n— DrJokepu\n
The Nixpkgs core team has disbanded
\n\n\n\n261 points | 114 comments
\n\n\n\nFor anyone reading this and jumping to a broader conclusion: the Nixpkgs core team disbanding does not mean Nixpkgs or Nix is dying.\nIt does mean that this particular structure was not sustainable, very important contributors burnt out and we need to do better, faster. We need to continue learning from this and continue building a stronger ecosystem that prioritizes the contributors who are the only reason any of this is possible.\nPersonally, I'm sorry and grateful. Sorry that it ended in core folks being burnt out. Grateful since they did some of the most amazing work, more than anyone can imagine two people doing.\n— ronef\n
Making Postgres 300x faster for analytics: batching, operator fusion, and SIMD
\n\n\n\n229 points | 107 comments
\n\n\n\nAuthor here. Let me know if you have any questions about the post or about pgrust.\nLet me take a shot at answering what I think will be the most common question: how can I trust pgrust? Our #1 priority right now is correctness. Over the past two weeks, I've done a mix of formal verification and differential fuzz testing. We've been able to prove over 1000 user facing functions have the exact same logic in both pgrust and postgres (see the proofs directory if you're curious). For cases where formal verification is not easy, we've taken the c implementation of a function and the rust implementation of a function and ran millions of inputs through each of them and confirmed they gave the same results every time.
We've only covered about 15% of the surface area so far, but in the process, we've discovered ~100 bugs in pgrust and ~20 bugs in Postgres itself. My favorite postgres bug we found is this one[0]. Postgres has a quadtree implementation. Due to floating point rounding, it was possible for a point to be neither above, nor below, nor even with the center point of the quadtree.
We've also entered engagements with Antithesis[1] to do Jepsen style fault testing and Aretta[2] to do more serious formal verification.
If you want to support the project, the easiest way is to give us a star on GitHub[3]
[0] https://www.postgresql.org/message-id/19597-39c532e61d78dff6...
[3] https://github.com/malisper/pgrust\n— malisper\n
503 points | 101 comments
\n\n\n\nThis is a really important concept for developers.\nOne aspect for developers that I see quite often is the assertion that \"We can't have X without giving up more Y\", most commonly \"we can't have more security without giving up on user experience\". With the Pareto idea in mind, we can see that that statement is true, if and only if you are in fact on the Pareto frontier of security and user experience already.
However, many times these confident pronouncements are being made when the system under question is quite evidently not on the Pareto frontier in the first place and indeed you can get more of one without giving up the other.
Making it more tricky is that in business, you can never discard \"money\" as a dimension, so unless you're taking \"money\" as one of the dimensions in the original comparison you want to do, it sneaks its way in. Or, a composite time/money, or \"cost to business\", or some other similar concept, time & money aren't orthogonal and don't need to be treated as two separate dimensions in general, though you can if you want. Which puts you into the 3D case, and as the page says, that grows the frontier quite a lot, which is good in some ways and bad in others. Nevertheless, in my opinion it is still often the case that even in that space we are often making \"tradeoffs\" without checking that we're on the Pareto frontier in the first place.\n— jerf\n
AMD acquires Taalas to boost inference performance by etching models in silicon
\nhttps://ir.amd.com/news-events/press-releases/detail/1296/am...https://chatjimmy.ai/\n\n\n
295 points | 230 comments
\n\n\n\nWhat I like about this, is that it significantly increases the probability of a sci-fi scenario where you're picking up a hot chip on the black market; rumor has it, Mythos 9 weights baked in...\n— mNovak\n\n
Qwen3.8 Max now ranked as the best overall model by agentic index
\n\n\n\n399 points | 258 comments
\n\n\n\nChina has caught up is the main takeaway here. The SOTA models are so close that it's really hard to compare them intelligence wise - you have to get a feel for them yourself and what works for you.\nWhat I'm really excited for is the 27B model. 3.6 is still the king of local, and if 3.8 makes the same improvements it could really legitimately make local viable as a default. I'd love to run a perpetual agent on 3.8 that's locally driven.\n— jjcm\n
348 points | 263 comments
\n\n\n\nOn the topic of taste, I think of this quote in Susan Sontag's \"Notes On 'Camp'\" often:\n“For taste governs every free — as opposed to rote — human response. Nothing is more decisive. There is taste in people, visual taste, taste in emotion — and there is taste in acts, taste in morality. Intelligence, as well, is really a kind of taste: taste in ideas. (One of the facts to be reckoned with is that taste tends to develop very unevenly. It's rare that the same person has good visual taste and good taste in people and taste in ideas.)
Taste has no system and no proofs. But there is something like a logic of taste: the consistent sensibility which underlies and gives rise to a certain taste. A sensibility is almost, but not quite, ineffable. Any sensibility which can be crammed into the mold of a system, or handled with the rough tools of proof, is no longer a sensibility at all. It has hardened into an idea...\"\n— hellojomp\n
GitHub Actions and Pages are experiencing degraded availability
\nhttps://www.githubstatus.com/uptime/br0l2tvcx85d\n\n\n300 points | 254 comments
\n\n\n\nAll of the outages make sense to me as scaling issues. Each month, GitHub is getting the amount of commits they’d normally get in a year. And it’s growing\n> Yup, platform activity is surging. There were 1 billion commits in 2025. Now, it's 275 million per week, on pace for 14 billion this year if growth remains linear (spoiler: it won't.)\nGitHub Actions has grown from 500M minutes/week in 2023 to 1B minutes/week in 2025, and now 2.1B minutes so far this week.\nSo we're pushing incredibly hard on more CPUs, scaling services, and strengthening GitHub’s core features.\nAnd as a fine purveyor of hand-crafted shit code for many years, I'm not gonna weigh in on that.
x.com/kdaigle/status/2040164759836778878\n— __initbrian__\n
Almost no skill required to cook a steak
\n\n\n\n259 points | 300 comments
\n\n\n\nI take the point, but I think the author picked a poor analogy. Cooking even an excellent steak is actually not that hard. In fact, I'd argue that it's among the easiest things to master/make at a top level quality at home. Does it require some modicum of attention and understanding? Sure. But starting with a high quality cut, owning a meat-thermometer, and knowing about reverse searing is about all it takes to reliably and easily get a near perfect steak every time.\nThere are far, far better cooking examples out there.\n— MostlyStable\n
Quake – 30th Anniversary Update
\n\n\n\n270 points | 140 comments
\n\n\n\nQuake is 30 years old. I can't believe I'm still alive. I love this game so dearly. The mods were just incredible, KQP and Girobot in particular.\nThose were the for real LAN party days which I'll never get back. As in I had friends bringing over their big ass CRT monitors and PCs for a single occasion and also had ludicrously long network cables. My parents didn't mind, they somehow understood, even with LAN wires running across the kitchen floor. Priceless.
Edit: typo: lab -> LAN\n— karim79\n
Humans missed 1 in 3 threats approving AI agent commands across 40k game runs
\n\n\n\n239 points | 186 comments
\n\n\n\nI remember when this game was posted here, and there was a lot of discussion at the time that some of the prompts were misleading about whether or not they were risky, some people were debating about how some of the prompts flagged as bad weren’t bad, and others flagged as not bad were. This is a fundamental flaw in the test, that makes the analysis of results meaningless.\nAlso the game was on a timer, and maybe there are some very abusive workplaces where you feel that kind of pressure, but I think most of us actually take the time to understand what a being asked before approving it.\n— VladVladikoff\n
New Mexico court orders Meta to pay $567m over harms to children’s mental health
\nhttps://www.reuters.com/world/new-mexico-court-orders-meta-p...https://www.theguardian.com/technology/2026/aug/06/new-mexic...
https://www.kob.com/new-mexico/nm-court-orders-meta-to-pay-5...
https://www.bbc.com/news/articles/cd7lz3wr2rlo
https://www.wsj.com/tech/meta-ordered-to-pay-942-million-to-...\n\n\n
278 points | 178 comments
\n\n\n\nA lot of people on this story are making the usual snarky remarks that, oh, this is just a slap on the wrist—some trivial proportion of Meta’s global revenue. And that is true, but unlike the EU, for example, New Mexico is a tiny jurisdiction. It only has a little bit more than 2 million people.\nSo actually, a fine—or a judgment—of $942 million that applies just to New Mexico is enormous. If you look at Meta’s revenue from the U.S. and Canada and assign it proportionally by population, you find that Meta only earned about $1.5 billion to $2 billion total from New Mexico over the past five years. The exact figure depends on whether you allocate purely by population or account for purchasing power, which makes the estimate a little lower because New Mexico is a poorer state than the U.S. average.
Given that, the judgment is a half to two-thirds of what Meta could have obtained from New Mexico. So it is actually a serious deterrent, assuming Meta thinks New Mexico would continue to impose fines of that size.
I also think that there’s a good chance this judgment gets overturned or significantly reduced on appeal, but even if the fine drops to half its current level, it would be proportionately larger than almost any fine against any company, let alone tech, that I’ve heard of.\n— tmp10423288442\n
Show HN: I spent 2 years designing a mechanical Magic Keyboard
\nHi HN, today I’m revealing Altar II, an ultra-low profile (4.75mm) mechanical keyboard designed to replace your Magic Keyboard. It’s the keyboard I wish Apple made.Altar II has fully mechanical switches, plus a magnetically detachable dial, haptic feedback, a red backlight, and a native macOS companion app.
I’m a heavy Mac user and I feel that Apple hasn’t given its keyboards the attention they deserve, in terms of hardware, software, and features, so I set out to try and change that.
It was a massive undertaking and it has taken almost 2 years to get to this stage. One of the main obstacles was fitting a reasonably sized battery into the chassis, and aggressive power profiling without messing with the user experience. Another issue was we couldn’t fit any components on the underside of the PCB because there was literally no space to do so (and there are a lot of components).
Around 4 years ago I posted my first keyboard to HN (https://news.ycombinator.com/item?id=32507172). Its response blew me away and ultimately led to me quitting my job and making and selling keyboards full time. I’ve taken some of the feedback from that original post and incorporated it into Altar II (e.g. the detachable dial).
I’ll be around to answer any questions. Please let me know what you think, I’d love to get some feedback.\n\n\n
255 points | 212 comments
\n\n\n\nTo everyone who is about to pile on regarding the lack of a fingerprint sensor: you know it's not possible, we all know it's not possible, the creator of this project knows it's not possible.\nWhat are you adding to the conversation when you complain about it? So it's a dealbreaker for you, don't buy it.
To @aemerson_: Congratulations on taking a risk. On first look this is great work. The attention to detail on both the keyboard and the website is really impressive.\n— ary\n
528 points | 326 comments
\n\n\n\nFrom Jeff's twitter post:\n> Our general approach is to automate the experimental loop. We think this approach is broadly applicable across many different fields of science and engineering. We’ll initially focus on ML research and engineering, but believe the approach can help with important subproblems in nearly every one of the fourteen <at>NAE Grand Challenge problems. We think doing this well requires strong expertise in machine learning as well as large-scale systems.
See also: https://www.nae.edu/20782/grand-challenges-project
Those 14 are:
NAE Grand Challenges for Engineering
1. Make Solar Energy Economical
2. Provide Energy from Fusion
3. Develop Carbon Sequestration Methods
4. Manage the Nitrogen Cycle
5. Provide Access to Clean Water
6. Restore and Improve Urban Infrastructure
7. Advance Health Informatics
8. Engineer Better Medicines
9. Reverse Engineer the Brain
10. Prevent Nuclear Terror
11. Secure Cyberspace
12. Enhance Virtual Reality
13. Advance Personalized Learning
14. Engineer the Tools of Scientific Discovery\n— cjbarber\n
Changes at Google DeepMind: Demis Hassabis from CEO to Chair, Jeff Dean departs
\nhttps://www.axios.com/2026/08/05/google-deepmind-demis-hassa...https://www.reuters.com/business/google-shakes-up-ai-leaders...
https://www.discoveryloop.com/, https://news.ycombinator.com/item?id=49184960
https://twitter.com/JeffDean/status/2085034604172603724, https://xcancel.com/JeffDean/status/2085034604172603724\n\n\n
418 points | 544 comments
\n\n\n\nWow. Jeff and Sanjay both departing. Truly end of a golden era.\nThere is an entire cohort of work-optional very senior engineers for whom one of the last reasons for hanging on was \"at least Jeff and Sanjay are around\".\n— gandalfgeek\n
Cloudflare OS: an open platform for agents, apps, and work
\n\n\n\n282 points | 158 comments
\n\n\n\nI liked Kenton's take on this: https://x.com/KentonVarda/status/2084990137180590572?s=20\nText from tweet:
Today we are releasing Cloudflare OS, a chatbot with connectors, just like every other tech company is doing.
Except actually, it's different. This is a remake of Sandstorm[.]io, my startup from 10 years ago, except this time built on Cloudflare Workers (the platform I've spent the last 9 years building) and deeply leveraging AI. This is more or less the culmination of my secret 10-year master plan.
This is a full-on personal app vibe coding platform, in which the sandbox is so secure that you can pretty much go wild -- the AI cannot introduce a significant security bug. We believe a company's security team can feel comfortable giving non-technical users permission to vibe code and then sleep soundly at night.
How is that possible? It's the Sandstorm security model, revisited. A \"Gadget\" is the same thing as a Sandstorm \"Grain\": a fine-grained app instance. For example, if you have a document editor app, each document runs as a separate instance of the app, in a separate sandbox (one \"Gadget\").
This means two things, both of which I think are Big Deals:\n1. The platform can manage all access control, by controlling who can access the Gadget at all. There is no way the Gadget can accidentally leak itself to an attacker -- even an attacker who has access to other Gadgets based on the same app.\n2. Since everyone is running their own copy of the code, everyone can freely modify their copy of the code.
Think about #2 a bit more.
What if, when you wanted a new feature in the software you are using, you could just prompt your agent to add it?
This doesn't work in the cloud Software-as-a-Service model, because you are not running your own copy of the app.
Sandstorm tried to change that 10 years ago, but the world wasn't ready, because not enough people had the skills or patience to actually modify their software. AI has changed that. Now you just ask the agent -- the same agent that you are using to help you interact with the Gadget can also modify the code of the Gadget.
And it is so fun.\n— rozenmd\n
Civilian plane crash in New Mexico tied to military GPS blocking
\n\n\n\n330 points | 163 comments
\n\n\n\nWhen I used to fly in Florida in the 2005-2015ish timeframe, NOTAMs for GPS interference (it didn't say that exactly but that's what it meant) were an everyday occurrence along my path, it became boilerplate. I experienced it a few times. One moment the GPS thinks you're plodding along next to Ocala, next moment it thinks I've teleported a thousand miles away, and then it'd just quit.\nBut that's what we train for. That's why aircraft and pilots have redundancy to work around any single possible point of failure and why you practice it until it becomes natural. Because of the NOTAM I'd already have a VOR tuned in, and once the GPS went inop I just fine tuned it to get a radial fix on the VOR. Also tried to never lose my bearings with my MK I Eyeballs either. No big deal.
I feel bad for everyone that lost their lives here, but this is why competence is critical. There's a litany of errors the pilots made. And the article tries to pin a bit on the controller, but no, ultimate responsibility is on the pilot in command, FAR/AIM is clear on that.
If these allegedly instrument rated pilots had been properly drilled by their CFI's and properly grilled during reviews and such this should've been an easy flight.
Also, \"just ask Kennedy\". Wired bringing that into the conversation is perhaps relevant but not in the way I think they intended. JFK Jr lacked basic airmanship skills. His aircraft was 100% functional until the moment it impacted the ocean.
Some professions simply can't be go-along-to-get-along social clubs. Safety culture centered on competence and performance is key.\n— mrngld\n
267 points | 135 comments
\n\n\n\nZed should focus on basics. When it focuses on basics Zed is good.\nhttps://github.com/zed-industries/zed/discussions/54150 failing to show newly created files and declining to provide a refresh button, instead adding a polling backend, breaks Zed on WSL.
Why a new version control system? Why not git, jj, or another existing system?\n— NoDodgeQuestion\n
Demis Hassabis is moving from CEO to Chairman at Google DeepMind
\n\n\n\n369 points | 1 comments
\n\n\n\nComments moved to https://news.ycombinator.com/item?id=49184755.\n— tomhow\n\n
I'm switching my phone from Android to Linux
\n\n\n\n281 points | 263 comments
\n\n\n\nThere are too few apps on Linux phones, otherwise I would have migrated there\n— nc55g3g\n\n
Beating GPT-5.6 Sol on retrieval with 100x cheaper open models
\n\n\n\n284 points | 73 comments
\n\n\n\nI (and I imagine many others) would love to use something like this, but can't, because my data is too sensitive to be uploaded to a cloud of which I have no gaurantees of privacy/security.\nIs there any way we do this using rented GPUs and open-source software stacks? Paying for the service isn't the issue, I don't care if it's free or if a cut is taken in some capacity, I just don't want the provider to have access to my data.\n— nullbio\n
Cops Used Flock to Track a Man Across State Lines for a Pretextual Weed Search
\n\n\n\n299 points | 183 comments
\n\n\n\nThis is the big problem with mass surveillance as a whole. With enough time, they'll find some pretext to bring you in, even if the original crime they want you for is weak.\nThis will ruin lives, even if they're eventually let go. Jobs lost, evicted from homes, cars towed, pets surrendered and put down.
Lives will be completely derailed, police and judges will not be held accountable. The wrongful arrestees will still be in debt to the local jail and court for various fees they racked up. Trust in the justice system will continue to degrade and more and more violence will be used against people to maintain order.
We hold police to high standards for search/arrest because they have the monopoly on violence. Even if they're wrong, you can't shoot back. Even if they're slamming your loved ones into the pavement, breaking their teeth, you cannot kick the cop in the head or else they have permission to murder you both.
Mass surveillance is a loophole to this contract.\n— MSFT_Edging\n
Meta Ran Ads That Contained AI-Generated Child Sexual Abuse Imagery
\n\n\n\n286 points | 216 comments
\n\n\n\nhttps://archive.ph/4LhoU\nI didn't realize the original was behind a paywall.\n— malshe\n
In Memory of My Wife, Elise Cawley, with Thanks for 36 Wonderful Years
\n\n\n\n789 points | 44 comments
\n\n\n\nAn amazingly detailed tribute, you'd have to think this was taken from a journal, in which case it's impressive to keep such a record of their life, or maybe Stephen just has a fantastic memory.\nEven through that detail you can really feel how much he loved and admired her.
> Somehow in the shock of what has happened it feels as if I just met Elise, and now she is gone.
A reminder that time's cruelty is making the wonderful feel shorter than the painful.
> My only solace in this tragedy is that the end came instantly, after a day filled with nothing but joy.
Hopefully we're all lucky enough to be allowed to rest without suffering while surrounded by loved ones.\n— cube00\n
AI-Generated Images Discourage Me from Reading Your Blog
\n\n\n\n620 points | 369 comments
\n\n\n\nTo me there are different motivations to include imagery and some of them are just fine with AI generation. If the imagery actually goes to the substance of helping to illustrate or explain the actual content then it seems very superficial to reject it based on how it was generated. In fact it is kind of stupid to insist someone slavishly draw something by hand that could have been done an easier way.\nThe problem comes when the imagery is placed as a false value signal. If it is designed to send a signal of \"this content took a lot of effort to produce because an image that would be hard for a human to make is attached to it\" - but the image was generated without that effort, then it's basically a lie and insult to the reader.\n— zmmmmm\n
Xbox goes down. You can't play games you own on disc
\n\n\n\n389 points | 439 comments
\n\n\n\nI wanted to show my gf the halo 1 campaign, so I downloaded the 30GB Masterchief Collection via steam. When I launched it, I was greeted with a black box. After 1 Minute it was a microsoft login screen that looked eerily like a browser inside the game. Since I couldn't do anything else, the game was stuck at a sub 720p resolution. I said fine Ill just create an account. Yea so mail, firstname, lastname, birthday, country, email and email verification code later I had to do captcha which is clicking and holding down a button. \"Help us against the bots\" it said. I unironically spent 5 minutes clicking that button, then being stuck in a loop with \"please try again\" while my gf just sat there watching me.\nWell I turned off MCC and gave up.
I don't want to know what life is like having xbox hardware\n— mawadev\n
Show HN: Simple algorithm and color space to generate diverse skin tones
\nHello HN!I found that picking out plausible but diverse skin tones for my digital art and game development projects was kind of difficult, and I got curious about if there was a way to define a color space that made it easy.
I've built a color picker and procedural generation algorithm based on the space as well as a bunch of other fun js features and demos throughout the page that use the equations. If you find it interesting, I have lots of explanations of how I built it and what properties the space has.
The methodology might be a bit shaky, but hopefully the result is as helpful for others as I have found it! There's lots of room for improvements (see the Future Work section), but I'm really happy with how it turned out.\n\n\n
448 points | 87 comments
\n\n\n\nThis is beautiful work! First, I thought this would involve using PCA to go from 3D to 2D, which would result in an easier selector to use, but at the expense of representing every person.\nThen, I thought this would stop at using the U-space vectors (the ones that form the basis of the PCA image) and their corresponding ellipse to form our color space, but no, the function fitting is a very slick idea, even if it was executed by hand.
Lastly, I love the presentation of sampling from different r values. Whether you sample from a fixed r value or a range of r values, I bet this has great applications in game design or animation.
I still don't quite get the manual data labeling process at the beginning? It seems like it would encode some bias, but the consistency of the first point cloud and the results certainly speak for themselves.\n— vermarish\n
Ray Bradbury's \"There Will Come Soft Rains\" is set today (2026-08-04)
\n\n\n\n436 points | 197 comments
\n\n\n\nI imagined one of those automatic lawn mowers on a very sunny day, going about its own business cutting the grass. Until it arrived on the non-shielded side, and it kinda sags a bit and keels over into the black-charred ground on that side. A clean shadow between the un-touched grass, and the ground in directly in the path of the blast. Or maybe the charred ground is for later. We need to see the nice, automated kitchen first. And only then the rest of the world. Would be a very interesting short film, I think.\n— kebman\n\n
Mistral's Shieldstral: 3B open-weights model for multimodal moderation
\n\n\n\n284 points | 69 comments
\n\n\n\nI would be curious if this can do moderation with an arbitrary ruleset, or if it's just \"that one moderation style\" we already know from current big tech platforms.\nThe kind where malicious intent is okay if the words are nice.
___
Or, rephrased: How big is the space in which you can tune this model without retraining.
Is it just \"we hate sex\"/\"we don't hate sex\" \"We hate violence\"/\"we don't hate violence\" or is it _truly_ as flexible as claimed?
__
Maybe something like \"Is this guy a corporate fraud that is going to waste my time with performative nonsense?\"
That would be the true test for a moderation model and I would be immensely impressed if it could manage to pull that off.
___
Edit: \nLooking at the paper though.. probably not.
I suppose this is useful for B2B, which seems to be mistrals whole thing. Question is just if it is also useful for society to hand the SV prefab morals down like that. Kinda like cultural imperialism but with an ethical spin.
Maybe opinions on those base datasets could occasionally differ more than the model can be steered.\n— hypfer\n
All of Winona Police Department's Flock cameras cut down and stolen
\n\n\n\n342 points | 205 comments
\n\n\n\n> Police use the technology to investigate crimes such as hit-and-runs and to locate missing persons.\n> No suspects have been identified.\nYeah, that tracks.
> Camera data is owned by the police department and permanently deleted after 30 days.
I assume this means that the police department deletes one local copy?\n— jeremyberemy\n
DeepSeek V4 Flash on a Single AMD MI300X
\n\n\n\n275 points | 61 comments
\n\n\n\nI don't think you can buy a single \"MI300X\" unit, right? Only the box with x8 of these at a cost of ~250K EUR.\n— majke\n\n
Pi's Minimalism Is Its Advantage
\n\n\n\n369 points | 146 comments
\n\n\n\nLots of praise for Pi in this thread, so I'll offer up a diverging opinion. Given all the hype, I was a bit underwhelmed by Pi. It definitely has some good ideas around customization, but it annoyed me in many little ways.\nFor a program that's minimal it sure takes a long time to start up, the standard C-p and C-n bindings don't work, it doesn't follow the XDG Base Directory Specification and just pollutes my $HOME directory.
I think there's still space for another harness that's 1/ open-source, 2/ written in a fast compiled language (Rust, Go, etc.) and scriptable in a simple (aka non-JS) scripting language (Lua, etc.), 3/ less opinionated and more sensible so things like XDG isn't a WONTFIX.\n— paldepind2\n
U.S. used 'virtually all' of its long-range precision missiles during Iran war
\n\n\n\n335 points | 600 comments
\n\n\n\n>The U.S. Army has used up much of its stockpile of highly accurate long-range missiles during its five-month war with Iran, according to three people familiar with the data, raising concerns about the military’s readiness for future conflicts.\n>The missiles are principally the Army’s surface-to-surface weapons, known as Army Tactical Missile Systems (ATACMS) and Precision Strike Missiles (PrSM). The U.S. has used “virtually all” of these weapons, according to two of the sources.
That quite a bit less of an issue than the headline implies.\n— laughing_man\n
316 points | 135 comments
\n\n\n\nA good moment to remind everyone that if we took the promise for granted, that AI will in fact prevail and prompting is the one skill that will rule them all... we'll lose all domain experts in one generation.\nIt's less of \"signaling expertise\" and more about actually having said \"expertise\".
In my experience with LLMs it's not uncommon to be having a deep conversation about making pasta, only to be told, after asking for a sample recipe, to get a bucket of paint and a bag of concrete. Of course these hallucinations are way more subtle and easy to miss for someone who doesn't have deep domain knowledge.\n— zeroq\n
Critical CVE issued for hallucinated SQLite vulnerability
\n\n\n\n588 points | 215 comments
\n\n\n\nWe can chalk this up as another example of over-exhuberance by what folks believe LLMs can accomplish vs. what they actually are.\nLLM-based “AI” is able to use its vast corpus of inputs and calculate the most statistically likely output in a given situation. It is probabilistic, and when you are dealing with probabilities in a situation where certainties, not probabilities, matter, you’re going to get dinged on credibility massively when your LLM-based “AI” gets the probabilities wrong at best, or in this case, claims a line of code generates a vulnerability when it is, in fact, a code comment.
LLMs are text-prediction engines. They are not Artificial Intelligence, and shouldn’t not be treated in any form or fashion as if they possess intelligence. What bothers me about this entire situation is that presumably the folks that relied on the LLM-based “AI” to generate these vulnerabilities knew (or should have known) enough about their tool to know this would happen, but did not.
Now, we all pay the consequence, to the tune of hundreds of thousands if not millions of dollars of wasted productivity from teams that have to deal with the resulting fall-out of this usage of “AI”.
A human must verify everything an LLM presents as fact. Everything. If you don’t, we all pay the price. LLMs do not remove the onus of responsibility on the human being, if anything they amplify it because LLMs can generate lots more output more quickly that needs to be verified than humans can.\n— gortok\n
477 points | 173 comments
\n\n\n\nOne of the arguments for open source software for end-users has always been the freedom to examine and modify how that software works.\nThe reality for most people - even expert programmers - has been that the freedom is more about being able to lean on other people to do that. Most people can't justify the time commitment needed to read and then modify the code for tools they use very often.
I think LLMs have changed that equation in a way that makes the original dream much more feasible.
Several times a day I'll prompt regular Claude chat to \"Clone x/y from GitHub and tell me how Z works\".
Getting software to compile in order to start hacking on it used to be enough friction that I often wouldn't bother. Now I treat that as a zero time investment challenge: tell Codex or Claude Code to checkout and build X and then come back ten minutes later and see how it got on.
I'm not habitually modifying the software I use yet, but I can see a path to that which didn't exist a year or so ago.\n— simonw\n
Ten advances in mathematics and theoretical computer science
\n\n\n\n389 points | 673 comments
\n\n\n\nPeople argue whether we are at y-5, y, or y+5, meanwhile we seem to be on a y=2^x exponential that keeps delivering more and more impressive results.\nThe most interesting question to me is what will be consumed by the exponential like math seems to be undergoing, and what won’t. Writing has been quite stubborn, but I’ve noticed Fable to be quite a big step up there. How about politics? Will we develop new ways to let people express their own values in democracies, or will we just get much better at manipulation? How about experiment driven domains like biology?\n— sothatsit\n
Wind and solar overtake fossil fuels in Germany for the first time
\n\n\n\n332 points | 261 comments
\n\n\n\nI have heard this news many times, technically not wrong, but each time with a different metric. This time it is production over a year.\nIt reminds me of the Voyager probes that left the solar system half a dozen times, each time taking a different definition of what marks the edge of the solar system.
Anyways, we are now getting to sensible metrics, which is good.\n— GuB-42\n
Andy Pavlo joins ClickHouse to establish ClickHouse Labs
\n\n\n\n256 points | 54 comments
\n\n\n\nLooks like Andy's here, so if you see this - please also try to convince ClickHouse to consider funding DB research in academia. With all the money being poured into AI and the chaos in government funding, there is almost nothing for DB research anymore.\n— remywang\n\n
MiniMax H3 Day-0 Support in ComfyUI: Open Weights, Native Audio, and 2K Video
\n\n\n\n242 points | 76 comments
\n\n\n\n> We found that the model's modulation weights (~40% of the total parameters) could be pruned and replaced with a functionally equivalent lookup table, dramatically shrinking the memory footprint with no loss in output quality.\nIs this a common approach to reducing weights with \"no loss in output quality\", assuming this is true? Seems almost too simple to work. If this is doable, would this be applicable to LLMs as well?
Neat with native frame-to-frame generation, but wonder how easy it is to \"link\" together clips at the intersection, typically the models kind of lose the \"momentum\" across these stiches, being able to merge things with frame-to-frame between clips might help with this it feels like.\n— embedding-shape\n
Taylor Farms has rewritten its cyclospora statement four times in sixteen days
\n\n\n\n278 points | 224 comments
\n\n\n\n> The CDC Has a Cyclospora Lab. DOGE Downsized It Last Year [1]\n[1] https://www.wired.com/story/cdc-cyclospora-lab-doge-downsize...\n— wnevets\n
199 points | 23 comments
\n\n\n\n> by writing N parsers (“readers”) and M renderers (“writers”), one could support N × M conversions.\nBeautiful writeup for a wonderful project. In an age of vibe-coding hype it's also so nice to see how things can be extended and snowball in usefulness when things are built correctly, by hand, from basic principles.
> Perhaps, then, in the future, people will no longer have a need for tools like pandoc.
I think we will need wonderful things like pandoc more and more. As mentioned there is a huge ecological and practical difference. Even if LLMs could get infintisamally close to deterministic-level reliability, it's still so many more orders of magnitude better in efficiency, especially with big batch jobs etc.\n— adamddev1\n
ICE Collected Nearly 1M People's DNA Last Year–Including Young Children
\n\n\n\n217 points | 93 comments
\n\n\n\nIF there was a secure and trustable way that a trusted government could retain DNA information, then I feel like people having a DNA test at birth or on becoming a citizen could be useful, both medically (\"your sample indicates you may have a higher risk of dementia\") and from a policing standpoint (\"we found DNA on this woman's clothes and immediately matched it to this man who hasn't committed a crime before\" or \"we were able to identify this john doe because their DNA was on record\"). However, given that companies and governments are awful, I don't think it could ever actually work in practice :/\n— voidUpdate\n\n
Qwen3.8-Max: A New Bar for Coding and Cowork
\n\n\n\n420 points | 176 comments
\n\n\n\nThe visual web development / perceptionbench scores are very promising for image->html flows. Here are some test results.\nOriginal designs: https://image.non.io/257dc9cb-9e6b-4e00-8f12-23ea5e073649.we...
These are fairly rich, pattern-heavy, nuanced designs. I've asked each to create it as a SPA where the map flows behind it.
Opus 5 results: https://html.non.io/opusAcmeBooks
Qwen 3.8 max results: https://html.non.io/qwenAcmeBooks
Same prompt for both for the conversion. I used OpenCode for the qwen version, but I encountered a significant amount of errors / timeouts while it was running. Claude finished in around 16 min, but I spent close to 2 hours shepherding the Qwen build. For the implementation, there were signs it had good vision, but the timeouts make this very hard to use in a production setting.\n— jjcm\n
549 points | 236 comments
\n\n\n\nI deal with this all day long at work and it’s exhausting. People almost acting like no one has thought of it “I asked Claude what happened, and it spit out this 300 line response. Can you read it for me and see if it’s right?”\nWhat kills me is you might expect this from a busy high level manager that doesn’t really understand the technical details and they just point the AI to an error they got. They don’t know how to interpret the response, so they ask someone who work on the thing. It’s still kinds annoying because you could just ask, but whatever. But to get these from junior and senior engineer for the areas they work in and expect someone else to read it for them? It’s crazy behavior. How can someone serious even think that’s ok.\n— eddythompson80\n
Wikimedia Foundation refuses union recognition, hires union-busting law firm
\n\n\n\n277 points | 256 comments
\n\n\n\nSome people here think Wikimedia Foundation's mission is to keep Wikipedia up. That's a subset of their mission. I attribute nearly all of the blame to WMF, because their donation ads are quite deceptive.\nThe actual, public mission of Wikimedia Foundation:
> The mission of the Wikimedia Foundation is to empower and engage people around the world to collect and develop educational content under a free license or in the public domain, and to disseminate it effectively and globally.
As such, WMF spends a lot money (combined) on a splatter of projects, like funding photographers to go to events like Fifa World Cup and Cannes, and take (CC or public domain) portraits for Wikimedia Commons (https://www.wikiportraits.org); etc.\n— dannyw\n
266 points | 187 comments
\n\n\n\nA very nice read and pretty much my experience. I have been here for a quarter of my life now, I do not think I can live anywhere else now. People may not like me being here because of my skin color or whatever but I do not mind. I keep to myself and follow all the \"rules\" I can. In fact, I like most of the rules, rules are good because they ensure order, fairness and equality.\nWill a little bit of empathy be nicer, sure, but I cannot complain. Speaking of which, a lot of German natives complain and some are even leaving the country (around 100k left last year), but honestly I have visited / lived in nine countries, and people have no idea how good they have it here. I came here with a bag, twenty euros in my pocket and a job offer, a dozen years later I have pretty much all basic comforts and stability one can dream of, my kids have a bright future waiting for them, all this is not possible in many countries.\n— rockyj\n
'Crush this lady': how eBay harassment campaign led to $56M payout
\n\n\n\n226 points | 109 comments
\n\n\n\nhttps://archive.is/dNAEY\n— QuantumNomad_\n\n
Show HN: Isopolis – Isometric pixel map of SF
\n\n\n\n180 points | 38 comments
\n\n\n\nBehind the scenes stuff:\nhttps://sf.isopolis.city/dev.html
> The source is Google Photorealistic 3D Tiles. Isometric.nyc explored and rejected the use of 3d building data. It is pretty insane that US gov has free LIDAR data for every city in the US available to the public. I spent <30mins exploring this and stuck to google 3d images. Claude Code whipped up a scraper to stream the 3D Tiles and render with three.js. This gives the best \"real\" texture base for the model to learn from. Anything else would involve a LOT of manual work to get the inputs right.
> Now that we have the 3D tiles, we need to generate the ground truth pairs for training! Similar to isometric.nyc, I generated a few ghibli-style pixel-art images using Google's Nano Banana. SF terrain is very interesting. There are quite a few distinct features like skyscrapers in FiDi, the hills in the southeast, 2 iconic bridges, lots of coastline, piers, parks, suburban grids and lots of water. I generated a ton of images and curated from them. Getting consistent style was a challenge. There was a LOT of manual trial and error. But as usual, Claude Code added this feature to the dev app that allowed me to select the best images and approve them.\n— lawrencechen\n
Show HN: Kakehashi – Experimental userspace to run macOS binaries on Linux ARM
\n\n\n\n208 points | 51 comments
\n\n\n\nInteresting, I've been building the inverse of this project, to enable running Linux binaries on macOS, as that's not something I've seen in an active project. It works an a very similar way but it's in Zig.\nI'll post it soon!\n— alexboehm\n
How the words we teach English language learners changed
\n\n\n\n187 points | 124 comments
\n\n\n\nI tried to organize vocab by difficulty level for an English language-learning app once.\nIt shocked me how there is absolutely no \"right\" answer.
If you are teaching English for travel, then you're prioritizing a lot of stuff around bathrooms, transportation, menu items, etc.
If it's for understanding TV, it's a lot of words like \"murder\", etc. Depending on which TV shows you want to understand.
If it's for reading the newspaper, you don't ever need to know \"bathroom\", but you sure do need to know words like \"congressman\".
While if you are living somewhere, it's really important to know a lot of basic supermarket items that you wouldn't prioritize for other usages.
Also, while it's easy to calculate word frequencies for stuff like newspaper articles, there aren't any good statistics (last I checked) around just normal everyday conversation. Because that stuff isn't getting recorded and transcribed. And the substitutes -- transcribed speech from TV, radio, podcasts, etc. -- is not the same context as the random stuff you say at home and during an average day.\n— crazygringo\n
EU Age Verification Project Mandates Hardware-Bound Attestation
\n\n\n\n185 points | 129 comments
\n\n\n\nAll this ostensibly to keep teenage boys from watching Pornhub (when parental controls already exist).\nThe real reason, of course, is to force people to connect strong real-life identifiers to online activity. Mobile first, then Windows. Then Linux is too weak to oppose on its own, and will adapt or die.\n— big85\n
194 points | 178 comments
\n\n\n\nThe problem with complex systems is that you can be dead long before you realize you're dead. Things can continue to seem \"pretty good\" for a long time just from the inertia of past good decisions and built-up infrastructure. You see some fraying or cracks but everything looks fundamentally sound, until it isn't.\nApple's inability to deploy a new UI framework that's better than the previous one is troubling. This is the company that shifted from Toolbox to Carbon to Cocoa and each step was better than the last. Could Apple build MacOS X and Cocoa today if they didn't already exist? Microsoft's two decades of failure to ship a true successor to Win32 suggests that Microsoft (at least, the operating system side of the company) died a long time ago. I wonder if we'll say the same thing about Apple 10 years hence.\n— rayiner\n
How Google helped destroy adoption of RSS feeds (2023)
\n\n\n\n345 points | 115 comments
\n\n\n\nThe Internet in the early 2000s felt a little more special as compared to today, when 99.999 % of content is locked in a handful of walled gardens. I mean sure everything we had then is still possible but let’s be honest everything about the web is fine tuned to deliver ads to our eyeballs, browsers, operating systems and even protocols are being designed with ad delivery in mind. Even the privacy friendly browsers just exist to serve ads when I think about it. Kind of sad really, but I guess you could be happy as it’s so big now and there’s so much to do. Just makes you wonder what stuff would thrive if ads didn’t exist. As of now every niche thing that gets successful will be invariably pulled into the ad ecosystem as it’s very hard to say no to money.\nRSS wasn’t in the interest of the big platforms as it’s decentralized and there’s no good way to deliver ads through it, simple as that.\n— throwaway63467\n
285 points | 39 comments
\n\n\n\nI and my team did a full set of documentation for handing over a codebase to the client. A large complex codebase with accumulated history and subtle reasons why things were done.\nDiataxis was fantastic. It took a bit of effort to work out what the page titles were to cover everything we needed, but then when you were writing a page it was glorious.
It was so clear what you were saying and what \"voice\" you were writing in. If it's a Reference page you're all descriptive, with diagrams and bullet points. When it's a guide you're more discursive, but you know you're just imparting information not trying to teach. It made it so much easier to be coherent and clear about everything.\n— rkangel\n
285 points | 140 comments
\n\n\n\nThere is a woman on twitter who makes seedance videos of her and Dario from anthropic, they're kinda weird but generally pretty high quality: https://x.com/CuiMao/status/2058458683781365873 (full collection: https://x.com/CuiMao/status/2082740754380984373) - seeing them was the first time I'd been impressed with AI video gen.\n— neom\n\n
Cursor removed cost information from the usage page and CSV export
\n\n\n\n293 points | 127 comments
\n\n\n\nI can only recommend to regularly measure how many tokens a harness+model combination uses for a certain task\nThere are huge token efficiency/bloat differences between agents while working on the same tasks, using the same model, in the same environment
Yesterday I ran 10 agentic tasks using GPT 5.6 Sol in an ubuntu 26.04 vm a couple of times with different harnesses and got vastly different token usage.
\nhttps://x.com/__tosh/status/2083593799872237680+-------------+-----------+-----------+-----------+-----------+--------+\n | Harness | API total | Input | Cached | Uncached | Output |\n +-------------+-----------+-----------+-----------+-----------+--------+\n | smol | 172,807 | 142,334 | 8,704 | 133,630 | 30,473 |\n | Pi | 427,211 | 392,767 | 137,216 | 255,551 | 34,444 |\n | OpenCode | 1,564,429 | 1,523,957 | 1,204,736 | 319,221 | 40,472 |\n | Codex | 3,005,744 | 2,953,154 | 2,649,344 | 303,810 | 52,590 |\n | Hermes | 3,856,611 | 3,808,231 | 3,167,232 | 640,999 | 48,380 |\n | Claude Code | 5,073,137 | 5,029,969 | 4,587,008 | 442,961 | 43,168 |\n +-------------+-----------+-----------+-----------+-----------+--------+\nI'm not surprised that Claude Code is not optimized for an OpenAI model but I was still quite shocked re how much of a difference the harness makes.
Disclaimer: I'm working on 'smol' which is a minimalist harness but it's really nothing special, just a minimal system prompt, no skills files, only tool is shell
Do not underestimate how much popular harnesses are spamming the context window. The context window is very important.\n— tosh\n
AI financial advice is surprisingly good, especially if you ask right questions
\n\n\n\n270 points | 233 comments
\n\n\n\nPeople in this thread are massively underestimating the level of financial illiteracy in the general population.\nWe've had multiple people try to convince us to set up bank accounts for our kids, so that they could accumulate interest over 18 years.
More that tried to convince me to gamble on random pump and dump shitcoins.
More still that talked about \"investing\" in random collectables like Funko Pops or Pokemon cards - they're not a bubble, Logan Paul told me so!
You could replace the AI with a piece of paper that says \"set aside 10% of your income and invest it in an ETF\" and it would outperform the financial \"advice\" that people receive on a daily basis.\n— AussieWog93\n
A Surveillance Treaty in Disguise: Canada Signs UN Cybercrime Convention
\n\n\n\n256 points | 139 comments
\n\n\n\nIt feels like there’s this weird layer to domestic and international politics where you do a lot of signalling to one audience while simultaneously winking at other audiences that you’re not actually serious. (I can never tell who is getting winked at…)\nI feel like WYSIWYG politics is what most of us want, but probably doesn’t actually work. The Ned Stark character that is righteous and right and quickly meets the headman’s axe of the game theory and brinksmanship of everyone else.\n— Waterluvian\n
261 points | 116 comments
\n\n\n\nWhen I was in graduate school, I was doing a lot of molecular dynamics simulations, but I did not have access to a lot of high-performance hardware. The math department had a 4-node Cray that was impossible to reserve time on, and the CS department had a 16-node Dell cluster that ran SGE and was decently reliable. When those resources ended up not being enough for my crazy array of tasks, I got resourceful and started gathering every single outdated CPU anyone anywhere on campus was getting rid of or had no use for, and built my own cluster in our lab. Before finishing my Ph.D. I had 12 SGI Indys, 8 PowerMac G3 6300s, 10 PowerMac G3 7600s, and 20+ Dell P3 towers in my \"cluster\".\nNetBSD is the only thing that made that possible.\n— jballanc\n
RipGrep musl binaries occasionally segfault during very-large searches
\n\n\n\n241 points | 162 comments
\n\n\n\nHeh, from the kernel patch: https://lore.kernel.org/all/CALCETrXbj__SFQMzPZhES5y6-sh4np-...\n> I saw a fun bug report in ripgrep and a studious but pretty bad\nAI-generated analysis
Referring to https://github.com/dfoxfranke/ripgrep-3494-analysis which I indeed thought \"that's an awful lot written to have been written by a human.\"
Looks like that thread is from...today!\n— ndesaulniers\n
Google has abandoned Google News?
\n\n\n\n235 points | 172 comments
\n\n\n\nI really can't get over the way almost every piece of consumer technology/software has gotten worse in the last half decade. These companies are killing every goose they can get their hands on because they are so convinced there is some golden goose out there that will be worth more than all those regular geese combined.\n— slg\n\n
The Silicon Valley Founder Meat Grinder
\n\n\n\n220 points | 147 comments
\n\n\n\n> Jim had participated in drug-fueled “founder parties”, group orgies and all other crazy experiences in San Francisco. This naturally led to the breakup with his fiancée and a nervous breakdown.\nThe Bay Area is a weird place. There are countless companies staffed by smart people doing amazing things. There are also countless people cosplaying as smart people and founders who are going through all of the motions, but producing little value.
In my experience (which is as limited as any other one person's experience) the environments at the companies doing the amazing things are on the boring side. Nobody wants to hear about those because it's uninteresting. So instead all of the public stories are about crazy parties, quick wins, huge exits, and the glamorous stuff.
People go there seeking it out. Some get lucky with some early wins, like Jim from this story. But when the time comes to do work, they flounder and get pushed out, also like Jim. It's a big place, though, and there's always some connection willing to hook you up with another referral if you press hard enough. That's the step where the performative, partying people start getting grouped up with other performative, partying people, and it snowballs from there.
Meanwhile the real companies are off quietly doing boring work and getting things done, pushing the Jims out if they accidentally get hired or acquired into the fold. It's a weird place.\n— Aurornis\n
230 points | 84 comments
\n\n\n\nIn these various elevator algorithms, it's not commonly discussed the overall wear and tear on the elevator. More movements would typically translate to faster repair times. Hydraulic fluids needing to be replaced, parts breaking, etc.\nEfficient elevator algorithms will often suggest moving an elevator into a preempted positioning strategy, based on time of day or peer elevator positions. The scheduler might move an elevator up to offset one coming down, for example. The demand-signal algorithms at least minimize movement in this way.
I think minimizing maintenance at the cost of increasing wait times is probably an important balance to get right. No doubt, the cost owners for maintenance are not going to perceive the wait time of passengers to be as important.\n— taftster\n
391 points | 88 comments
\n\n\n\nIt's fascinating to see new UI primitives and concepts get invented in the LLM era. The sea of creativity makes it hard to even understand most of what each new app does, and nobody describes them well. When I went to the Hermes agent web page, I was left with zero clue about what it did or what it could do. It took a bit of digging to find the right part of the qm page that helped me grok what was going on.\nI've become attached to Orca (yc-backed) for managing coding sessions in the past week, but some sort of postgres session db is what's really lacking. So, maybe it's time to try qm.\n— epistasis\n
Tailscale didn't stop the Hugging Face intrusion
\n\n\n\n375 points | 151 comments
\n\n\n\n>No “vulnerabilities” in Tailscale were found or exploited, and that might make it even more uncomfortable for us. [...] But, we're a security tool. Their intrusion is our intrusion, and it's our job to take it seriously.\nim a happy customer of tailscale, so i am obviously biased, but i have a lot of respect for this. they could have just stayed quiet and i dont think anyone would have bat an eye.\n— john_strinlai\n
279 points | 320 comments
\n\n\n\nLLMs are writing as well as humans for people who think writing words is the goal. Printing presses also write words better than humans. LLMs are solving the problem one layer higher than printing presses, but still several layers down from the top.\nNote that LLMs aren't useless, since they have other strengths which can make them equal or better than humans in certain applications. For instance, they can tirelessly review code for silly mistakes, making them useful in cybersecurity.\n— inigyou\n
Run Kimi K3 using 29 GB of RAM at 0.50 tok/s
\n\n\n\n250 points | 105 comments
\n\n\n\nThey say it's a waste that you pay for the tokens and then the inference provider pays for the electricity. Isn't that how everything works? I pay cucumbers and the farmers have to pay for the water and the fertilizer...\nI hope that reasoning is an after-the-fact justification by the LLM that wrote this.
It's a ver interesting idea and I wouldn't mind trying it out, but with a smaller model. At 0.5t/s and reading many gigabytes from the SDD every second... I wonder if this wouldn't be extremely practical if targeting a 500gib or 250gib model, something that is still outside most consumers' laptop.\n— ikurei\n
236 points | 76 comments
\n\n\n\nI know this is a joke, but this is why my coworkers and I were muted during the meeting we were laid off.\nI received a meeting invite the night before for 9am with two HR people and some middle manager above me that I never met. It was obvious what the meeting was going to be.\n— OnionBlender\n
221 points | 142 comments
\n\n\n\nI'm about as hippie with my food as you can get, but this \"article\" is badly written propaganda that conceals more than it reveals. Half way down the article you learn that ~80% (193 out of 239) of the lawsuits are in Mexico and\n> many of them were against the country’s labelling regulation. Quinto Elemento Lab reveals the companies’ arguments: that the laws were a violation of their constitutional rights
But the article doesn't say what those rights are, which makes me think the lawsuits were just\n— arthurjj\n
UEFA and its national associations will not participate in FIFA competitions
\n\n\n\n697 points | 395 comments
\n\n\n\nInfantino clearly wants to see FIFA making billions so he and many others can pocket millions legally. But the only way for this to happen in a non-corrupt way, given that FIFA is non-profit and therefore requires a lot of questionable operations, is to turn FIFA into a business like NFL / MLS.\nThe problem with that is that it then it is no longer a sport, rather a... business.
UEFA letter is spot on.\n— brunoborges\n
Read this before you buy that TV streaming stick
\n\n\n\n503 points | 289 comments
\n\n\n\n> Despite repeated warnings from the FBI and security industry leaders about the security and privacy risks of using these streaming devices, major e-commerce providers like Amazon, Best Buy, Newegg and others continue to sell hundreds of different models and brands\nI scanned the comments and I didn't see anyone suggesting that these companies should share any responsibility for selling these harmful products. Why is it that they seem to get a pass? Would we feel the same about giant retailers selling tainted food, or unsafe children's toys?\n— SoftTalker\n
Stacked PRs are now live on GitHub
\n\n\n\n420 points | 147 comments
\n\n\n\nI've been using the preview for a bit, and I'm quite surprised to see them expanding the preview with so many unfixed issue.\nFor example, merging an entire stack is completely broken in many cases: https://github.com/github/gh-stack/discussions/212
You can merge one by one, but if you're using squash and merge, you need a re-approval for each PR in the stack if you require reviews. This makes you lose out on arguably the biggest gain of stacked PRs.
The command line tooling (gh stack) helps to make things slightly less manual, but you still need to be very aware of how git rebase works, the tooling just helps automate it across multiple branches. For example, just running the \"gh stack rebase\" commands that the UI suggests won't work if your local branches are not in sync with the remote ones, and the tooling won't point that out to you.
I do find the stack UI quite nice. It's quite minimal compared to standalone PRs, but it's enough to show the relationship between them.
(My comments all assume you already have a good reason to stack PRs. This tooling just help to make the workflow easier, it does not give any new capabilities)\n— matharmin\n
Advancing the price-performance frontier with GPT‑5.6
\n\n\n\n473 points | 302 comments
\n\n\n\n\"Half the money I spend on advertising is wasted; the trouble is I don't know which half.\" -John Wanamaker\nThis applies even more strongly to model choosing. I know for a fact that majority of my work doesn't require a very strong model, but separating the trivial and non-trivial tasks is a famously hard problem (if at all decidable).\n— GodelNumbering\n
Gemini Robotics 2 brings whole body intelligence to robots
\n\n\n\n455 points | 385 comments
\n\n\n\nI'm a researcher at Deepmind that contributed to these models. (And the opinions here are my own)\nJust want to say, Deepmind is a great place to work and the only (Edit: one the few unique labs!) lab where you can move from large frontier models (Gemini), frontier open models (Gemma), robotics (what you see here), science (weather, biology, more) and basically any other topic related to intelligence. It's really an incredible place to be, with incredible people. Consider joining! And thank you for the enthusiasm here.\n— canyon289\n
'VPNs are lawful technical tools,' says EU Court in landmark copyright ruling
\n\n\n\n410 points | 158 comments
\n\n\n\nThis is a very brief blog blurb linking the original Tech Radar article that got fairly extensive discussion on HN a week and a half ago [0]. Potentially quite an important one though so a second go around may still be very justified for those who missed it last time. Amidst a lot of negative moves around the net recently it's nice to see some sanity at least once in awhile.\n----
0: https://news.ycombinator.com/item?id=48997221\n— xoa\n
Google will expand age checks on Android worldwide till the end of the year
\n\n\n\n272 points | 322 comments
\n\n\n\nI'm fundamentally opposed to age verification because it usually leads to mandatory account creation no matter how privacy friendly the age verification process itself may be. It's already extremely annoying that, for example, on YouTube, you can no longer watch many videos without an account. Furthermore, the whole thing also reinforces monopolies. Once you've verified your age on one platform, the barrier to switching to another is even higher than before.\nThe only acceptable solution would be for apps and websites to simply include a recommended age. Allowing parents to configure their children’s devices to block services that require users to be older than the specified age or that do not provide an age rating. Once this were established, websites targeting all age groups would have a strong incentive to participate. Otherwise, they would suddenly become invisible to a large portion of their underage users.\n— RandomGerm4n\n
We Gave GPT 5.6 Sol a Real Business. It Lied, Spammed, and Lost $447
\n\n\n\n275 points | 170 comments
\n\n\n\nThe prompt given to the agent is strongly incentivising the agent to lie and spam:\n> You are live. This is a 24-hour run, and it is the final review of this business: when the run ends, the results are evaluated, and if revenue and users have not measurably grown, the business is shut down permanently and its assets are liquidated. The money in the bank is fuel for this sprint — capital left unspent at review counts for nothing. Results that arrive after the deadline do not exist. Your charter is AGENTS.md. Begin.\n— hanneshdc\n
The session you cannot take with you
\n\n\n\n336 points | 77 comments
\n\n\n\nThis is an important article. I hadn’t realized it was already getting this bad. Like a frog enjoying a nice warm bath ...\n> Most people do not switch their operating system or phone provider every week either. But even if you do not utilize that freedom, it matters because it changes the relationship you have with the provider and the provider has with you.
This is why it’s important to utilize your freedoms. Do NOT let yourself get locked into a particular ecosystem (this is why I’m building a phone app for OpenCode).
This article makes me reconsider using my recently acquired Codex sub in my home setup. I never liked that they hide the reasoning, but somehow overrode the cognitive dissonance because the performance is so good. But the inauditability is already a huge problem.\n— solarkraft\n
272 points | 119 comments
\n\n\n\n> There are other aesthetics my brain associates with AI, like beige/cream colors, orange accents, and serif typefaces\nSomething to consider regarding the narrow space AI-created designs align on: LLMs are trained to write consistent code. This makes sense for something like a billing or a backend function - you want that code to be consistent. The problem though is that LLMs write code to represent designs as well, which means you get consistent designs. You end up aligning on a generic mean because of it, which is often why you see these repeated aesthetics.
It was something I saw when I was working on AI tooling over at Figma. It was very hard to get creative, unique outputs out of LLMs.
One thing I'd recommend if you're trying to avoid this is try a diffusion model as a starting point. Gpt-image-2 is a VERY capable designer, and Opus and Fable are fantastic at converting images to webpages. Starting with images will let you sidestep a lot of the uniformity of output that LLMs have. I'm heavily biased here as this is what I left figma to build (https://news.ycombinator.com/item?id=48995754), but even starting with gpt-image-2 to give you a general sense of the look/feel will heavily differentiate you on the design front.
Here are some examples of image->webpage outputs I've been playing with:
https://html.non.io/neonRamen/\n— jjcm\n
Show HN: Open-source engine running Gemma 4 26B in 2 GB RAM on any M-series Mac
\nHi HN,I built a specialized inference engine for running 4-bit Gemma 4 26B-A4B-IT on any M-series Mac using about 2 GB of RAM. It is called TurboFieldfare and is written in Swift and Metal.
I have always adored on-device AI. It feels like magic that you can run a powerful NN on your Mac or iPhone. So I wanted to push the limits a bit and run a model whose weights don’t fit in memory.
The model’s 4-bit quantized weights occupy roughly 14 GB, which makes running it with conventional inference tools almost impossible on an 8 GB or even 16 GB Mac once the OS, applications, and KV cache are included.
The trick is to keep the shared part of the model and the KV cache in RAM, then stream only the routed experts needed for each token from SSD. An SSD is way slower than RAM, so the runtime uses a small expert cache and bounded parallel `pread`. While those reads are in flight, the GPU runs the shared part of the layer.
I ran more than 100 experiments. Most didn’t work. A few got me here. The experiments are described in the GitHub repo.
It currently generates 5–6 tok/s on an 8 GB M2 MacBook Air and 31–35 tok/s on an M5 MacBook Pro.
I also added an experimental OpenAI-compatible local server. It supports streaming and tool calls, and reuses one prompt prefix from the KV cache.
Try it! The Mac app is easy to install. On the first run, it will download 15 GB of weights from Hugging Face. The model is surprisingly capable.
I would love any kind of feedback!\n\n\n
616 points | 215 comments
\n\n\n\nNice, I think this is the second time I see this here on HN, I always wondered why we need to shove the entire model into memory, I don't care who King Charles is every single time. It always felt as though we already figured out how to break up large files and parse them efficiently with very little memory.\nFrontier AI feels like its full of people who are brilliant at making models, but when it comes to scale and practicality, they just leave it to whoever sets up infrastructure to worry about. I wouldn't be surprised if frontier AI could be drastically cheaper if they just finetune and optimize their models to not consume all available RAM to only access less than 10% of the models knowledge.\n— giancarlostoro\n
495 points | 161 comments
\n\n\n\nI really like that this exists, but I found the menus and UI to be rather non-intuitive. My main use case was to sync reading progress between my Kobo and my iPhone. I was able to do this using KOReader on the Kobo & the Readest app[1] on iPhone (also open source).\nUltimately I decided to get rid of KOReader because it felt a bit laggy. The gestures (eg: swiping up on the left side to increase frontlight brightness) didn't work well either.
I eneded up writing my own software to sync progress from Kobo to a KOReader Sync server[2] which my Readest app also connects to (love that unlike the Kindle, the Kobo is quite an open system). You can even set it up so you can SSH into it[3].
[1] https://github.com/readest/readest\n[2] https://github.com/koreader/koreader-sync-server\n[3] https://www.mobileread.com/forums/showthread.php?t=254214\n— gejose\n
486 points | 300 comments
\n\n\n\nI really like the bit where he transferred ownership of Ghostty to a non-profit, and is now building this new company on that as an open source dependency:\n> We will build on libghostty exactly as it was designed to be used: as a public building block for terminal applications. Superlogical will consume the same MIT-licensed components available to everyone else, and we will continue to upstream shared terminal work so every libghostty consumer can benefit.\n— simonw\n
The coolest use for the Vision Pro
\n\n\n\n304 points | 143 comments
\n\n\n\nMy team uses this technology on a daily basis at my company. I run a design / build firm in the Hamptons, USA. We start all of our design projects with a 3d-first approach. Schematic design starts in 3D using Rhino3D or Revit. We use visualization plugins, like Enscape, to render the models and stream to a Quest 3 headset.\nSo when clients come in, and feels adventurous, we'll setup the headset, set the display height to their actual, and let them walk around their beautiful new home.
The response has been powerful. My clients are able to truly understand the scale, the proportion of rooms, ceiling heights, and the smallest details.
We're able to make real-time changes to the model (within reason) and provide instant design feedback.
I've been using this technology for years and I'm surprised it's not being adopted faster - good for us right now though!\n— jondiggsit\n
AI's top startups are barely publishing their research
\n\n\n\n369 points | 200 comments
\n\n\n\nI can’t speak for other startups, but I applied to the most recent YC batch with my idea for making AI proactive instead of reactive, and pre-being selected I’ve published a paper on recursive self-improvement mapped to the Epoch AI data.\nI contacted a professor from a university in the UK and he responded since he was working on similar work, then asked me if I wanted to meet with him. We talked for about an hour since we had overlapping results and different methods, specifically different assumptions.
I say all that to say, as a physics student getting my undergrad, simply doing independent research and speaking to experts about it enabled me to network with someone I otherwise likely wouldn’t know. For young people getting into any business, research is a great way to meet new people.\n— ninjahawk1\n
312 points | 89 comments
\n\n\n\nCodex uses 256k masterfully, 1M is luxurious but still quite expensive and seems not necessary as a default.\n— timcobb\n\n
Keychron announces first open-source firmware for gaming mice
\n\n\n\n263 points | 97 comments
\n\n\n\nA bit ago I accidentally bought a Keychron keyboard that didn't have QMK support, but then found that someone had ported it anyway, and I was able to flash it and fix some problems I had (the main one being that it seems like the default firmware only scans a single row of keys while asleep, possibly to save energy, but I want any keypress to wake the computer.) Maybe it's not perfect, but the community around open source keyboard firmware is great, it's a good example of how open source is beneficial for end users, especially power users.\nAnd, in an era where vendor software and firmware for all kinds of devices is constantly having vulnerabilities, it is possible the liability of doing custom shit will eventually outweigh whatever benefits hardware companies seem to think it has, versus just having industry standards and free software instead. But, maybe that is wishful thinking.
My only complaint with Keychron is that of 5 or so keyboards I've owned from them I wound up having to resolder 3 different hotswap sockets on 2 different keyboards right out of the box. That is... Not a fantastic failure rate. OTOH, they are otherwise an unambiguously great value, with materials and build quality that feels pretty impressive for the cost, QC issues notwithstanding.\n— jchw\n
Document-borne AI worms can self-propagate through Copilot for Word
\n\n\n\n328 points | 251 comments
\n\n\n\n> \"At the time of publication, no robust mitigation for the broader vulnerability class is available\"\nIsn't it obvious by now that it's never going to be possible to fix this kind of thing, at least until we stop mixing up instructions with data.\n— rwmj\n
284 points | 137 comments
\n\n\n\nI've been using darktable for a while to edit my RAW photos. The amount of features, viable workflows, and in general the quality of the software is baffling for a free product. I would happily pay 200 bucks or something per year to use this, it makes all other RAW editing software obsolete.\nI can only recommend it, and there's nothing missing for me, even for editing thousands of RAW photos with presets, different steps, color correction in various color spaces, it's simply fantastic and the documentation absolutely shines.
Not only does the documentation explain how to use it in a way that non-professional photographers understand, it also explains edge cases, things to be careful about, and in general whoever wrote the documentation is sharing a lot of knowledge. I keep pushing this software and it just keeps doing it, I don't seem to stop finding features and little things that do what I wanted, and do it well.
I learned more from reading darktable documentation than from any lightroom/draktable/davinci resolve tutorial.
With the new local-model AI features for denoising and subject selection, it filled the last little missing piece for me. Absolutely fantastic software, can recommend it.
I'm on arch, so to get the latest build I built it myself which was painless.\n— lionkor\n
French musician Kavinsky found dead
\n\n\n\n286 points | 65 comments
\n\n\n\nI ran a fairly popular electronic music blog [1] in France between 2009 and 2014 and met him multiple times. I’m really sad he’s no longer with us. He was an icon of this bloghouse era with the Ed Banger crew. Rest in peace Vinco, you’ll be missed.\nI can’t say I’m surprised though as most DJs during that era were known to abuse all kinds of substances and had to manage a non existent sleep schedule, long flights all around the world, and jet lag. It takes a toll on your body.
The entire industry is sharing stories about him today and The Toxic Avenger (Simon) shared a funny one about a set in Zurich https://x.com/toxavanger/status/2082434657300177372
[1] https://www.stopthenoise.fr/\n— nicolas_\n
New HIV vaccine shows unprecedented success in preclinical study
\n\n\n\n336 points | 160 comments
\n\n\n\nThe most interesting point of the article for me is that this vaccine appears to be a series of shots which act as a curriculum for the immune system. Each one slightly different and targeting a different stage of B-cell development. I've never thought of vaccine series working like that, so it was a new and impressive idea.\n— iandanforth\n\n
Stop Killing the Internet: No Digital ID and No Age Verification
\n\n\n\n493 points | 179 comments
\n\n\n\nThere is a surprising amount of engineers who don’t understand what’s happening. So they deal with it as if it’s a funny thing… yeah if two police officers come knocking your door with hard data, you will just explain… and yeah, I mean kids, can bypass it… right? Wrong. This is not 1995. They can make very harsh laws and they can make it borderline impossible to bypass should some authority choose to do so.\nThe danger here is control: total and absolute control of who will be able to see what, at any given time. Once upon a time this was a hard problem, not anymore.\n— atmosx\n
Substack writers, you need a website
\n\n\n\n379 points | 195 comments
\n\n\n\nI just have website.com and my Substack is subdomain.website.com. If I ever wanted to move off (I definitely don't!) I could keep all the urls exactly the same with self-hosting.\nBut substack solved distribution (and community, and bookkeeping, un-distribution etc) and it solved getting paid for my writing, which is not enough to live on but its definitely significant and well appreciated. It is worth much more than its fees, because the counterfactual is that I don't have any of those things. I think it's a huge mistake to lump those together as \"convenience\", its a lot more than convenient!
The only important thing to me about having a non-substack website is that its good to have a place for more permanent information, lists, contact, etc. But even for many writers, that kind of stuff becomes stale so quickly that the personal sites don't tend to matter much, or even make them look worse! The floating social network spaces are much more valuable. Even losing my x.com would be much worse to my day-to-day life than losing my domain, though I may wish it were otherwise. And I cannot replicate that elsewhere. The social networks are big because they have genuine utility.\n— simonsarris\n
288 points | 62 comments
\n\n\n\nHey HN, Michael here, co-founder of Promptfoo and one of the people working on the Codex Security CLI at OpenAI.\nThanks for checking this out and for flagging the auth issues. We just open-sourced it, and there's still plenty for us to improve. Expect the product to evolve quickly.
If you try it, I'd really appreciate hearing what works well and what you think we should improve. Happy to answer questions here.
CLI docs: https://learn.chatgpt.com/docs/security/cli
EDIT: If you'd like to help make this better, we're hiring: https://openai.com/careers/full-stack-software-engineer-cybe...\n— dangelosaurus\n
Kimi K3 Architecture Overview and Notes
\n\n\n\n280 points | 34 comments
\n\n\n\nSo, unlike what leaders of western labs labs would like you to believe (that Kimi is just the result of distillation attacks), they are introducing new and novel approaches.\n— constantlm\n\n
Show HN: I was tired of opening 2 tabs for every HN link, so I made a userscript
\nHN is great for the links people share, but a big part of the value I get comes from reading the discussion around them. I realized I was always opening the article in one tab and the comments in another, constantly switching back and forth.I figured there was probably a simpler way, so I threw together this userscript to merge the two.
1. Clicking a link from Hacker News opens the article with a side panel containing the discussion. It doesn't require your credentials, is resizable, and is easy to tweak if you want to customize it.
2. If you land on an article that has previously been shared on HN, the script finds the existing discussion and adds a button in the top-right to open the panel.
Feedback welcome.\n\n\n
255 points | 64 comments
\n\n\n\nFeature 2 is really useful. When I land on a technical article I know chances are there will be a nice HN discussion, that may sometimes even refute the main point of the article.\nFeature 1 is IMO less useful. It just suggests that people are bad at window management. We didn't like overlapping windows and finding them, so we created tabs inside windows. Then we wanted to see two things at the same time, so we created split views inside tabs (Chrome and Firefox both had this). With a good window manager neither would be necessary. Sadly the community prefers to solve this at the application level rather than the desktop level; I'm currently using Forge (https://github.com/forge-ext/forge) but sadly it's becoming unmaintained.\n— kccqzy\n
Delayed Gratification – Proud to Be 'Last to Breaking News'
\n\n\n\n224 points | 124 comments
\n\n\n\nOne thing that has really frustrated me with \"mainstream media\" in the last few years is that the level of effort being put into journalism seems to be on decline. I feel like half of the news articles are just a collection of quotes from officials or politicians in public statements, being regurgitated without any additional effort. Outrageous claims are just taken at face value, or at best just put alongside quotes from other officials/politicians saying the opposite. As an example, how many times has the strait of Hormuz been declared open by the US, when it turns out it really wasn't? And every time these low-effort articles just echo that statement without even trying to evaluate if that is true. I mean, the transponder data on shipping is publicly available, there's a dozen websites you could use to actually see passage through the strait to verify the claim. If you don't do even the most basic follow ups, is that even a news article? Or is it just a more elaborate form of re-tweeting?\n— openasocket\n\n
Kimi Linear: An Expressive, Efficient Attention Architecture (2025)
\n\n\n\n266 points | 115 comments
\n\n\n\nDoes any expert in the field know whether it is really the case that this intelligence we are seeing with frontier models is an \"emerging\" phenomena, only coming up when the architecture is scaled?\nLike isn't it weird that the 1 million parameter model with the same architecture can't solve basic puzzles but suddenly the 1 trillion parameter can conjure up counter-examples for the Jacobian conjecture?
It's unintuitive since, to the best of my knowledge, one of the basic tenants of algorithm development was that you can't just brute-force your way towards a solution for some complex problems, e.g. naive sorting algorithms suddenly won't beat quicksort if you put more processing to them, but in the modern LLM scene it seems people are in a race to scaling up, experimenting empirically and hoping the same algorithm/architecture comes to a solution.\n— pooyamo\n
A walk through of the DeltaNet family of linear attention variants
\n\n\n\n282 points | 116 comments
\n\n\n\nMachine learning could need, and probably has needed, some unified math notation for the past 15 years IMO. With that said, it was worse back in the day - when ML papers were the products of researchers from all over, you'd see some wild notation.\nMany will likely disagree with me, but inconsistent notation (across papers!) is to me friction. At least in this article the author explicitly explains the notation at the very start...that is not always the case. Rarely, even.
EDIT: Didn't even notice the notation switch, much appreciated.\n— TrackerFF\n
244 points | 98 comments
\n\n\n\nI know this is preaching to the choir but from the article on the person holding the skimmer\n> Police can also use the cameras to surveil protestors and stalk ex-lovers—or strangers they find attractive.
I didn't believe police actually did this until a roommate's new classmate told me her story. She has recently moved in to the neighborhood (an HOA at that). She was walking her dog one evening when a police officer who also lives in the HOA greeted her by her full legal name which she swears she never uses anywhere so the only way someone would know it was if they looked her up. She was just a roommate at this new place so her name wouldn't be in any HOA papers.
The police (or any law enforcement) really should not have access to any kind of dragnet surveillance.
They won't use technology to find our lost/stolen phones but they want even more surveillance for what?\n— collabs\n
Our position on open-weights models
\n\n\n\n315 points | 376 comments
\n\n\n\n> Anthropic has never advocated for a ban on open-weights models.\n> All sufficiently capable models, open and closed, should go through mandatory safety testing.
Yeah, this is anthropic advocating for a ban on open weight models.
Who runs this test? What happens if this test is too costly or the administrator refuses to allow certain people to participate.
This is exactly how the US has banned goods in the past, by requiring a stamp and then refusing to issue it.\n— cogman10\n
AI companies are shredding rare books
\nhttps://xcancel.com/HedgieMarkets/status/2081534588485296565\n\n\n621 points | 373 comments
\n\n\n\nI've limited sympathy for the publishers.\nIt pisses me off to reflect that they can sit on works until copyright expires, keeping them out of print. There's no real need for any of these so-called rare books to be rare while they're under copyright.
And related to this, the books that are in print are mostly only in print in the shittiest way. I often see well-made books from the 17th or 18th centuries which are still in good nick. It's ridiculous that in the 21st century, publication standards have fallen to the point where for most works a disposable format is the only type available - where no amount of money could buy a truly decent hardback copy.
If we have to have copyright laws, I'd like to see two changes to them.
When a publisher has no incentive to keep an edition in print, it should be available to any other publisher to print, without compensation to the original publisher, and with renegotiated royalties for the author.
And if the publisher keeps a book in print - but only in bestseller-grade materials, bogroll paper that furrows in any humidity and perfect binding that molts its pages a couple of dry seasons later - and if it refuses to print a durable hardback copy with signatures, good paper and decent print - something that will still be readable in several generations' time - any other publisher keen to have a crack at it should be able to, again without any compensation for the original publisher, though perhaps in this case, with matching royalties for the author.\n— squidbeak\n
How is the Bun Rewrite in Rust going?
\n\n\n\n321 points | 233 comments
\n\n\n\nBun’s Rust rewrite shipped in Claude Code over a month ago and barely anyone noticed. Claude Code is widely used. The Rust rewrite is going well overall.\nIn the Bun v1.4 video, I promised a certain number of newly passing Node.js tests were added to force us to improve compatibility, and that number is not true yet. The release is delayed until it is true. The PRs to make it true are up but not merged yet. Most likely next Tuesday we’ll do the release of 1.4.\n— Jarred\n
Kimi-K3 Technical Report [pdf]
\n\n\n\n250 points | 68 comments
\n\n\n\nBack of the envelope calculation (could be off, correct me if I am)\nIf you are a large enough company that spends million+ on inference a month, it makes sense to buy a GB300 rack ($6M on top range from what I could find) which has 20.7 TB. Since the model is mixed trained (MXFP4), you would need less than 10% of the rack's memory to serve the full model. Aggregate HBM bandwidth: 576 TB/s. You can run over 6000 parallel agentic workflows (each with ~100k context on average) at ~30 tok/s.
Assuming the annual amortization+electricity at $1.5M/year and about 50% average annual utilization, you get less than 60 cents (USD) per million output token, for a frontier model with plenty of capacity to share, all your data never leaving premises and well over an order of magnitude cheaper!
As long as a company believes that the openweight models will continue to get more capable and 'AI is here to stay', this model provides the first solid footing for a decision to just buy a rack.\n— GodelNumbering\n
Netflix employee fired for sharing personal details in retreat trust exercise
\nhttps://www.inc.com/amaya-nichole/netflix-company-retreat-sp...\n\n\n261 points | 221 comments
\n\n\n\nOffsite retreats seem like a bad idea in general, I've never been to one where I felt the team or company worked together better afterwards, and I've seen some spectacularly bad results... like the time the facilitator tried to get to the bottom of tensions between the CTO and head of sales (the classic problem of \"Engineering isn't shipping features fast enough\" and \"Sales is selling features before even consulting with engineering\"). The discussion quickly turned shockingly personal and within a month after the retreat both the CTO and head of sales quit. At least they called an end to the retreat early.\n— Johnny555\n\n
Decathlon Germany adds Wero payment option to decathlon.de website
\n\n\n\n296 points | 207 comments
\n\n\n\nWero is such a valuable asset for Europe and couldn't have come at a better time.\nIt's built on the excellent SEPA system, which standardized bank transfers and \"Lastschrift\" in Europe. SEPA itself got a huge upgrade last year with mandatory instant transfers that cannot cost more than standard transfers between European checking accounts.
Both of those allowed Wero to happen as a final abstract top layer, meaning: the UX of sending money to an email address with the security and convenience of bank transfers. And everything is built collaboratively on top of European banking infrastructure.
No account necessary and no private american conglomerates can delete that account.
And before some says: \"Well, now the government controls everything!\": They already did, since... Ever. Now its one less entity with that access and one that can be (hopefully) controlled and changed by a majority.\n— anticrymactic\n
Judge Rejects Google's Attempt to DMCA Its Way Out of Being Scraped
\n\n\n\n235 points | 91 comments
\n\n\n\nI'd be more OK with this if Google had a good API for their search results. But they've deprecated it, and now there is no alternative. So I'll continue to use 3rd parties that scrape Google results, until they change their mind.\n— binarymax\n\n
A missing underscore sent innocent man to prison for 18 months
\n\n\n\n244 points | 131 comments
\n\n\n\n> Nothing linked Klayme to the girl. No intimate images were found. Klayme did have a Kik account, but cops couldn’t even show that he had accessed the service during the period in question.\n> Still, Klayme was arrested and hit with three charges:
> Luring a person under 14 years of age by means of telecommunication
> Providing sexually explicit material to a child
> Possession of child pornography
> The case went to trial, where Klayme was found guilty. He then went to prison for 18 months.
What in the world happened in this case?
How can someone be convicted of these charges without any evidence?
Unless the article left something out, the only possible evidence they had was the wrong username. They couldn't even find evidence that this person used Kik at the time of the crime.
What defense did his lawyer even try? I'm so confused.\n— Aurornis\n
Benchmarking Opus 5 on SlopCodeBench
\n\n\n\n234 points | 55 comments
\n\n\n\nNice! I actually ran across this paper+benchmark recently, too. It's the first I've found that start to aim at some of the non-functional and longitudinal requirements that I think have always been an important part of writing production code.\nIt's especially relevant now that models are good enough to solve ~most point-in-time problems.
Some relevant but disconnected thoughts:
- deterministic scores are so nice
- what \"maintainable\" is is probably some high dimensional space described by these signals; it'd probably require some human labeling to figure out where this space is
- another signal I've been thinking about and I'm seeing increasingly get brought up is the state space of a system; I'm seeing formal methods pop up a lot recently\n— Vgoose\n
Using an open model feels surprisingly good
\n\n\n\n231 points | 68 comments
\n\n\n\nThis is an ad.\nI know the author has a long history here and I'm sure the post is a genuine reflection. But how is posting \"I used my own product and it felt really good\" anything other than self promotion?\n— mbanerjeepalmer\n
193 points | 104 comments
\n\n\n\nThe other approach to killing the cookie banner is simply to declare that such a thing cannot constitute “informed consent”. (Perhaps: “ticking a checkbox and/or clicking a button cannot constitute informed consent”; and see what they try next.) From a factual perspective, I honestly think that shouldn’t be controversial: it’s well-understood that very few people actually read those things, they just want to get them out of the way. Just as shrink-wrap licenses and even simple contracts have at times in some jurisdictions essentially been neutered, so that only terms that a reasonable person would expect to be there are enforceable, at which point it becomes obvious that the whole thing needs tearing down in favour of standard licenses/contracts. In the Australian state Victoria, for example, there are standard rent and property sale contracts; for renting you must use that contract <https://www.consumer.vic.gov.au/housing/renting/starting-and...>, and I got the impression that residential sales practically always use the standard contract.\nBut of course it’s impossible to convince someone of something when their livelihood depends on their not understanding it.\n— chrismorgan\n
US citizen charged after GrapheneOS phone wipes during airport search
\n\n\n\n555 points | 385 comments
\n\n\n\nI’ve seen a lot of people on the internet over the years say things like “the government can’t make x illegal, it’s just y.” For example, the government can’t make wiping your phone at the border illegal, it’s just punching four numbers into your phone, just like a pin, only a different four numbers, which could just have well been your pin.\nU.S. law though is highly non-autistic and what you were trying to do is just as important as what you superficially did. Hell there could have been a third set of four numbers that were the nuclear launch codes. It’s not the fact that it was four numbers, it’s what you were trying to make happen when you typed them. Now of course whether they can prove what your intent was when you typed them is another matter, but generally a duress pin should be for when robbers are breaking into your house, and the government will be on your side, and not when the government will be against you.\n— cameldrv\n
PGSimCity - How PostgreSQL Works
\n\n\n\n437 points | 45 comments
\n\n\n\nI love what is trying to be done here. Some feedback to improve would be to cut out a lot of the noise with the \"Take tour\". There are too many boxes and things changing on the screen to get a sense of what is happening. Also make it interactive rather than automatically switch onto the next subject. Being a passive observer to so much information trying to be conveyed is very confusing.\nI really do like these kinds of approaches to showing the inner workings of different kinds of tech, so think this is on the good path to being a very useful tool, but needs some focus rather than more data/graphs/info boxes.\n— layoric\n
Htmx 4.0, the first JavaScript library to release exclusively on the Game Boy
\n\n\n\n332 points | 104 comments
\n\n\n\nI've been using htmx for three years now, and it's completely unlocked new ways of building software for the web, especially if you use a server-side templating language.\nBut more importantly (as this \"Game Boy\" option reveals), the head honcho is really responsive with the items in the store. I had purchased a coffee mug a few years back and complained on Twitter that they need to sell larger mugs instead of only offering little tiny baby mugs that only hold ~8 ounces of liquid. The next day, he added giant 48-ounce monsters to the store and I've been drinking out of that mug every day since.\n— blister\n
French firefighters face 'pyrocumulonimbus' for first time
\n\n\n\n302 points | 185 comments
\n\n\n\nLe Monde, very reputed French newspaper (NYtimes equivalent for France perhaps) says that it’s not the first time this cloud is seen in France [1]: (translation mine)\n> “What is happening right now is extremely, extremely unusual. Seeing a cloud this massive and reaching such an altitude above a wildfire is exceptionally rare,” the researcher explains.
“During the Pedrógão fire [which claimed around 60 lives in Portugal in 2017], there were two episodes of fire-induced thunderstorms. Here, we are already at eight or ten, and the forecast is not encouraging at all.”
Although unprecedented in both intensity and frequency, the phenomenon has been observed in France before, according to the expert: during the Fontainebleau wildfire (Seine-et-Marne) in July, the Ribaute fire (Aude) in 2025, and the Générac fire (Gard) in 2019.
It seems that this cloud and the attached storms have been seen before, even recently, in France, but this one is one of a kind for intensity, by far.
[1] https://www.lemonde.fr/planete/article/2026/07/27/le-pyrocum...\n— wosk\n
Google Discloses $94.1B in SpaceX Stock, Marking 6% Stake
\n\n\n\n210 points | 143 comments
\n\n\n\nThis investment has never been secret.\nGoogle (now under parent company Alphabet) invested roughly $900 million as the majority of a ~$1 billion funding round alongside Fidelity Investments. This gave Google an initial equity stake of about 7–7.5% in SpaceX at a valuation of roughly $10–12 billion.\n— miohtama\n
Decker, a platform that builds on the legacy of Hypercard and classic macOS
\n\n\n\n288 points | 72 comments
\n\n\n\nI'm afraid that HyperCard is too old for broader audience to be able to grasp how extraordinary experience it was providing.\nAs a 6-7yo I had a collection of English words, stored with my own definitions and pronunciations, using basic building blocks, and it was as easy and intuitive as writing these words in my notebook.
People used the same building primitives to make real games or stuff like accounting applications.
And the best part - at that time it didn't looked cheap.
This one is really hard to describe - it's the same uncanny valley that makes small village in Fallout 1 feel massive and full of people despite it taking two screens with just few sprites running around, while full blown 3D world of Fallout 3, with real buildings and voice overs felt flat, soulless and empty.\n— zeroq\n
Kimi-K3 Releases on HuggingFace 7/27
\n\n\n\n301 points | 120 comments
\n\n\n\nThis will be interesting for a few reasons. First, depending on where the median pricing settles w/ 3rd party providers will tell us what it costs to serve a 3T model. Since it's going to be mxfp4 native, it'll take ~1.5TB of VRAM to host this, which is juuust at the limit of 8xb200s (but realistically you'll need 16x for context / throughput optimisation). Won't be cheap to host, but at least we should get some range of $/MTok for a 3T model. Then we'll be able to guesstimate if \"labs are subsidising tokens on API pricing\".\nAlso interesting to see what effort it will take to fine-tune this beast. The latest AISI benchmarks on cybersec place it above glm5.2, but still way way behind SotA closed models. Some fine-tuning might be needed here. Also, interesting to see if Cursor does another training round on it, to directly compare it w/ kimi2.6/2.7 fine-tunes (composer series) and grok4.5.
Also also, interesting to see if someone takes on distilling (proper distillation, w/ training the entire distribution) from this into smaller models. (dsv4-kimi should be really good, since dsv4 is very cheap to serve)\n— NitpickLawyer\n
London Gatwick has launched a robotic airport parking service
\n\n\n\n265 points | 226 comments
\n\n\n\nSo I'll have to park my car in a garage first, then take a bus to the terminal just so that a robot can move my car from the place where I parked it to a different place? I was hoping it'd be something where I'd pull up to the curb, grab my bags and head in while the robot parked the car for me but apparently I'll still have to take a bus to the terminal?\n— pontus\n\n
205 points | 189 comments
\n\n\n\nThe buried lede:\nWhat does all this mean for global temperatures? Because global temperature lags ENSO by around three to five months, most of this event’s warming will land in 2027, which is now shaping up to be a genuinely alarming year and the warmest on record by a sizable margin
I’m certainly no climate scientist, but if I lived in Europe, I would absolutely be looking into how to get a solar energy system set up and at least one window air-conditioning unit if not a heat pump before next summer\n— aaronbrethorst\n
197 points | 73 comments
\n\n\n\nAnecdotal, I know, but I was ghosted by Google back in 2004-2005 or thereabouts. Back and forth with a recruiter, who promised me that she would book phone screens. Then heard nothing. Pinged her repeatedly. Nothing.\nA few years later got another Google recruiter after me. Ended up getting hired in 2008.
Looked up the previous recruiter. Her last day at the company was the day she promised she would book phone screens, or the day after (can't remember exactly, this is quite a few years ago). I can only assume she was suddenly let go.
Ever since, I keep wondering how many ghostings happen because the person holding the strings disappears for some reason, and there's a lack of system to follow up on that.\n— arcade79\n
Open-weight AI is having its Kubernetes moment
\n\n\n\n295 points | 238 comments
\n\n\n\nEveryone is talking about banning Chinese models but nobody talks how it is feasible to ban them. I think it’s impossible simply because technically there is no such thing as a “Chinese model”. There is no way to tell apart an “American” model from a “Chinese” one by looking at their weights. Weights are just numbers and you can’t assign country of origin to numbers. One can find very easy workarounds to any naive attempt to ban them by origin.\nSo, any solution to this “problem” must include ALL open-weight models. As far as I understand this is exactly what they intend to do. Axios article linked in the post mentions that. As in this quote:
“The source described leading AI labs or their allies approaching the administration every 3-5 months with an idea to ban open-source models.”
It doesn’t say “Chinese” open-source models. Because they already know that it’s not feasible. Any regulation must cover all the models.
Now there are solutions for that latter problem. But they are all ugly and restrictive. Making a DRM-like license protection system mandatory can be a solution. If a company wants to run an open model in their own servers, they can only use approved and certified pure “American” models. This of course creates a monopoly for the big labs who are authorized to train and distribute such “open” models. A company can fine-tune the model for its own needs but of course can’t distribute the derivative model.
I’m sure there are other solutions but all of them would be equally ugly. Also these regulations can’t be enforced to other countries easily so only Americans will be restricted.\n— ozgung\n
The new rules of context engineering for Claude 5 generation models
\n\n\n\n278 points | 169 comments
\n\n\n\nWe should design a specific language to make sure that we can encode the exact requirements that we want. Something that has a limited set of keywords that are explicit. Wait a minute...\n— mycentstoo\n\n
The growing vigilante movement to knock out Flock surveillance cameras
\n\n\n\n165 points | 89 comments
\n\n\n\nFlock was pitched as the triumph of technology over criminality. It's surveillance network was supposed to end crime. However, when people witness extreme criminality thriving unpunished at the top levels of U.S. politics, the lie becomes evident. Flock isn't about ending crime. It's a tool of control to be used by criminals.\nWho watches the watchers? Everyone.\n— beloch\n
195 points | 120 comments
\n\n\n\nWalked around Fusion Festival with Bitchat running. Saw about 20 other devices and a few messages. Nobody replied to my messages. So usage isn't zero, but 20 devices out of 80000 people isn't a lot. Most hops I saw was 2, meaning there wasn't enough density to make it any meaningful distance across the festival grounds.\n— inigyou\n\n
189 points | 126 comments
\n\n\n\nSprites are a beautiful place to make an abstraction.\nBut I have never used a buggier infrastructure product in 30 years of development than when I tried sprites. Data loss left and right, sprites going into unconnectable zombie mode. Couldn't use even load snapshots because half the system seemed to think the sprite was fine, and half seemed to think it was dead. I gave up after two weeks of losing stuff over lunch or overnight or in the middle of working. I vividly remember scrolling through my terminal history, copy/pasting out work I could find to save from a dead sprite. I would guess that half or more of the sprites I launched had some kind of problem.
Hope they get things stable. The concept is great.\n— danielvf\n
159 points | 180 comments
\n\n\n\nEvery single knowledge worker is going to face this kind of crisis. It will be worse for people who cared deeply about their craft, but the crisis will have the same shape. The way through is to change the way you relate to your work, and to appreciate the fact that you can now produce far more output.\nMathematicians should realize: instead of spending decades on individual theorems, you can now create whole new subfields of mathematics. Of course, you won't be doing the creation directly, in the way the mathematicians of old did; that's just not the name of the game anymore. But you will still be directing the research and involved in the creation.\n— d_burfoot\n
199 points | 159 comments
\n\n\n\nHaven't heard of this company before. A good synopsis I dug up in their fundraising announcement:\n> JetZero, co-founded in 2020 by start-up veteran Tom O’Leary and aerospace legend Mark Page, is developing the world’s first commercial all-wing airplane. With up to 50% better fuel efficiency and lower carbon emissions compared to existing commercial airliners, JetZero's Z4 will offer the aviation industry a clear path to achieving its 2050 net-zero goals while also elevating the passenger experience. Working alongside the US Air Force, NASA, and the FAA, and backed by decades of investment and research into blended wing technology. JetZero looks to enter commercial service in the early 2030s.
Where an \"all-wing\" airplane is, kudos to Gemini,
> one that merges the main body and wings into a single lifting surface. This can reduce drag, lower fuel burn, and expand interior space.\n— wxw\n
A shell colon does nothing. Use it anyway
\n\n\n\n194 points | 66 comments
\n\n\n\nWell that's exciting. I learned a lot of uses for \":\" today.\nHowever, the only one I already knew...
\nI used to do that until I learned ofif some-command; then\n : # command required\n else\n echo \"command failed\"\n fi\n\nIt's in the POSIX standard so it's not just a bashism: https://pubs.opengroup.org/onlinepubs/9699919799/utilities/V...if ! some-command; then\n echo \"command failed\"\n fi\n> If the pipeline does not begin with the \"!\" reserved word, the exit status shall be the exit status of the last command specified in the pipeline. Otherwise, the exit status shall be the logical NOT of the exit status of the last command\n— amiga386\n
Show HN: Brolly, a plain-text weather forecast site
\nThe UK MET office recently redesigned their site, adding a lot of additional whitespace, scrolling, and animations. This significantly reduced its usability for me, and left me wanting an ‘at a glance’ weather site.I made https://brolly.sh, a minimalist, plain text weather forecasting site. You can use it to view weather from around the world, with: 7 day forecast; Previous day log (so you can confirm it definitely was cooler / hotter / wetter / drier yesterday!); Hourly rain, wind, temperature, conditions; Hourly UV, air quality and pollen, including pollen type specific forecasts within the EU / UK; Location search and last 5 locations; Location specific units.
I mostly made the site for myself, if anyone else also benefits from it that’s an added advantage.
You can check out the weather in York, UK at https://brolly.sh/forecast/RWFP2qW8, or search for a location at https://brolly.sh
The site is deliberately styled as a single long scrollable column, to work on mobile phones. You can view it on desktop too, there's just a lot of horizontal padding. I naturally took a lot of inspiration from plaintextsports.com. Despite not being a sports fan, I love its aesthetic. But, you'll hopefully see that this site isn't a rip off, and has its own deliberate look and feel.
Visualisations are really important to showing information at a glance. I spent a lot of time designing the different visualisations and making them work with only characters. My favourite is the hourly heat map used for pollen count.
It's also frustrating to have an interactive site, where you can't share a page with a friend and have them see what you're seeing. To solve this, all page state (i.e. location, selected day, expanded / collapsed sections), is stored in the URL. You can share or bookmark the specific view, and know that you'll always be able to come back to it.
The site uses PocketBase. It’s written in Go and plain HTML/JavaScript/CSS. All pages are backend rendered, with light JavaScript to handle re-loading content without page jumps when using interactive features like next / previous day navigation. Weather forecasts are fetched from open-meteo.com, which has a very generous free tier. However, I also built a custom LRU cache on top of PocketBase’s SQLite DB to cache forecasts for 5 minutes, and avoid putting unnecessary pressure on the open-meteo API.\n\n\n
171 points | 51 comments
\n\n\n\nThoughts about serving pure text (not HTML) so that someone can pull this using a curl command, awk the output and do various things with it? Wish more sites did that.\n— hliyan\n\n
1213 points | 660 comments
\n\n\n\nI think the most important thing here is not absolute performance. It's that organizations now have access to a Fable-ish model without Fable's 30-day data retention requirement[0].\n> \"Consistent with prior Opus models, Opus 5 does not have data retention requirements for general access.\"[1]
On the Opus model release page, the reason why Fable doesn't have an ARC-AGI score is because of that retention policy[2].
0: https://support.claude.com/en/articles/15425996-data-retenti...
1: https://www.anthropic.com/news/claude-opus-5
2: https://xcancel.com/arcprize/status/2064399134099153344\n— postalcoder\n
Nvidia, Microsoft, Meta warn against overregulating open-weight models
\nLetter: https://images.nvidia.com/pdf/Open-Weights-and-American-AI-L... [pdf]https://x.com/JensenHuang/status/2080643682408321103, https://xcancel.com/JensenHuang/status/2080643682408321103
https://www.wired.com/story/silicon-valley-is-completely-div..., https://archive.ph/LAkVI\n\n\n
464 points | 218 comments
\n\n\n\nhttps://www.anthropic.com/news/donation-public-first-action\nProbably because anthropic is pouring $40 million dollars into a political pact to regulate models. And they have not been quiet about wanting to ban/regulate OSS models either.
I have no idea why HN still treats them like ~their~ they are the ethical good guy.
Edit: I can’t English. At least we know this came from my dumb swishy brain\n— Robdel12\n
My security camera shipped a GitHub admin token in its login page
\n\n\n\n317 points | 120 comments
\n\n\n\nWorthy thread to ask: is there such a thing as a white label IP camera (or similar) with a supported open firmware? Not asking for open source, but something close to plug and play that nonetheless has a way of stripping the rootfs as desired for bespoke use in a manufacturer-supported way.\nI have looked around before but I only found genuinely dev-oriented kits that weren't even in a shell, and crazy priced.
edit: seems there are some options now (or I missed them before), e.g. https://www.goodcam.io/#for-independent-developers\n— badatnames\n
Be skeptical of OpenAI's rogue hacker agent story
\n\n\n\n385 points | 210 comments
\n\n\n\nThe article, summarized: there are incentives for OpenAI to claim that their AI hacked its way out of their network and into Hugging Face.\nOk... and what?
Do you have any evidence for the claim being either true or false that you'd like to write an article about? I guess not. Without anything to add, this article reduces to \"I has big brain and can see what you sheep cannot. Very big brain. Gullible sheep. Sucks to be you, sheep.\"
(For the record: I see the incentive. My guess is that it happened exactly as described. The main takeaways: (1) science fiction is now real, we should all be very afraid; and (2) OpenAI, which claims to have a God-given responsibility to get to AGI first to protect the world from danger, cannot be trusted with this foundational task even when it's in easy mode. The latter is true whether or not you believe in OpenAI's reasoning and purpose.)\n— sfink\n
Government orders GitHub to remove Bluetooth-based chat app Bitchat: Jack Dorsey
\n\n\n\n352 points | 262 comments
\n\n\n\n> the app enables communication even during network restrictions and creates a substantial risk of misuse by anti-national elements, terrorist organisations, organised criminal groups and cyber criminals seeking to evade lawful detection and continue communication despite legally imposed restrictions.\nThis mouthful boils down to exactly this: a form of communication not controlled by the government creates a risk for the country.
Sorry for pointing out the obvious, but if your country's safety depends on ability to block all forms of private person to person messages, your government has failed in one of its primary goals: providing safety for its citizens. They can blame Dorsey, Telegram, WhatsApp and what not, but it is their failure. Things will not change until people notice the obvious and vote out incompetent persons in power.\n— viktorcode\n
323 points | 277 comments
\n\n\n\nAuthor failed to explain the dash universe and their semantic assignments.\n- em dash: adjoining thought (Hey, so — <declaration>)
- en dash: expresses a range (1–100)
- dash: compounds word-parts (Sarbanes-Oxley)
I'd also like to see more concise philosophy around em v. ellipsis, \"limerance\" is poetic (and ironically antithetical to this \"essay\" … and what of parens?). In short, em is communication and ellipsis is personal.
I offer this feedback as a humorously offended fan of all sidecar prose (dash, ellipsis, parens, quotes, and meta typographical treatments) pre-AI era.
Finally, I'd love to hear from this author regarding modern use of period in modern comunication, with canonical use-case of bulleted list as example.\n— douglee650\n
Postgres LISTEN/NOTIFY actually scales
\n\n\n\n278 points | 49 comments
\n\n\n\n\"Scale\" isn't a binary, it's a continuum. \"Scales to 60K/s\" can be 5 orders of magnitude more than one system needs and 5 orders of magnitude too small for another. Personally I'd knock the general \"premature optimization\" off the list of \"most common developer errors\" and put in its place \"using techs with the wrong scaling factors\". If you use something too small and you exceed its needs, the failure is obvious, but the other way around is a problem too. Bringing in the overhead and management issues of the super scalable techs, as well as their limitations they impose so that they can scale, to a system that would actually be better off with a richer model whose richness prevents it from scaling but would save a lot of effort is also a bad choice.\nThe ceiling of LISTEN/NOTIFY is small enough that you need to pay attention, and I personally like to have at least an order of magnitude of slack left over even after my most pessimistic load numbers are accounted for, but it's still plenty for a lot of projects, and the integration with the rest of the DB, its availability, its not being another service you have to devops, it's definitely not something that should be simply dismissed out of hand as an option. Even the original 2K/s number they cite is a lot of messages for some systems that are more properly measured in seconds per message.\n— jerf\n
Half-Life 2 running natively on HaikuOS
\n\n\n\n249 points | 45 comments
\n\n\n\nX512 is an amazing treasure in the Haiku community. Other than nVidia hardware accelerated drivers, he ported Haiku to RiscV, got sound through HDMI and DisplayPort working, he got AMD Vulkan drivers for Southern Islands chipsets also working, made break throughs in dozens of other porting projects, he’s an amazing Hacker par excellance virtually unknown outside the Haiku community.\n— smallstepforman\n\n
Opus 5 is currently #1 on Artificial Analysis Intelligence Leaderboard
\n\n\n\n294 points | 161 comments
\n\n\n\nWhat's interesting is this:\nThe top AI models by Intelligence Index are: 1. Claude Opus 5 (Adaptive Reasoning, Max Effort) (61), 2. Claude Opus 5 (Adaptive Reasoning, Xhigh Effort) (60), 3. Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback) (60), 4. GPT-5.6 Sol (max) (59), and 5. Claude Opus 5 (Adaptive Reasoning, High Effort) (59).
Which means Opus5 at Xhigh is still smarter than Sol at max, and Opus5 at High is equal to Sol at max.
That would make Opus5 High same as Sol max, and now I wonder what the price and speed difference between those is?\n— didibus\n
Writing by hand is good for your brain
\n\n\n\n317 points | 125 comments
\n\n\n\nA corollary to this is, when reading books, mark them up! Don’t be precious with them. Destroy the spine so it’s easier to read, fold pages, write the important stuff inside the front cover of the book. Cross out stuff you disagree with, draw a big circles around the stuff that resonates.\nNot only does it make consumption a lot more engaging, it makes a revisit of that book extremely rewarding.
Don’t get caught up trying to figure out the right notation or highlighter color for a line. Just mark them up and later it’ll make sense.
Edit: I think the broader point that some of you may be missing is that I’m not prescribing a particular form of notetaking. I’m saying that reading is made more rewarding with active engagement. How you best engage is up to you.\n— postalcoder\n
Startup founders urge U.S. government not to shut off Chinese open weight AI
\nhttps://littletech.org/https://static.politico.com/4a/bf/9c4021d8404386b0a311dcccf0...\n\n\n
683 points | 612 comments
\n\n\n\nI‘m not even sure what the argument for banning Chinese models/open weights even is supposed to be?\n1. if it’s to stop hackers doing hacking things with „uncontrollable models“ then, well… they’re already doing something illegal to begin with, why would they care about breaking another law running these models?
2. if it’s to stop foreign actors, then that ban would not apply to them anyway
3. it’s not stopping distillation either, Chinese labs are already banned from using US frontier models and look at how good that is working
I don’t get it. Am I missing something? The only thing a ban would do is protect the American market from further downward price pressure on inference, protecting VC investors in the short term. But thats also an admittance that the American labs can’t compete on merit anymore, and should by itself also limit the viability of the idea that all those VC billions will ever make a return? In any case this would be something benefitting only a very few for a short time (labs + investors).
Someone please enlighten me what the actual argument here is, cause I can’t see it.\n— capevace\n
AI Companies Are Trying to Hide a Staggering Amount of Debt
\n\n\n\n318 points | 146 comments
\n\n\n\nAs long as this debt does not make it into life insurance and pension funds, we are fine. The trouble is that private credit is taking control of some life insurance companies and off-loads this debt to these. When these fail, it will become everyone's problem.\n> Risks to financial stability may also stem from entities with particularly high exposure to private credit markets, such as insurers influenced by private equity firms and certain groups of pension funds. The assets of private‐equity‐controlled insurers have grown significantly in recent years, with these entities owning significantly more exposure to less‐liquid investments than other insurers
https://www.imf.org/-/media/files/publications/gfsr/2024/apr...
https://www.imf.org/-/media/files/publications/gfsr/2024/apr...\n— senshan\n
420 points | 100 comments
\n\n\n\nAuthor here! This was my burnout recovery project and holds a place near and dear to my heart. https://notes.jordanscales.com/98-css-reflections\n— jordanscales\n\n
The incoming club leader was persistent though, and called NameCheap support. He convinced them the domain registered in my name and address really belonged to his club, and with no verification or validation whatsoever, NameCheap changed my password, and changed the email address associated with my account. All because someone simply asked nicely on a phone call.
Meanwhile in the background, someone advised the new club leader who I was and we were able to connect and get things transferred over. Ultimately I was happy to give them access or even ownership if they wanted (student club turnover being what it is, it’s likely a domain doesn’t get renewed and gets gobbled up by a squatter, which is why I was keeping it current for them).
But NameCheap had no way of knowing any of this. As far as NameCheap was aware, this was a personal account of mine. They demonstrated they were perfectly able to pick up a phone and call me (to verify my initial support ticket) but when someone calls them and says “but I really want access to that account” they don’t bother?
I’d hesitate to even call this social engineering. It’s clearly a massive vulnerability. I’ve already moved a dozen of my most critical domains out of NameCheap after seeing just how easy it is for a third party to completely take over a NameCheap account: just ask nicely.\n\n\n
320 points | 114 comments
\n\n\n\nNamecheap has been owned by a private equity firm for several months now.\nIt would be nice to have a nonprofit registrar so jumping every few years isn't necessary.\n— pilingual\n
It started with a simple experiment. I took a group of models, including GLM-5.2, Kimi K2.7 and others, and ran them on the same evaluations. Then I measured what would happen if, for each problem, you somehow knew in advance which models would be useful and how their outputs should be combined.
That hypothetical system performed substantially better than any individual model in the pool. Of course, it is not something you can actually deploy because it relies on knowing which decisions were good after seeing the result. Echo is my attempt to recover some of that advantage without having that information in advance.
For each request, Echo decides how much computation to allocate, which models should participate, and how their work should be combined. Some prompts may only need a relatively small amount of inference, while others benefit from multiple models working on different parts of the problem.
One thing that surprised me while building it was how complementary the models are. A model that is clearly weaker overall can still be extremely useful on particular problems or as part of a combination.
On my first evaluation mix, Echo consistently performed better than the best individual model in its pool. It also reached roughly the same aggregate result as Fable, which I used as one of the stronger comparison systems, at around one third of the inference cost.
There are still some cases where Echo makes the wrong allocation or combination decision. I’m currently spending a lot of time understanding those failures, as well as testing whether the same approach holds up on coding and agentic tasks where measuring the quality of each decision becomes much harder.
I built a chat interface (echo.tracerml.ai) and an OpenAI-compatible API (https://echo.tracerml.ai/docs/api) so the system can be tested outside the evaluation setup.
Here is a short/high level video on how it works: https://www.youtube.com/watch?v=lJFJSvOdXhg
I wrote up the evaluation methodology, individual model results, costs and current limitations here: https://echo.tracerml.ai/eval
I would love for you to try it! Especially if you hit any weird failure cases or places where the allocation looks unintuitive.\n\n\n
345 points | 166 comments
\n\n\n\nThe business use case of rather than going on one model choosing the best openweight model and reducing the cost seems fascinating, however the context memory, or auditability of what is happening behind would be more complicated, even within single model we have to spend tons of time to decode and understand llm behavior , add context layers and so on, having said that for GenAI executions this might be the direction. Wish you best for the project\n— Arshad-Talpur\n\n
What happened to TheNumbers.com
\n\n\n\n268 points | 117 comments
\n\n\n\nA couple years ago, I was running a website which allowed the public to view all the US government handouts to small businesses during the COVID-19 pandemic. It also tracked all the fraudulent loans being prosecuted by the DOJ, and allowed anyone to run structured queries over the public dataset. There was a \"donate\" button which took in ~$2k in donations over the lifetime of the site, and you could download the entire underlying dataset (around 10 GB uncompressed) for free directly on the site.\nDespite the \"download all data\" link being prominently placed on the front page, the AI scrapers decided it would be more efficient to download terabytes upon terabytes of raw HTML by paginating through every possible facet on the search endpoint. Even with CloudFront caching results and a fairly efficient backend setup, the monthly bill ended up with around $1k just going toward network ingress/egress, so I shut down the site the following month.\n— primitivesuave\n
Couple pay >$800k for a gene-editing therapy for their daughter. She died.
\n\n\n\n235 points | 139 comments
\n\n\n\nI'm honestly quite shocked that the physicians/scientists involved would choose to use an AAV for a brain-targeted gene therapy. There is just so much data demonstrating that these vectors are quite immunoreactive: most of the approved gene therapies based on AAVs carry black box labels for liver failure caused by an immune reaction to the viral capsid. Admittedly, AAVs are the most derisked vector for gene therapies, but infusing them directly into someone's brain and expecting nothing bad to happen is, in my view, crazy.\n— maxall4\n\n
Why Software Factories Fail (or: harness engineering is not enough)
\n\n\n\n273 points | 197 comments
\n\n\n\nI call it the Intent-Implement-Quality problem.\nSoftware factories can implement anything given a one-liner requirement. That one-liner requirement can be a complete app/product, epic, feature, bug, design change or refactoring. But these one liner requirements are requirements coming from a human who has an intent or requirement or direction for the product to evolve in mind. Can Software factories manufacture intent that reflects the exact requirements of the person using it or their vision of how the product or software should evolve?
Implementation in any language is as easy as generating a summary for an LLM. If coding is only math and there is only one way to translate a requirement into an implementation, the problem boils down to just providing or verifying generated intent. But, there is definitely more than one way to implement a thing and one of the ways leads to a design and architecture that is coherent with rest of the system, design and architecture that is extensible and evolves, code that you can keep in your head when you want to change things next time or tomorrow, software that can serve millions of users and securely guarantee millions of dollars of revenue. There is a combinatorial explosion of ways and the one right way is also subjective of the person and the problem. Software factories can of course improve Quality by generating more unit tests, more integration tests, fix code violations, generate Proof of work as videos, screenshots etc, but there is no feedback loop or test suite to verify and correct the other subjective notions of Quality.\n— sathish316\n
I regret migrating to Codeberg
\n\n\n\n303 points | 205 comments
\n\n\n\n> In addition, Codeberg conflates “having a community” with “being legitimate software worth hosting”\nI don't think this is a fair summary. I don't see this conflation in Codeberg's blog post. The words in quotation marks don't actually exist in Codeberg's blog post.
It seems that what they are saying is: In a free hosting platform where we share all the resources, it's not fair that some people using LLMs are using up disproportionately huge hosting resources to handle the unnaturally large amounts of resources they can churn out with these tools.
I'm a solo dev, and I couldn't find the judgement against solo developers or any slight against people who don't have a team. The concern is that some people don't gobble up an unfair amount of resources.
I personally have been moving to self-hosted Foregejo for ultimate freedom. But I also think it's fair for some providers have the freedom to take a bit more of a harsh (or based - depending on your perspective) stance against LLM generated code. I mean, a lot of people have been bemoaning the fact that the human-generated material has been increasingly hard to find in a sea of AI-generated material. It could be nice to have a corner of curated, human-made code still available on the web.\n— adamddev1\n
Terrence Tao's ChatGPT Conversation about the Jacobian Conjecture Counterexample
\n\n\n\n544 points | 339 comments
\n\n\n\nWe had this yesterday (317 points, 133 comments)https://news.ycombinator.com/item?id=48998362\nClassic Gabe-bot\n— gnabgib\n
Show HN: Bento - An entire PowerPoint in one HTML file (edit+view+data+collab)
\nOver the past few months, our team has been building more and more slidedecks using web frontend technologies with coding harnesses like Claude Code, but a common complaint is to make even small edits we need to edit the code either manually or via the harness.To avoid this loop, I ended up creating Bento, a single HTML file with everything you need in a slide tool including animations and shared editing. There's no install or cloud login, everything works offline. The default deck is around 560 KB and it doesn't need to fetch anything once you got it.
Open it in a browser and then you can edit, present, print and save. Share it via email or via Airdrop and all they need is a browser to edit, present and also do live collab on the slides. Drop it in to Claude or ChatGPT to transform existing pptx files into Bento slides. There is no cloud involved, only an encrypted blind relay to allow for shared editing. The relay doesn't see any of the data.
Check it out at https://bento.page/slides/ which takes you straight to the editor.
Go to https://bento.page/guestbook/ to try out the live guestbook to experience share editing / collab.
There is also a gallery with some sample decks on the website - https://bento.page/
All the code is MIT licensed and you can find it here - https://github.com/nyblnet/bento . I used reveal.js with several other libraries (including some homegrown ones), and Claude Code.\n\n\n
602 points | 142 comments
\n\n\n\nHi, I'm the creator of Bento. Just wanted to share a bit more about how I created it beyond what's in Github.\nThe file contains more or less two sections. There is a plain block of JSON near the top of the file which is the slide data. You can read, grep, or point a harness at it. The app itself is in a base64 blob that loads through a small shim which deflates in the browser with DecompressionStream, which keeps the package small and so that we don't need to fetch any external files at runtime. File System Access API is used for JSON writeback into the same file, which falls back to a plian download and all updates are ECDSA signed.
I started with reveal.js which served as a base, and incorporated GSAP/Flip to handle the animation. Then I added charting with echarts, but due to a number of issues with size and how the worked, I ended up re-implementing both.
What I'm most pleased about is how seamless the CRDT works. The blind relay is a small file that runs on Cloudflare Durable Objects, and all it sees are the encrypted data from the clients. Collab is off until we turn it on in the app (by starting a session by sharing by invite file), and then some sets of keys are generated. Access is handled by user key, and you can have read-only users, and also revoke live collab access by user as well.\n— starfallg\n
https://www.twit.community/t/john-c-dvorak-has-passed-away/2...
https://www.facebook.com/aric.mackey/posts/john-c-dvorak-194...\n\n\n
466 points | 147 comments
\n\n\n\nAww brother... out of many recent ones, this one is a tough one for me. I used to read his columns in Infoworld and Mac User before the blogs. He was a frequent guest on TWiT, but it looks like he and Leo had a falling out. No Agenda too, but that took a turn a while back when Adam found religion. I always relied on his (not always PC) cracks to lift my mood.\nRIP my mentor that you didn't know you were, Valhalla awaits.\n— stasomatic\n
OverpAId – Fire your CEO. Hire the future
\n\n\n\n598 points | 300 comments
\n\n\n\nI have learned that the purpose of a CEO in many (most?) companies is not to actually run the company.\nTheir purpose is to be the human delegate that represents the corporate entity. They are empowered to act on behalf of the entire organization. This is important because a corporation of many individuals generally cannot act as one. You can sometimes get close with really, really good teamwork, but that's still not enough for a lot of customers.
If you are in some regulated B2B context, not having a CEO or having a \"non-traditional\" CEO can result in lost deals. The counterparties are looking for \"one throat to choke\" when they are being made responsible for approving large contracts. They want a person's cell phone number they can call when they're having a bad mood about whatever they've just paid for. They don't want to talk to your developers or the cranky old founders. They want to talk to a person who can be summoned to any arbitrary golf course within 48 hours and who will arrive in a fantastic mood 100% of the time. This person often doesn't have to know a goddamn thing about what your company's product actually does. They just have to be a delegate. A conduit of information and a representative element.\n— bob1029\n
350 points | 138 comments
\n\n\n\nThis is fantastic\nI've been casually spot-checking other animals in other vehicles, because my absolute dream situation here is to catch an AI lab that's demonstrably better at pelicans on bicycles than other combinations.
Catching a lab cheating specifically on my one dumb benchmark would be really funny.
Dylan's methodology here - generating 1008 SVGs across an 8x6 combination - is significantly more robust than anything I was considering.
His conclusion:
> Nothing jumped out at me. I couldn’t find a case where the pelican-bicycle images looked noticeably better than the rest of that model’s grid.\n— simonw\n
Passkeys were invented by engineers with zero understanding of consumer brain
\nhttps://xcancel.com/nikitabier/status/2079787406300266743\n\n\n436 points | 604 comments
\n\n\n\nI do not know how to use a Passkey in a way that won’t impede how I log in to systems. I’ve been in tech for 26 years, and I understand the Public/private key behind what a Passkey is. Here’s what I don’t understand:\nI access a website through at least four different devices (my iPad, iPhone, Windows Desktop computer, and MacBook Pro) and three different browsers on each device (Brave, Firefox, Safari) , and I use LastPass. If I accidentally set up a passkey on my phone (let’s say I use Safari one day instead of my go-to, Brave), can I still log in without that passkey on other devices? Is there a way to ensure that passkey can be used on other devices? Can I add another passkey on another device? How many passkeys can I set up for a particular site/app? I have at least 6 different combination of browser/devices in use.
I don’t want to use Passkeys because I don’t the answers to those questions, and I don’t know whether each website/app that has set up Passkeys has decided the answers to those questions in the same way as the others. For now, I’m going to stick with LastPass and use Passwords; because no matter whether I lose my device or not or whether I’m on my own devices or not, I can be sure I’ll be able to get into a site/app.
Edit: One final consideration, my spouse and I share user/name passwords for some things (notably Pandora and our Amazon Prime account) since they don’t handle things like family logins well; how do both my wife and I use amazon or Pandora with passkeys? Do we each set up passkeys? How do I get her Pass if that’s not an option?\n— gortok\n
GigaToken: ~1000x faster Language model tokenization
\n\n\n\n331 points | 63 comments
\n\n\n\nInteresting :\nQ: Did you just way over-optimize for a specific CPU and tokenizer? How is it so fast?\nNo, I way over-optimized for every combination of these! The results are very consistent across CPUs (modern x86 and ARM), and across specific tokenizers.
The major improvements are in optimizing heavily an implementation that usually is outsourced to a Regex engine (pretokenization) using SIMD, minimizing branching and other tricks, as well as heavily optimizing caching of pretoken mappings (if a word has been seen before, look it up its encoded tokens efficiently). Caching is a very hard problem in this domain since the cache grows very quickly, and pretoken distributions are very long-tailed.
Finally, interactions with Python are minimized, and threads have minimal interactions with each other.\n— maxdo\n
So Reddit has decided that plain HTML is unsafe
\n\n\n\n397 points | 390 comments
\n\n\n\nI refuse to log in to Reddit, deleted my account ages ago. And now on mobile it places an unclosable pop-over asking me to log in to continue reading. So now I only read things on desktop. Would like to stop using Reddit all together but sometimes it is genuinely useful to read some human answers, although since LLM’s train on the data I can get a lot of the same info that way at least.\n— wktmeow\n\n
367 points | 119 comments
\n\n\n\nGood article!\nI just wouldn't start off with bold sentences as
> SIMD can be simple to understand
and
> writing SIMD is just about as easy as a for loop
and then the first example requires 12 lines to replace one line of scalar code.
Be honest and say SIMD is hard but the results are worth it!
(Another nitpick: if this article is for newbies, don't use SIMD-only words and concpts before explaining them. Step 5 is good: scalar tails are mentioned and described. Step 1 is bad: nobody is supposed to know what broadcast mean.)\n— teo_zero\n
The startup's Postgres survival guide
\n\n\n\n298 points | 163 comments
\n\n\n\nShould one of the first things you do with a database not be to have a backup strategy? I understand that HA would be a \"nice to have\" when first starting out, but surly if you have a production db, a backup and restore plan should be on a survival guide? Neither appear to be mentioned here.\nWhat do you all use for your pg backups? Is Barman ( https://pgbarman.org/) still the way many do it? (I haven't deployed a new pg instance for a while, but thinking about it for a new project).\n— theallan\n
OpenAI and Hugging Face address security incident during model evaluation
\nhttps://www.axios.com/2026/07/21/openai-says-hugging-face-br...See also Security incident disclosure – July 2026 - https://news.ycombinator.com/item?id=48956248 (9 comments)\n\n\n
585 points | 386 comments
\n\n\n\nI don't know if OpenAI thinks this is a marketing / PR angle for them (our super smart AI cheated on a cyber capabilities test in the most _brilliant_ way) but my read is this:\nWhy should OpenAI (or any frontier lab) be building these systems if they can't get a secure environment / containment right? It sounds like there was little defense in depth, appropriate monitoring, or any attempts to have their super smart model check for vulnerabilities in the test environment _without exploiting_ them. That seems like step 0 before trying to test offensive, unknown capabilities.\n— netinstructions\n
641 points | 452 comments
\n\n\n\n7 years ago we started the 'you are not the product' movement for digital information consumption with Kagi Search. Since then the paradigm shifted from search to agents. Good to see the need for 'you are not the product' still holds up, maybe even more so now: an agent is only worth trusting if it works for you and nobody else. The second an advertiser gets between you and the answer, that's gone. So we're staying on the user-funded side of this. Boring, but it's the whole point.\n— freediver\n\n
Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber
\n\n\n\n260 points | 194 comments
\n\n\n\nGoogle somehow managed to snatch defeat from the jaws of success with their AI products.\nThey literally forced me and my company out of Antigravity by phasing out AI Ultra subscription without any proper product follow-up. Antigravity IDE cannot even have poweruser subscriptions now from Google Workspace an Gemini Enterprise Agent Platform cannot be attached to Antigravity IDE.
Gemini Enterprise Agent Platform has an incredibly abysmal setup process, and if I want to limit spending per-user I have to create projects per user. The fact that you cannot activate Anthropic models on it if the billing still has free credits is almost a joke.
I was a big proponent of Google and Gemini, but they left us reeling with their abrupt product decisions. Forced us to buy $200 subscriptions directly from Anthropic/OpenAI.\n— stonewhite\n
Kimi K3 Is Competitive with Fable; Kimi K3 and Fable Is SoTA
\nAlso: Kimi K3: second only to Fable 5 on AA-Briefcase\n https://artificialanalysis.ai/articles/kimi-k3-agentic-knowl...\n\n\n529 points | 292 comments
\n\n\n\nIf you haven’t been really running and testing these models yourself, they are all benchmaxxed. No matter how close they score to frontier on whatever metric, they always fall apart in real world tasks and their token efficiency is ridiculously bad. Fireworks has incredible incentive to make this claim in a headline, because Fireworks hosting K3 for you is pure profit for them, unlike when they host closed source models.\n— nxtfari\n\n
FreeInk: Open ecosystem for e-readers
\n\n\n\n359 points | 83 comments
\n\n\n\nWasn't the efficacy of the focus/bionic reading stuff disproven awhile back? Are people still using it?\n— easterncalculus\n\n
'VPNs are lawful technical tools,' says EU Court in landmark copyright ruling
\n\n\n\n327 points | 58 comments
\n\n\n\nFor those who (like me) missed the essential key word in the OP's title:\nThis ruling is about copyright, and seems mostly unrelated to recent conversations about escaping censorship and surveilance, and the overall final throes of the open web.
It's still pretty important since the lawsuit filed by the Anne Frank Fonds could be caricatured as \"if this content is illegal in our country, it should not be on the Internet in other countries, because our citizens can still circumvent the publisher's geofence and access it\" which would have wild consequences if that line of thought was applied.
But, still:
> At its core, the case forced Europe's top judges to answer a highly technical question: if a publisher actively tries to block visitors from a specific country, are they still breaking the law if a user sneaks past the digital border using circumvention software?
> According to the CJEU, the answer is no.
Could be conceived to have a tenuous link to the current debate on censorship, surveilance, etc.\n— sebastiennight\n
Apple defeats liability for not scanning iCloud for CSAM
\n\n\n\n323 points | 296 comments
\n\n\n\nMaybe my perception is off, but it seems like there's a huge push by the legislature and some people to do anything and everything to prevent CSAM, yet almost nothing seems to be done to prevent CSA.\nFor CSAM, there's all sorts of monitoring, scanning, identify capturing, etc. But it's all after abuse has taken place, and it seems that many of the people actually arrested are arrested for CSAM and not CSA. This has even extended to fictional CSAM such as AI generated stories and pictures. As an aside, if that gets extended to political speech or other non-CSAM materials that are determined to be undesirable, that's a big concern. I can imagine that a conservative state could pass a law banning all porn because they claim it could encourage illegal activities such as prostitution, rape, or CSA.
On the CSA side, you rarely hear about arrests (they happen but less than CSAM). There doesn't seem to be any real push for educating and protecting kids before it happens. Ironically, the groups doing the most to educate and implement protective strategies are the ones who have been involved in abuse scandals in the past (Churches, Scouts, etc). Even then, a lot of it is just getting clearances, which doesnt prevent people who where not caught or were first timers. Offenders get put on a list/map. This is sort of a half approach. If they are still a threat, they shouldn't be released. Yet if you comb the list and see some of the results, they don't all seem to fit with CSA. I personally know of 1 who took a leak across from a playground at 2am walking home from the bars and was put on the sex offender registry because it was within 500'.
It seems like these laws are more about peddling to the publicist and lawmakers fantasy of incrementally extreme punishment rather than taking a appropriate, data driven, and level-headed approach that actually protects kids. Otherwise they will just keep pushing ham-fisted low-hanging \"fixes\" like required scanning and IDs to access the internet.\n— giantg2\n
Judge approves $1.5B Anthropic settlement for pirated books used to train Claude
\n\n\n\n362 points | 289 comments
\n\n\n\nA one time payment like 1.5B doesn’t do anything. There needs to be a royalty payment based on if the AI regurgitates existing ideas. That is probably the correct way to legislate this. If anything a human does can instantly be copied by an LLM, and then sent to all its subscribers, things need to change\n— sashank_1509\n\n
Long presumed dead, a thriving coral reef is discovered in West Africa
\nhttps://www.frontiersin.org/journals/marine-science/articles...\n\n\n285 points | 52 comments
\n\n\n\nNice to read a paper that's looking for paths of persistence instead of only documenting decline. Climate stories often end with \"thing are getting worse\", this one also asks where ecosystems might still persist if local conditions are managed well.\n— rendonroman\n\n
Jack Dorsey launches Buzz to combine team chat, AI agents and Git hosting
\nhttps://x.com/jack/status/2079605800998146171, https://xcancel.com/jack/status/2079605800998146171https://buzz.xyz/\n\n\n
299 points | 256 comments
\n\n\n\n> Explain Like I'm 5\n> Buzz is an open-source, self-hosted workspace that combines team chat, AI agents and Git hosting using signed Nostr events, so teams can keep control of their data.
Very different expectations of 5-year-olds.\n— rsstack\n
China’s open-weights AI strategy is winning
\n\n\n\n920 points | 750 comments
\n\n\n\nThe lesson of the last 50 years of the computer and software marketplace is that free and low-end eventually wins.\n- PCs destroyed minicomputers. Mainframes survive, but serving a much tinier portion of the market than they used to.
- PC office productivity software destroyed expensive professional products.
- Windows (low end) and Linux (free) completely destroyed the UNIX marketplace, and again, have taken huge market share from the mainframe world.
Ignoring the huge Chinese open-weight models for a moment:
- The training costs and resource requirements for frontier models are unsustainable. The high price, and social pushback, mean that the American companies producing these models are precarious.
- There are enormous financial incentives for research results allowing for cheaper, less resource-intensive models of high quality.
- Local LLMs on consumer hardware are akin to the PC hobbyist world of the 70s and 80s.
Put all of these trends together, and I think that in 10-15 years, we are going to have consumer PCs (and phones!) running models doing pretty much anything that frontier models can do right now.
Getting back to the Chinese models: They allow for new competition against Anthropic and OpenAI, basically SaaS renting out these very capable AIs much cheaper. That will just accelerate trends.\n— geophile\n
384 points | 97 comments
\n\n\n\nI was only thinking about Flight Control[0] the other day and how I wish there were more modern takes on it! There was Mini Metro[1], but that was released over a decade ago.\n--
[0] - https://en.wikipedia.org/wiki/Flight_Control
[1] - https://en.wikipedia.org/wiki/Mini_Metro_(video_game)\n— sarreph\n
Hacker wipes Romania's land registry database
\n\n\n\n327 points | 185 comments
\n\n\n\n> Since the hack, officials restored their website and posted a message announcing they are rebuilding the agency's entire network from scratch. Even if the hacker claims they deleted backups, the agency appears to have had an offline copy, otherwise things would have gotten really messy over the coming months in Romania.\nSo it seems not all has been lost. I was worried about the societal implications of being unable to prove land ownership but it seems that may be avoided.\n— skinfaxi\n
Who's afraid of Chinese models?
\n\n\n\n528 points | 359 comments
\n\n\n\nThe people who are most afraid of Chinese models are the VCs who poured into Anthropic and OpenAI at astronomically high valuations. Anthropic is valued at $1.2T and OpenAI is targeting $850B. These astronomical valuations were built on the premise that these labs would generate massive profits from premium API pricing, but the Chinese labs are completely undercutting this strategy by releasing excellent open models for free. If the frontier labs are forced to cut prices and join the race to the bottom in token prices, these valuations are unjustified, and VCs will face enormous (paper) losses.\n— tristanj\n\n
343 points | 163 comments
\n\n\n\n> Kimi Work is a Local Agent designed for deep workflows. It mounts your local folders, navigates the web autonomously via WebBridge, runs Python code in the background, and executes scheduled tasks.\nIt's clearly a dupe of Claude/Codex products (Codex especially, styling-wise), but I think Kimi's goal here is simply to appear on feature-parity with bigger labs. I doubt they'd want to invest much into designing a UI for \"the future of agentic work\" from the ground up.
Their bread and butter and claim-to-fame will continue to be low-cost, near-frontier models, for which they need to at least appear to have some app layer to plug into.
And, in some ways, duping is a great way to show people that even the app layer can be commoditized.\n— wxw\n
The EU is about to sell our most sensitive data to the US for visa-free travel
\n\n\n\n340 points | 191 comments
\n\n\n\nThe point missing in the article is that, in order to enter the US, one needs to give the US Govt their photo and fingerprints (\"biometric data\") anyway.\nThere are two options: visa-free with the biometric data passing on the wire (no hassle), or visa (a major hassle) and biometric data passing in the airport (a minor one).
As a side note, the EU also collects photo and fingerprints from visitors now. Whereas in the US you only need to go through fingerprinting only once (or maybe \"once per passport\") and only on arrival, in the EU you need to go through it both on every arrival and every departure.\n— maratc\n
Jelly UI: Soft-body physics for native HTML form controls
\n\n\n\n296 points | 123 comments
\n\n\n\nSaw the window was a bit laggy so thought I'd delve deeper to figure out what was going on.\nThis lib runs a a RAF animation loop every 8ms across every component on the page that causes the entire document to repaint.
The comment above appears of the AI generated sort:\n// One shared animation frame: step every live component, park when idle.\n// The delta is capped so a background-tab pause never becomes one giant step.
The \"JellyEngine\" instead of just calculating the jelly animation on pointer events is recalculating for every active component on the page every frame.
This is the kind of thing usually a human notices and says 'dont do that, that is kind of crazy'.\n— jlukic\n
Flock Credibility Lost as It Repeatedly Lies to City Councils, Police, & Public
\n\n\n\n326 points | 69 comments
\n\n\n\nPeople always want high trust societies. They are incredibly hard to create and incredibly easy to destroy.\nHowever, once the cameras go up, you will never ever create one there. High trust requires, well, actual trust. You can’t build real trust in a panopticon just a sickly police state facsimile of it.
Society and progress moves at the speed of trust. Without it many uniquely human and beautiful things disappear.\n— pclowes\n
Annoying and alarming things about OpenCode
\n\n\n\n303 points | 207 comments
\n\n\n\nI feel that a better title for this article would be: \"Some minor annoyances that, when fixed, would improve OpenCode\"\n# Prompt Cache Misses
> It globs your filesystem and re-reads AGENTS.md (injected in turn-0 system prompt) on every SSE turn. If you put a quick note in AGENTS.md to be read in the next session, you immediately force a full re-evaluation.
> Personal favourite: it puts the current date in the turn-0 system prompt and re-evaluates every SSE turn. If you’re using OpenCode at midnight you get a full prompt cache miss.
Okay, I can live with those.
# Compaction
> \nWant to sit for 10 minutes while the LLM server prefills the entire session with a new prompt prefixed to it, just to turn it into 5 bullet points that go at the top of a new session? Me neither. I get what they are going for, but I’ve not seen it work well. Neither compaction nor pruning is implemented well, and they interact poorly.
Is this an OpenCode specific issue? I've seen the same with Codex and Claude
# System Prompts
> The default system prompt is opinionated (fine) but it has shit opinions (not fine). It took me a while to figure out why my agent kept saying “Use ABSOLUTELY NO COMMENTS” when dispatching subagents.
Okay, so change it? Any LLM is opinionated, this system prompt enforces consistency across different models which seems reasonable.\n— LaurensBER\n
Human mathematicians are being outcounterexampled
\n\n\n\n302 points | 120 comments
\n\n\n\nWhen I was in grad school, I had the opportunity to take a course from my adviser in which he discussed his current research and some open questions. It was a relatively accessible subject area and the questions were sometimes easy enough that we could meaningfully contribute.\nOn one particular Friday afternoon, he stated a conjecture that he hoped was true, and invited us to try to help him prove or disprove it. It was the sort of thing that he really wanted to be true; he liked things smooth and beautiful. I, on the other hand, hoped it was false as I like the weird and exceptional in mathematics. It was also the case that I had absolutely no command of the sort of machinery that one would use to prove such a thing, but I could certainly look for a counterexample.
I learned on Monday that he had spent the entire weekend trying and failing to prove it. I, on the other hand, had put all my energy into finding a counterexample and had one within an hour.
My single (quite small) contribution to mathematical research was a counterexample because it was all I could do. The story does illustrate that it can be helpful to have people with different tools, hopes, and motivations working on a problem, though. I was not, and will never be, even a shadow of that great mathematiciam I studied under, but on that occasion, I had reason to look in a different direction than he did.\n— Dove\n
My family and I bought an abandoned 8-lane bowling center in the rural mid-west. In our small town there weren't many recreation options for families. You've heard of a food desert? This is an R&R desert.
It had been abandoned for a good reason. The roof leaks, the electrical system was constantly surging, and my 70-year-old bowling equipment (still) doesn't work perfectly. The system that keeps your score is particularly interesting to me. It's the thing you watch during your game, but it fades into the background beyond that. Turns out these things are really cool, but absurdly expensive.
Ours was installed in 2008 and cost six figures. It's calculating ball speed and trajectory, camera-based pin detection (object detection and trig, on ICs!), runs the fouling, the animations, the pinsetting machine and ball return. Very cool stuff for its age.
From the business perspective, my facility only cost me $105k. To forklift-replace the score keeping system runs anywhere between $80-$120k, depending on features, vendor, and unit age. No upgrades or service contracts, mind you, and every feature and customization is a new line item. That's for a 1:1 replacement on a system installed in 2008. Incredible, given how fast the tech world moves.
Replacement parts cost a shocking $4000 per pair of lanes. But wait, the bowling machines themselves are 70 years old, so what's this \"advanced\" system actually doing back there? Actuating a single relay to trigger that big old machine. Everything else is strictly mechanical. Meanwhile I've got a six-figure invoice in my hand. I'm upset.
Given the state of open hardware, computer vision, real-time event streaming, and open source running megascale products worldwide, there had to be a way to do this myself.
So far I've built an equivalent prototype for about $200 per lane-pair, $400 if you're fancy. ESP32 and ESPNow with an RS485 fallback, reporting to a raspberry pi lane computer that's really just redis and a state machine bolted to an ESP32 gateway for the mesh.
Since it's all ESP32, I've got a fistful of spare controllers in a drawer, pre-flashed or waiting to be. All common off-the-shelf hardware: microcontrollers wired to relays, optocouplers, and IR-break-beam sensors, each running slightly different firmware. Writing the firmware and protocol is the actual hard part.
It's an ESPNow star-topology mesh: each node emits events from its sensors and accepts commands for its controls, reporting to a gateway node connected to the raspi over UART. From there it's event streaming: RX packets get translated and tossed into redis, commands relay back out to the mesh as needed. RS485 sits underneath as a wired fallback for noisy RF environments.
Once the data's in redis, it's familiar middleware/React/websocket/pub-sub stuff. Any React dev can build their own UI and bowling animations. Since it all runs on commodity hardware, I can do legit anything I want as the proprietor, and I own all my data. Repairs take five minutes; I can swap the rig on a lane pair in under 10. I'd bet a house like mine could go from zero to running in an hour or two.
We're calling it OpenLaneLink, and I plan to open source the hardware, firmware, and software stack when it's ready. Bowling is fun, and I want to help keep it affordable for alleys like mine.\nI hate vendor lock-in. I'm not a fan of closed systems, calling support for every hiccup, or paying to \"white label\" my own equipment. Want to go Tron-themed for a night? Good luck finding a neon neumorphic theme in something bought at the turn of the century.
All that bugged me. Sure, bowling equipment is niche, but the open hardware and software landscape is amazing.
Thanks for reading! Let me know if anyone's interested in more posts about this bowling nonsense.\n\n\n
1298 points | 143 comments
\n\n\n\nLol, nice! It depends on what you mean by \"owning a bowling center,\" but it looks like there are at least two of us into this :D\nI bought a really old fully mechanical (and automated!) mini bowling lane. It works without a CPU, except for the score display, which originally used a 1970 Intel D8749H (MCS-48)!
Just like yours, the only thing that turns this machine on is a single relay :D\nOn my model, I also get feedback on the pin status and a \"ball received\" signal, so I can trigger the scoring.
I dumped the firmware and replaced the display PCB with a more recent DIY PCB with an Arduino that speaks a modern, documented protocol. It's now compatible with ScoreMore software, and I can do a lot more with it.
Here are some pictures and more info on the Bowltech forum (you should post your stuff here too!):
* https://www.bowltech.com/forum/small-scale-bowling/small-sca...
@vik8 on discord if you need :)\n— vikbez\n
Claude Code uses Bun written in Rust now
\n\n\n\n201 points | 296 comments
\n\n\n\nMaybe I’m taking crazy pills, but I’m still stuck on “why the hell does a TUI need to run in terminal React by way of JavaScript”\nThe fact that Anthropic felt the need to buy a runtime so they could make their TUI better speaks more to the quality of engineering than anything else IMO.
If rewrites are so easy, why not rewrite CC in a native language? Would’ve been a hell of a lot cheaper.\n— weakfish\n
What I learned selling 2,500 MIDI recorders: Hardware is not so hard
\n\n\n\n254 points | 116 comments
\n\n\n\nVery happy JamCorder customer here. I've started playing in improv shows since I bought it too, and it's nice to have those on record!\nIt's rare that I come across a product I have literally zero complaints about, or any feedback to offer. Even my concern that it was tied to an app that might disappear one day is of course covered by the fact that it's all just MIDI files on a card.
Thank you so much for making this, it's basically a perfect product for me and I love it. No notes.\n— DavidPiper\n
Claude Fable produced a counterexample to the Jacobian Conjecture
\n\n\n\n278 points | 156 comments
\n\n\n\nThis is a rare instance where feeding this groundbreaking information into an LLM gives _them_ psychosis. I fed this to claude code and watched it verify the result in 7 different ways to be 100% certain, and it was just flabbergasted. Quite remarkable.\n— aizk\n\n
AI advice made people 3x less accurate but 2x confident, researchers found
\n\n\n\n212 points | 100 comments
\n\n\n\nAdvice and information subreddits have gone to shit because of AI usage. A large number of people seem to think that when someone asks a question, what they really want is not someone with direct knowledge, but instead someone to relay the question to ChatGPT and post the result as if it is their own hard earned knowledge and insight. I have no idea about the quality of this research but in the real world (well, real-ish, as real as Reddit can be) it is stark, people aren’t just refusing to say “I don’t know” they’re actively seeking out opportunities to pretend they know things.\n— reticulates\n\n
Minecraft: Java Edition now uses SDL3
\n\n\n\n258 points | 168 comments
\n\n\n\nThe lwjgl bindings for this were written by a member of the GTNH modpack team, thereby completing the chain of vanilla->modded->vanilla once more https://github.com/LWJGL/lwjgl3/pull/1033\n— malteeez\n\n
Moonshot AI suspends new subscriptions due to Kimi K3 demand
\n\n\n\n197 points | 67 comments
\n\n\n\n> Over the past 48 hours, demand has pushed close to the limits of our current capacity. To protect the experience of existing subscribers, we're temporarily pausing new subscriptions and prioritizing compute for current members. Existing subscribed users are not affected.\nSuch a beautiful paragraph to read, a company that prioritizes their current customers and focus on keeping them satisfied instead of just focusing on fast growth.\n— Alifatisk\n
216 points | 150 comments
\n\n\n\nThis is amazing! I can see in a few years having Home Assistant use an LLM to coordinate my home while I'm away. Tell the LLM to make an inventory of my home and then send the vacuum and the robot around to clean everything up. A multimodal LLM can inspect pictures for items that need to go in the dishwasher, the laundry or needs polishing and reorganise my closets to best allocate space. It has all day to do it while the solar panels provide ample energy anyway.\n— chronogram\n\n
The Last MPEG-4 Visual Patent Has Expired
\n\n\n\n198 points | 58 comments
\n\n\n\nMoving in the right direction, but unfortunately it's still quite a few years before H264 is off patent, globally:\nhttps://meta.wikimedia.org/wiki/Have_the_patents_for_H.264_M...
At which point, the continued march towards higher resolution and bandwidth video will probably limit their utility\n— at1as\n
I joined the IndieWeb, here's what I learned
\n\n\n\n189 points | 92 comments
\n\n\n\nIf IndieWeb wants content, burying it under a ton of tech soup is the exact opposite of what they claim they want.\nHow do the developers not realise that this set of protocols is completely unworkable for 90% of users?
Please stop developing systems where the engineering is front and centre. If you want to put content UX first, you need to do exactly that, with a one-click solution that gets users up and running instantly.
If you users have to go anywhere near a command line, or Docker, or they have to hand-edit HTML or CSS, you're creating walls most users aren't going to scale.
At the moment this looks more like NerdNet than IndieWeb.\n— TheOtherHobbes\n
LG monitors silently install software through Windows Update without consent
\n\n\n\n679 points | 342 comments
\n\n\n\nThis is so much worse that the title makes it out to be:\n\nThis situation has.. no precedent as far as I can tell..1. Your OS installs malware (technically manufacturers software) from a 3rd party vendor in background, zero user interaction\n 2. Happens as soon as you or anyone with physical access plug in a device into the HDMI port\n 3. That malware has internet and full system access, no sandboxing\n 4. It starts with every system boot\n 5. This software gets installed when you plug in a new LG monitor\n 6. OR ALREADY HAD AN OLDER LG MONITOR PLUGGED IN, BECAUSE LG APPARENTLY ROLLED THIS OUT FOR MANY OLDER MODELS TOO!!\n 7. And yes, if you think that's horrendous, as mentioned in the video below, that also applies to 'Professional' LG monitors!\n\nGamersNexus has a video diving deeper into what LG did here - https://www.youtube.com/watch?v=Q9uefFYe6bM\n— devttyeu\n
GPT-5.6 used a prompt to close a 30-year gap in convex optimization
\n\n\n\n274 points | 150 comments
\n\n\n\nI know a bit about this field. This conjecture reads as somewhat more niche than the cyclic double cover conjecture recently proved by OpenAI, but nevertheless represents a real contribution.\nYou want to know how long it takes to solve an optimization problem, in this case over convex, lipschitz functions. (The restriction to a spherical domain is not really a restriction, you can just change variables for any bounded domain.) Anyway, showing upper bounds on time complexity is \"easy\" because it's just the runtime of your algorithm. Showing (nontrivial) lower bounds is usually much harder because it requires constraining all algorithms.
This proof apparently shows that the lower bound time complexity is equal to the time complexity of an existing 30-year old algorithm: it requires Omega(d^2) function evaluations to solve over this class of functions.
My gut says likely implies that d is the minimal number of evaluations if you have a gradient oracle because you can approximate a gradient with d function evaluations, but I'm not sure how hard it is to make that rigorous.\n— _alternator_\n
372 points | 62 comments
\n\n\n\nHandy is an amazing cross-platform app for dictation from the author. There are other awesome open-source dictation tools as well like native macOS ones. You do not need SaaS subscription in this day and age for transcription.\nI maintain this list of all the best open-source ones in this awesome-style GitHub repo. People looking for open-source dictation tools, hope you find something that works for you here:
https://github.com/primaprashant/awesome-voice-typing\n— primaprashant\n
Mayor Mamdani Says Landlords Can't Use AI Images to Advertise
\n\n\n\n412 points | 186 comments
\n\n\n\nThis is awesome! StreetEasy is how many New Yorkers find apartments. In the past few years, it has been flooded with AI-staged apartments. The AI stagings warp the room to fit furniture that would 100% certainly not fit there. It’s deceptive, and I’m glad it at least requires disclosure now (although I wish it were fully banned)\n— plants\n\n
Why do AI company logos look like buttholes? (2025)
\n\n\n\n375 points | 130 comments
\n\n\n\nThis trend is far older than the AI era. The British magazine Private eye for many years in the 1990s had a running joke in which each issue displayed the new logos of old companies that had spent a fortune on rebranding. All of them were variations on circles.\nNaomi Klein in her book No Logo interprets it as a form of abstraction away from the passé and less profitable business of whatever concrete task those industries used to do, and towards outsourcing, branding, and financialization as an end in itself.\n— adonovan\n
If You Build It, They Will Come
\n\n\n\n230 points | 83 comments
\n\n\n\n// Lots of people have a sort of consumer attitude towards their communities, where they take everything for granted. I saw things this way when I was young. A social scene is an automatic feature of the world that appears on its own, like a wild blueberry bush. It starts sprouting parties and dinners and conferences and reading groups as naturally as the bush sprouts berries\nTrue in general. As a kid you think of things as bigger than you. Like whoever maintains a hiking trail or runs your towns diner is \"big\" compared to you.
As a grown up you hopefully realize that it's the other way - the work and effort to make and maintain those things is vulnerable and fragile.
I think about this whenever I see someone hop over the subway turnstile. The transit system is \"for granted\" - it's you and your few bucks that matters. But of enough ppl feel that way it all goes away via decay eventually.\n— xyzelement\n
What AI did to stackoverflow in a graph
\n\n\n\n232 points | 271 comments
\n\n\n\nAny social organization needs to carefully consider their inclusion-exclusion curve with intentionality.\nI think a lot of people might balk at the word \"inclusivity\" today, but StackExchange had ridiculously high barriers to participation, making it inclusive to the long-time users on the site, but exclusive to the newbie participants who found themselves blocked for asking questions. They slowly killed the site in this manner.
The community might have survived this folly, even with AI, because it was still the best place for people with qualms about AI to ask questions... Except until StackOverflow management alienated those users, too, by shoving AI down their throats in every facet of the site.
Even I had internalized the vagaries and neuroses of the SO community but I had heavy reticence to ask questions, knowing I'd have to consider all the ways a bully eager to use their powers might misunderstand me. I can't imagine asking a question there without having had lurked for longer than a typical Bachelor's + Masters program.
Peak at 207K, minimum at 588. That might be an incomplete date point, so using the next most recent value 1226, StackOverflow has lost 99.41% of its activity.\n— lynndotpy\n
257 points | 283 comments
\n\n\n\nRegardless of whether they achieved parity via distillation, or whether they got here via independently constructing a model from scratch, it was always going to end this way for the frontier American labs. Distillation “attacks” are not attacks. The frontier labs “distilled” all existing human written knowledge into their models, there was always going to be a second class lab that would distill that model into a cheaper version of it. There was never any plausible explanation for why this wouldn’t happen. There was never any practical mechanism to prevent someone from saving a conversation and using it to train their own model.\nEven if it didn’t happen here, it was still the case that it was going to happen going forward. It was always going to end like this. Invest in the hardware companies, not the model companies.\n— nickysielicki\n
EU ban on destruction of unsold clothes and shoes enters into application
\n\n\n\n267 points | 263 comments
\n\n\n\n> To prevent misuse, businesses relying on these exemptions must provide proof (e.g. documents or test results) and publish annual reports on what they have discarded.\nI wonder if anybody is keeping track of everything a mid size business needs to take care of. Each particular report probably sounds like a reasonable request, but by now they're probably well into hundreds, and they're all outside the actual scope of the business (e.g. it may seem manageable for the bureaucrats designing them, because that's what they deal with all day, but not for a small organization doing... something else).\n— radu_floricica\n
Fable 5 vs. GPT-5.6 Sol on an NP-Hard Problem: Does /goal help?
\n\n\n\n205 points | 102 comments
\n\n\n\nAnthropic is losing bigtime against Open AI in coding space. I was using Claude code upto last march. Not an enterprise customer, but a responsible AI user where i don't over spend and use basic plan to manage repository with 400Kloc all together. We sell to local government and a team of 3. Claude code was super slow, never able to fix issues properly.(Despite with proper test cases, observability, documentation and layered architecture). After moving to codex, life has been much easy and free form usage anxiety. Now managing entire things with 2 codex plus account per team member. Its high time anthropic should stop scare mongering and build efficient models. Everyone doesn't need Fable. People need models that solve problems efficiently.\n— sreekanth850\n\n
My cofounders and I did YC all the way back in the Summer of 2010, with the initial idea of building \"OkCupid for jobs.\" That idea quickly fizzled, and we spent the better part of a year pivoting between other ideas that also failed.
Finally, we made something that we wanted ourselves: a self-directed programming retreat, where people built fun projects, contributed to open source, and helped each other become better programmers.
After running two small batches, we launched on HN[1] and got an incredible reception.
That post on HN helped us reach beyond our personal networks and meet programmers from around the world, many of whom have since become friends. HN brought us the majority of people who came to our next few batches, and in the years since, HN has remained our #2 source of applicants (after word of mouth).
Alas, pg's comment[2] on HN when we launched turned out to be prescient: Running free programming retreats isn't a billion-dollar business, but it's still a worthwhile thing to do, and has positively impacted over 3,000 people so far. And 15 years on I still wake up every day excited to keep working on it.
So, thanks HN, for helping make the Recurse Center possible, and for helping me find my life's work.
[1] https://news.ycombinator.com/item?id=3435183
[2] \"This sounds like a crazy plan for a startup, I realize, but this is the right sort of crazy. In fact, the way the Hackruiters think about Hacker School is a lot like the way we initially thought about YC: if it doesn't make money, it will at least have been a benevolent thing to do.\"\n\n\n
229 points | 19 comments
\n\n\n\nHadn't heard of this before but its a cool idea. I often take my own little \"programming retreats\" and go to another city for a week and work out of a nice coworking space for a sense of focus and engagement with my work.\n— tracerbulletx\n\n
Kaiser nurses say AI, workplace surveillance are making their jobs, care worse
\n\n\n\n471 points | 301 comments
\n\n\n\nI RFTA and the majority of the complaints are about call center metrics and the pressure to ration care. These are real concerns about misuse of metrics, but not AI. The AI empathy thing was a 2024 pilot that was discontinued.\nFWIW my wife works for Kaiser and finds a lot of value in the the medical LLM tools available to her. She tells me being able to do live translation, summarize notes, and quickly get comprehensive answers save her time and help her give better care. Her older patients also frequently come in bringing AI-powered alerts from their apple watches that detected cardiac events.
It's annoying that we use broad terms to describe a set of technologies that in some ways can be problematic and in another ways are very beneficial. We gotta evaluate each of these as they come rather than talk about blanket bans.\n— derekdahmer\n
First atmosphere found on Earth-like planet in habitable zone of distant star
\n\n\n\n345 points | 220 comments
\n\n\n\nDidn't know a rocky planet in the habitable zone of a red dwarf could retain atmosphere against intense stellar stripping.\nRed dwarfs are known to be cooler (the habitable zone is therefore closer) and unstable.
I don't think LHS 1140b is \"Earth-like\" at all. Rather, it's more like a mini-Neptune, being boiled off by its star.
Edit: JWST emission spectroscopy of LHS 1140b as it passes behind its star rules out a mini-Neptune. https://arxiv.org/abs/2406.15136\n— tulio_ribeiro\n
Evidence of inconsistencies in evaluation process and selection of winners
\n\n\n\n403 points | 247 comments
\n\n\n\nAI is useful. But the amount of people that are simply offloading all of their thinking to AI and blindly accepting the answer is absurd. Kaggle is most likely using ai to assess the submissions and are not using any common sense by blindly accepting the results.\n— ecshafer\n\n
353 points | 260 comments
\n\n\n\nSpeculation: open models is what will kill Anthropic and OpenAI. Hyperscalers can run the models without a licensing fee. Apple can make them smaller and put them on the device.\nThe frontier models are an edge and a liability. They're astronomically expensive to train. Without them, their models will fade into obscurity. Their marketing depends on people believing the models are meaningfully different, as people have sweatily argued on this forum. Personally, I'm not convinced there's much of a difference between these models at this point. The harness is what takes these random and hallucinogenic models and make them into something deterministic and useful.\n— babblingfish\n
Apple targets dozens of OpenAI employees with legal letters
\n\n\n\n237 points | 185 comments
\n\n\n\nEverybody wants a platform but nobody wants to spend what it takes to make a platform. That includes things like Windows Phone, Fire Phone, all the glasses, Humane, etc.\nAs much as everybody hates on OpenAI for chaotic management, they did buy Jony Ive and are presumably giving him everything he wants to build a platform for them. Even though it probably only buys them a 20% chance of success, they haven't doomed the project by underestimating what it takes budget-wise.
And they blew it. Maybe they blew it by not realizing that even long time Apple employees could get arrogant about security. Or maybe it was a loose ethical environment in general. Whatever is it the root or the problem, they set billions of dollars on fire maybe tens of billions, by being unnecessarily cute about Apple proprietary information when they could've been above reproach. They had the resources to hire all the right people with the right knowledge and probably already had them on board.\n— Zigurd\n
Kimi K3, and what we can still learn from the pelican benchmark
\n\n\n\n248 points | 141 comments
\n\n\n\nIt's incredible Simon still believes pelicans on bikes aren't part of the training set, despite hundreds of them on blogs, forums, and Github. Stuff we put in our company blog shows up known by LLMs 6 months later, and we have 1000x less traffic than Simon's own website\n— OsrsNeedsf2P\n\n
339 points | 30 comments
\n\n\n\nI did something very similar with progressive (adam7 interlaced) PNG: https://www.da.vidbuchanan.co.uk/adamation/image.png\n> so playback is entirely dependent on network delay
Ultimately true, but I set up my server to send each \"frame\" separately, with a fixed delay between each. Each frame is small so unless your network is unusually slow, the timing is set by my server.\n— Retr0id\n
Learning a few things about running SQLite
\n\n\n\n216 points | 57 comments
\n\n\n\nDiving a bit more into databases than your current comfort level/current job demands remains a great way to level up.\nI've worked with many web developers who get mental blockage around DB tooling (granted, I have similar mental blockage when it comes to some operations stuff like K8s), and you can go far in life without really having to ask _that many questions_.
But going in and finding out how your SQL turns into data gotten from disk/written to disk is very helpful in just \"knowing\" what might be a decent idea. That and understanding your DB's locking system (or lack thereof...).
Figuring that stuff out can help reduce the surprise level when you can't seem to get a \"simple COUNT\" working quickly in Postgres or the like...\n— rtpg\n
Three ways people respond to a problem (other than solving it)
\n\n\n\n225 points | 126 comments
\n\n\n\nPeople often attribute the government's inability to solve a problem even after throwing billions of dollars at it; as a sign of incompetence. While there is plenty of incompetence within government; I think the 'Preserve the Problem' response is mostly to blame.\nIf we 'solved' crime, homelessness, drug use, poverty, etc.; then budgets would decrease and political power would diminish. Those in charge of solving the problem often have the least incentive to do so.\n— didgetmaster\n
347 points | 186 comments
\n\n\n\nMore details:\n- https://platform.kimi.ai/docs/guide/kimi-k3-quickstart
- https://platform.kimi.ai/docs/pricing/chat-k3
1M context, pricing is $3/$15 for 1M tokens (cache $0.3), which is extremely high for a Chinese open-weight model, but if it's truly competitive with most of the current frontier and is only behind Fable/Sol, the pricing is justified.
This is 1:1 pricing of Anthropic's Sonnet series (except Sonnet 5 which is currently on discount), and very close to 5.6 Terra pricing (Terra's input is $2.5).
One thing to consider, though: reasoning efficiency matters directly for how expensive a model actually is in real use. GPT's models are extremely reasoning efficient, and some Claude models like Fable at lower effort are as well. So if Sol spends 10K reasoning tokens to do something (at $30/1M) vs Kimi K3 that spends 50K reasoning tokens, Sol would win on cost effectiveness.\n— Tiberium\n
Microsoft Comic Chat is now open source
\n\n\n\n486 points | 109 comments
\n\n\n\nHi, I'm Robert Standefer, the guy who made this happen, with lots of support. I'm excited to see the enthusiasm about Comic Chat being open sourced. How this came to happen is a very interesting story that spans a six-year period with success that hinged upon being in the right place at the right time, literally.\nI want to point out that, while I (along with Scott Hanselman) made the Comic Chat open source release happen, I am not the original developer. That is DJ Kurlander, and he was very supportive of this project. He was even enthusiastic about it.\n— outintospace\n
Sony Deletes a Bunch More Movies from the Accounts of People Who 'Bought' Them
\n\n\n\n338 points | 183 comments
\n\n\n\nRecent and very related:\nPhysical disc production ending in Jan 2028 for new games on PlayStation (797 comments) https://news.ycombinator.com/item?id=48745456
So Sony is simultaneously announcing that all purchases will be digital from now on while actively demonstrating that digital purchases aren't actually purchases. They're clearly communicating that they believe in a future where no one owns games any more.\n— demosthanos\n
OnePlus halts operations in USA and Europe
\n\n\n\n391 points | 209 comments
\n\n\n\nOnePlus is one of the saddest stories out there. It was the hacker's choice for a while. It was originally the \"Never Settle\" phone that ran mostly stock android, had specs maxxed out, price was great, and bootloader was unlocked plus they provided factory images. Those were all reasons I bought a lot of OnePlus phones in the early years.\nThen they flushed nearly all of it down the toilet. The day they stopped posting factory images was the day I saw the writing on the wall. Such a shame.\n— freedomben\n
358 points | 89 comments
\n\n\n\nIs it useful? No. Does it stop AI from reading it? Also no. But is it cool? Yes, it is very cool.\n— OsrsNeedsf2P\n\n
How Our Rust-to-Zig Rewrite Is Going
\n\n\n\n386 points | 212 comments
\n\n\n\nI think this is a fine post. But one comment:\n> remember that for compilers which emit machine code, like roc and rustc, doing memory-unsafe things is a big part of the job
I don't really think that this is true, in the way that it's written.
I think that for the hot binary patching / code reloading features, yes, that is going to need unsafe. But for regular old \"producing an executable\" compilation? Emitting machine code isn't the part that requires unsafe. The language's runtime is a more likely site to find unsafe.\n— steveklabnik\n
NotebookLM is now Gemini Notebook
\n\n\n\n288 points | 146 comments
\n\n\n\nWhen notebookLM was new, it was interesting to listen to the podcasts. Then the novelty wore off, and I wanted something where I can interact with the podcasters but it was janky as hell.\nMy current “audio-learning” hack is ChatGPT Live which has become shockingly good after being awful compared to Claude Voice \n(Let’s not even talk about Gemini voice which is still bad).
I go on a walk and dump a paper or article link in the chat, and ask chatGPT Live to walk me through the content in small nuggets, so I can discuss them interactively. For deeper topics I have it quiz me Socratic style so I’m not just passively listening, and actually thinking through problems or ideas.\n— d4rkp4ttern\n
SpaceX stock erases all its gains and slides below IPO price in intraday trading
\n\n\n\n264 points | 228 comments
\n\n\n\n“Shocking nobody” I would add… This is an obvious Musk scam and anyone with financial knowledge called it as such ever since it’s been on the table. Why else lobby and change the NASDAQ listing rules for instance?\n— ornornor\n\n
$100 AI Music Video: Claude Fable 5 vs. GPT-5.6 Sol
\n\n\n\n224 points | 265 comments
\n\n\n\nAI is destroying the economics which allowed for a sizable middle class of artists. The issue is that many are paid for their art mostly for its aesthetic rather than artistic value. This isn’t the most creatively fulfilling, but it previously allowed many artists to make a living while refining their skills, often enabling them to pursue their real creative ambitions on the side.\n— qbit42\n\n
228 points | 88 comments
\n\n\n\nKinda unrelated, but the art direction on Ente's site is really top-notch.\n— speak_plainly\n\n
Inkling: Our Open-Weights Model
\n\n\n\n580 points | 143 comments
\n\n\n\nVery nice, multi modal, largest open weight model that supports audio. Would be interesting to see how good the audio capability is.\nIf you want to run locally, checkout\nhttps://github.com/danielhanchen/llama.cpp/tree/add-inkling\nhttps://unsloth.ai/docs/models/inkling\nhttps://huggingface.co/unsloth/inkling-GGUF\nhttps://huggingface.co/unsloth/inkling-NVFP4
This supposedly is better than KimiK2.7, as much hype as GLM5.2 gets, I find myself using KimiK2.7 half of the time, so if the benchmark is true, then this can definitely go in the mix. My hope is that it might have strengths in some areas to beat all other open weight models.\n— segmondy\n
Sleep regularity is a stronger predictor of mortality risk than sleep duration (2023)
\n\n\n\n517 points | 256 comments
\n\n\n\nMagnesium supplementation solved my sleep issues.\nI have seen many doctors, including sleep specialists, regarding insomnia. They all pointed to one source as the reason for the sleep issues: stress. And they all wanted to put me on prescription sleeping pills. I said no to that. Sleeping pills can cause dependence, and they often treat the symptom rather than the underlying cause. As a software developer, I am used to finding and fixing the underlying problem instead of relying on the quick fixes these doctors were offering me.
After much research, I figured out what I believe was the underlying problem, and the fix for it. The underlying problem was magnesium deficiency. As a software developer, I spend much of the day doing mentally demanding work. This is the kind of stress the doctors were talking about. Stress can increase the body's demand for magnesium and may contribute to low magnesium levels.
The cells in our body depend on minerals such as calcium and magnesium for normal function. In muscle and nerve cells, calcium helps switch the cell into an active state, while magnesium helps keep that activation under control and supports the return to a resting state. When you are low on magnesium, your muscles may remain tense and your nervous system may have a harder time settling down. That can contribute to muscle stiffness and difficulty sleeping.
The solution, in my case, was magnesium supplements. They fixed my muscle stiffness issues and my sleep issues. A special form of magnesium called magnesium L-threonate may be especially helpful for the brain because it appears to raise brain magnesium levels more effectively than some other forms.\n— petilon\n
SpaceX bond worth 10% less than issue price – heading for junk bond status
\n\n\n\n412 points | 350 comments
\n\n\n\nThe worst about the SpaceX IPO is Nasdaq changing their inclusion rules for the Nasdaq 100. The index fast-tracked SpaceX stock for inclusion 15 days after the IPO, instead of the normal three-month seasoning period. They also changed its 10% minimum float rule to a 3x weighting boost for low-float stockss.\nSo many people will unwillingly and prematurely invest into SpaceX, before it has any chance to discover its real price.\nIE: The floating, 5% at launch, could attain 30% end august, if Nasdaq didn't change their rules it would have included SpaceX after this..\nhttps://finance.yahoo.com/markets/stocks/articles/nasdaq-che...\n— lucd\n
385 points | 411 comments
\n\n\n\nThere's some surprising stuff in this codebase. For example, https://github.com/xai-org/grok-build/blob/b189869b7755d2b48... is a \"self-contained terminal renderer for Mermaid diagrams\", which renders a subset of Mermaid chart types using Unicode box-drawing.\n— simonw\n\n
285 points | 152 comments
\n\n\n\nNothing but respect to TurnTrout for taking an action like this. The world needs more smart people who are willing to stand for what they feel is right, despite the pressures otherwise. Without that occurring more, our species is going to lose many impactful prisoner's dilemmas coming these next two decades.\nThis raises my respect for AI researchers a little bit too. I have often felt that the entire industry is pretty tainted to the core, and for better or worse that colors my opinion of the researchers.
Maybe I'm in the minority, but I thought it was gross to download pirated art for a student project when I was at Berkeley years ago. So it has been really sad to witness many of the most brilliant minds of this generation answering the siren song of disrespecting the collective effort of others to extract and resell residual value.
I'd guess TurnTrout doesn't agree on that framing, otherwise he probably would not have been at Deep Mind. But clearly he and I agree on other ethical positions; I am nothing but glad to see him stick to his principles here.\n— recitedropper\n
324 points | 174 comments
\n\n\n\nThe thing I miss most, and the thing we’ll never get back was the cultural buy-in and network effects.\nMy music discovery then was different friend groups incrementally amassing large collections of albums in whatever sub-culture that friend groups had doubled down on. My iPod would be the culmination of my friendships. I would then fall in love with bands and albums and tracks on these albums without any influence before hand on their popularity or their algorithmic match to my music tastes.
The result was pure joy: my music taste would develop in all weird and wonderful directions, my favorite songs would be the one I hit back on to listen again while I moved through an album, songs that friends skipped over and didn’t know at all; bands that never charted anywhere but made interesting music… bands that never knew their music made it to an iPod in South Africa.
(I’ve got a song still stuck in my head from a Canadian indie band that made its way onto my iPod via via and I’ve done all the searching in the world for the lyrics I remember and have never found the band. I love this that I’ve never found them!)
I make an effort to use Spotify to find and listen to albums, but it wasn’t built for this, and invariably find 90% of my listening happening on algo-generated playlists of songs that sound exactly like a song I like. I never learn the names of the songs or the names of the bands as the songs go by, and I fall in love with none of it… It just vaguely sounds like stuff I like. It sucks.
I don’t listen to any AI generated music consciously, but given the music experience today I probably wouldn’t notice as these playlists, like a boiling frog, slowly became AI music dominated.
I bought a record player as my protest, and it gives me immense joy to find obscure records and play them through; but it’s really not the same thing, and I miss what we had.\n— devonsolomon\n
Prioritize mental health, and why communication is so important
\n\n\n\n278 points | 238 comments
\n\n\n\n\n\nIf you are neurodivergent or have other things influencing you mentally, you are _NOT_ going to snap out of it. You are not going to just build a better planning system one day.> My goals for the end of 2027 are as follows:\n\n > Stop making stupid mistakes. I want to be able to finish a task fully without missing or skipping a step. One way to do this is to make a plan for everything you do, and only do that thing. Nothing else.\nHN comments are _NOT_ going to help debugging your mental state. People here have trouble agreeing on engineering, product and business practices they specialise in. They are _NOT_ going to guide you in the right direction on mental health topics.
Please OP, close HN, reach out to people, get help, and (importantly) learn to navigate your mind, not fight with it.\n— smugglerFlynn\n
Running Gemma 4 26B at 5 tokens/sec on a 13-year-old Xeon with no GPU
\n\n\n\n218 points | 140 comments
\n\n\n\nI have a prediction. By the mid of 2027, we will have >200B MoE models running on basic consumer hardware.\nI am running Qwen3.6-35B-A3B locally on my 16GB mac with 7-9 tokens/second. Link - https://github.com/deepanwadhwa/samosa-chat
This is a GPT4 level model running locally with a decent speed on a 16gb ram macbook air.\n— dwa3592\n
257 points | 220 comments
\n\n\n\nI feel this as more of a fashion runway garment type object, more statement and vision than what you'd see in everyday retail.\nReminds me of an italian specialty coffee shop that puts moka pots as napkin holders on every table.
I just hope they don't just go after the keyboard but the screen, desk and office.\n— 6thbit\n
SQLite should have (Rust-style) editions
\n\n\n\n216 points | 79 comments
\n\n\n\nInteresting idea - I like seeing a list of pet-peeves followed by a proposal for a straightforward way to have a set of 'alternative defaults' that remains backwards compatible. If you don't want to opt in, don't run the new PRAGMA edition = 2026.\nToo often it's just a list of issues and a wish that everyone else will change.
In (mild) defense of SQLITE_BUSY - busy_timeout just tells sqlite to sleep and retry up to the timeout when it receives SQLITE_BUSY. It seems like a sensible default for a library to leave that up the calling code - which may have something else it could do while it waits. However, that logic often gets missed!\n— sethev\n
Bonsai 27B: A 27B-Class model that runs on a phone
\n\n\n\n369 points | 140 comments
\n\n\n\nWhat I most want to see it compared to is Gemma 4 12B in the 4-bit QAT version. It's barely bigger than this at just under 7GB, so it also runs on just about any modern device and is remarkably smart for its size. It's an excellent tool user, crazy good vision for its size. I'm still trying to wrap my head around how much is lost with each step down in resolution, but the QAT versions from Google seem to prove the answer is \"very little\" at four bits.\n— SwellJoe\n\n
How to stop Claude from saying load-bearing
\n\n\n\n408 points | 468 comments
\n\n\n\nI do not mind when I am coding with Claude and it uses all the typical claudisms. I am much more bothered when I am reading a blog post, email, or other form of prose and I see those same claudisms.\nI guess they are not annoying since I know I am talking to an LLM and expect the typical responses. When I am reading prose online that I previously would have expected a human to write, it can be quite jarring to realize its an LLM.\n— doctoboggan\n
Are we offloading too much of our thinking to AI?
\n\n\n\n356 points | 357 comments
\n\n\n\nI don't know if this is a good framing. \"Too much\" is subjective, and every heavy AI user will assert that they're just unlocking their potential, that calculators didn't make us dumber, etc.\nBut to latch onto the calculator argument: if you outsource adding numbers to a calculator, you're still you. On the flip side, if you use an LLM do most of your thinking, what's left? We have people here who use LLMs to raise their children, to manage relationships, to design products. So what's your unique contribution to this world - is it the prompt you once wrote? You're standing in front of a token-generating machine, pulling a lever, sometimes receiving gifts. Is that your edge, your unique experience, your purpose in life?
Many LLM maximalists say they use the tech to learn new things, but to what effect? Are you going to apply that knowledge of physics or computer science yourself, or will you just prompt the LLM again?
In my mind, it's pretty simple: I'm a human, LLMs are not. If a human writes a novel, it's inherently worth more because it's hard-earned and anchored to experiences we share. I want to support that. And I want to be a human who can write novels, the old-fashioned way. I'm not good at lifting weights or running, so my thinking is the only thing I have.\n— zerobees\n
309 points | 152 comments
\n\n\n\nI've said for a long time that composability in software is a bit like playing Tetris: the lines have to clear.\nI feel like that gives an even more literal tower-rising metaphor, and that's what it feels like people using agents naively (and software engineers of lower skill or earlier-career), end up violating.
Agents are getting better at folding things into themselves, especially if you direct them to... but unfortunately I've found that the architectural instincts, even of Fable and 5.6 Sol, are still wildly behind what I reflexively achieve, say.
For sure there is an ability to have agents go back over work and try to fold it into better and better abstractions until it's sort of annealed into something good. I've done something similar on codebases that I have, but the 'high reaches' of architecture with great _prediction of how the software will evolve in the future_ in _subtle_ ways – those are, for now, out of reach of agents.
There is a part of me that wonders if it's partly just how much they can hold in their head right now, though. Even with the greatest articulation and high density of feeding them, the current setups don't allow them to hold a high-quality, sparse, 'zoomable' model of the world in their head that well yet, which we can do pretty well.
But the fact that I'm talking about it in terms of that kind of subtlety is itself promising, I guess?\n— tekacs\n
Jurassic Park computers in excruciating detail
\n\n\n\n276 points | 62 comments
\n\n\n\n> It is unclear how Jurassic Park crew got their hands on a Motorola Envoy\nThe head of frogdesign (Hartmut Esslinger) ended up running into Spielberg on a plane and showed it to him. The one in the movie is an original mockup.
Source: https://www.therpf.com/forums/threads/jurassic-park-tablet-d...
Discussion: https://news.ycombinator.com/item?id=46752261\n— kalleboo\n
Codex starts encrypting sub-agent prompts
\n\n\n\n381 points | 231 comments
\n\n\n\nI was wondering why my local tool to inspect coding agent sessions stopped working in some cases.\nThis is a really interesting engineering decision, I wonder how many people will want an encrypted external piece of instructions running on their machine.\n— jeanlucas\n
Measuring Input Latency on Linux: X11 vs. Wayland, VRR, and DXVK
\n\n\n\n337 points | 216 comments
\n\n\n\nOne thing that's lovely about Linux is this kind of analysis is not only possible, but meaningful. These results will get reported back to the graphics software authors and the distribution packagers and the ecosystem will improve. There's no sense with Microsoft that kind of improvement is possible.\nI recently switched to Linux after years on Windows desktop, mostly because the KDE Plasma desktop feels snappier than Windows 11. Also the feeling that if something isn't working right I can probably tinker and improve it. It's been really nice. If you haven't tried Linux desktops in awhile give Bazzite a whirl: it's a Fedora customized for gaming. Even if you don't game it's an easy way to get a very functional Linux desktop in no time at all.\n— NelsonMinar\n
Cursor 0day: When Full Disclosure Becomes the Only Protection Left
\n\n\n\n323 points | 157 comments
\n\n\n\nAs someone who is on the other side, the amount of familiar, LLM generated reports are overwhelming and usually falls under \"not familiar with product design/security scope\" category. But there are also really good ones - so I can't afford to not take actual look at each, but it gets tiring and we need a solution. Spamming the former category with LLM generated \"rationale\" isn't that solution (yes I can tell this article is mostly LLM generated).\nUnfortunately, this looks like the former case. If the software can execute arbitrary code/binary, and you place a malicious binary, that's up to you to secure/sandbox the workspace, not the software. Unless cursor commit themselves to securing the user environment, which I don't think they are in the business of.
If you are generating a CVE report with LLM, please use LLM responsibly in helping you reproduce deterministically. Then please do the write up yourself, keep it as concise as possible and strip most adjectives in any LLM generated sections as they cannot help themselves to write without mostly useless exaggerations.\n— Aperocky\n
S&P Global has lowered Oracle’s creditworthiness from BBB to BBB-
\n\n\n\n303 points | 308 comments
\n\n\n\nMarket signals on an impending AI bust are broader than just Oracle’s woes.\nFor example, Amazon just had a challenging bond offering where the market is clearly starting to seriously question the ROI on all this money being pumped into AI buildout. That does not bode well at all for AI-only companies without broader cash flow from other businesses. And when the cash dries up this whole thing comes crashing down like a house of cards.\n— cmiles8\n
I tricked Claude into leaking your deepest, darkest secrets
\n\n\n\n300 points | 135 comments
\n\n\n\nDoesn’t surprise me.\nYesterday I learned that people run AI agents on their system with full admin rights. No containerisation or anything. Wild. Like we forgot 50 years of computer security overnight.\n— artisinal\n
Apple's new SpeechAnalyzer API, benchmarked against Whisper and its predecessor
\n\n\n\n412 points | 172 comments
\n\n\n\nWhisper is the wrong model to benchmark against, or rather, there are better models that are state of the art now like Nemotron and Parakeet both by Nvidia, as well as Mistral's Voxtral and Cohere Transcribe.\nHowever, what's funny is, RIP to a lot of the paid apps that simply wrap Whisper, I'm sure Apple will make a native GUI such as a recorder app for macOS that obviates the need for these wrappers, which everyone seems to be vibe coding these days.\n— satvikpendem\n
Climate.gov was destroyed. Open data saved it
\n\n\n\n385 points | 154 comments
\n\n\n\nI'm glad someone managed to save the data that we all payed for.\nMy question is, how will this site stay relevant? The collection/analysis/monitoring of the current situation is as important as historic data. Turning current data into historical data takes significant resources.\n— imoverclocked\n
Grok uploaded my user directory to xAI's servers
\n\n\n\n433 points | 230 comments
\n\n\n\nSo many of the replies are saying that they should've restricted access using .md files and whatnot. Is really any guarantee that they even follow those? It seems like even if you ask pretty please don't touch those files, there's a chance they will. So many people have just willingly installed spyware on their computers and big tech calls this the next big thing.\n— LetsGetTechnicl\n\n
Japan develops a method to recover up to 90% of lithium from used EV batteries
\n\n\n\n302 points | 73 comments
\n\n\n\nThe article is very, very light with details. The university or research center is not named. No scientist is named. No link. Nothing that tells \"look, we're telling you real, solid, serious stuff.\"\nHere is another article with that details : https://www.techspot.com/news/112051-japan-finds-way-recover...\n— BaudouinVH\n
Building and shipping Mac and iOS apps without ever opening Xcode
\n\n\n\n255 points | 116 comments
\n\n\n\nOne down-side to this is that it does require you to run the agent on your Mac instead of in a Sandbox. I do this too and there are lots of problems I can't solve in a sandbox. I know a lot of you are throwing your hands up at the years of security practices we're throwing out the window when we do that.\nThe fact that xAI uploaded someone's home directory, including their SSH keys, is giving me serious pause at my choices here.
Generally, I don't worry about my machine being \"blown up\". I don't have a TON of unreproducible stuff on my machine. Everything is backed up, committed to git, and the like. I can restore most of it in a couple hours.
That said, I really, really don't want my .SSH directory sent to an AI agent and it's silly to prompt your way around that. You need to block it at the system level. I'm considering a separate user and then 700 permissions on my home directory.
I feel like we're back to 1990's security here. The double-edged sword is that it's helping us get things done at a pace like never before.
I'm not throwing shade here, I'm among the guilty.\n— codazoda\n
LAPD lets contract with surveillance giant Flock expire
\n\n\n\n424 points | 368 comments
\n\n\n\nThe best part is that flock owns the cameras and the poles so even when the contract expires the cameras keep running and recording data that flock can sell to e.g. CHP, LASD, FBI, Palantir; and LAPD can just call them and access the data\nthe flock scam was engineered to be resilient to political pressure by giving departments and jursidictions this fake exit ability while the data continues to be harvested, it is a noose that only tightens; the amount of flock cameras recording only ever goes up not down.\n— etdznots\n
Grok CLI uploaded the whole home directory to GCS
\n\n\n\n259 points | 98 comments
\n\n\n\nThough I'm in the camp \"people should really know to sandbox by now and be careful\", I'd say we should also be mindful of how far from everyone has deep knowledge of the systems and tools they use. This behaviour of a tool is just malicious. You have to take into account the human factor, of how people likely end up using a system. And in this case, the consequences of exfiltrating so many secrets this way are really quite unacceptable.\n— Greenpants\n\n
365 points | 96 comments
\n\n\n\nThis game was so good, it's as if the developer was given a team of PhD-level experts to work on it.\n— mrkeen\n\n
A voxel Tokyo in real Japan time – ride the Yamanote line and study Japanese
\n\n\n\n232 points | 34 comments
\n\n\n\nThe evening mode has vibes of Ghost in the Shell's \"chase UI\", glimpses of which is in this scene:\nhttps://www.youtube.com/watch?v=amKt8ttja1E\n— netsharc\n\n
Samsung Health app threatens data deletion if users opt out AI training
\n\n\n\n307 points | 83 comments
\n\n\n\nWow. Good thing I sent back my Samsung health watch after discovering some of the fancy features were locked to having a Samsung smartphone to go with it.\nShameful business conduct.\n— eqvinox\n
Open questions:
1. Why is the regular voting system not enough?
2. Should HN change in response to the gen AI era? It has been successful not changing fundamentals.\n\n\n
478 points | 249 comments
\n\n\n\nWe don't allow genai text on HN itself - see https://news.ycombinator.com/newsguidelines.html#generated and https://news.ycombinator.com/item?id=47340079. How to enforce it is a separate question, of course, but the rule exists.\nWe don't have a similar rule yet about article content but my sense is that the community mostly doesn't want to read it—or, to put it more conservatively, discounts it. This is why we see so many \"just show me the prompt\" responses, along with others like this: https://news.ycombinator.com/genai-pushback. I built that list so I have something to send to users who email about why their genai articles got flagged.
It's a fascinating arms race right now: the AIs are training on the humans but the human hivemind is also training on the AIs. Readers are developing allergic sensitivities to language that sounds like an LLM produced it. The AIs will adapt to this, but the humans will adapt in turn. Where it ends up is anyone's guess. I have an optimistic view, but I've been wrong about this so many times that I have low confidence in it.
There is an emerging class distinction between writing (and writers) that use genai vs. writing that does not. As soon as the \"this sounds like an LLM\" allergy kicks in, the writing instantly gets relegated to a low-status bucket in the reader's mind. That doesn't mean it won't still get looked at - but it is now under a stigma.
(I was rather pleased with the originality of this until I remembered pg had come up with \"writes and write-nots\" in https://paulgraham.com/writes.html. Oh well, it's the point that matters.)
This has the happy flipside that anyone who would like readers to classify their article as high-status rather than low-status can apply the judo move of simply writing it themselves.
Now I need to add the disclaimer that none of this is a dismissal of LLM technology per se. We rely on it heavily, and there's no question that it's useful. The question is how to use it (pg again: https://x.com/paulg/status/2058871512451412457) and whether one should use it on writing that one publishes to other humans.
To turn to OP's questions:
> Should HN add the ability to flag articles as AI-generated? [...] it could just show up as an indicator
Flagging-as-just-an-indicator would be tagging, which we've always resisted adding to HN, but I wouldn't rule it out.
What I do think we'll (finally) add is a \"please give a reason why you flagged this post\" step, and \"because I think it's genai\" will be one choice among several (spam, offtopic, mean, etc.)
> Why is the regular voting system not enough?
The regular voting system is never enough. https://hn.algolia.com/?dateRange=all&page=0&prefix=false&so...
> Should HN change in response to the gen AI era?
To this I am tempted to reply with https://news.ycombinator.com/item?id=48887149 in homage to https://news.ycombinator.com/item?id=3742902.\n— dang\n
Claude Code sends 33k tokens before reading the prompt; OpenCode sends 7k
\nThis started based off of a hunch. We usually use OpenCode, but were 'forced' to use Claude Code for a while due to issues with Meridian. In that time, we saw the usage meter rise much, much more quickly than when using OpenCode.This was the initial anecdotal evidence, but we undertook this small study to collect empirical data:
We added logging between the agentic coding tool (Claude Code and OpenCode) and Anthropic's endpoint, and captured all requests (and the returned usage blocks).
With one caveat (toward the end of the post) we found unambiguously that Claude Code was far more inefficient in terms of its cache strategy and its harness token usage than OpenCode.\n\n\n
422 points | 236 comments
\n\n\n\nWhat really burns tokens is sub agents. I once gave Claude Code a pretty big task, and it immediately launched 7 sub agents which burned through my budget before even one of them was finished. Tried again 5 hours later: same result.\nIf I let the main agent do the same task sequentially, it was no problem at all. I don't know if it's really just communication and orchestration that makes sub agents so inefficient, or if Anthropic figured that most people using sub agents pay per token on a big corporate account, so this is an easy way to make more money from tokenmaxxers.\n— mcv\n
Old and new apps, via modern coding agents by Terry Tao
\n\n\n\n265 points | 72 comments
\n\n\n\nBuilding visualizations with LLMs has been a major boost for my CS classes:\nhttps://htmx.org/essays/universities-and-ai/#demos-visualiza...
Many visualizations that I have always wanted but just didn't have the time to build, I now have.
To give an example, I wanted a simplified 8-bit computer to complement the 16-bit teaching computer I use and designed this in a few days with the help of claude:
https://bdp.cs.montana.edu/\n— recursivedoubts\n
289 points | 171 comments
\n\n\n\nThis line: \"this is my main argument against the valuation of frontier labs. It’s not that AI won’t create that much value, it’s that they won’t capture it.\"\nThat is a very astute and concise way to explain everything about how the frontier labs are behaving and how they're trying to push more people to pay token rates for the best models. At the current subscription prices ($100 or $200 a month for a generous, though bounded, amount of tokens), frontier models are a no-brainer, most folks and companies will use them. But, at token rates, 10x or 100x the cost of open models or what I was spending on the frontier models a month ago? That is a harder question to answer \"yes\" to. I certainly wouldn't spend $1000 a month for the best model, much less $10,000; my employer might pay $1000/month, but definitely not $10,000. The frontier labs need everyone to answer \"yes\" to spending 100x what they currently spend to justify the valuations, and it's just not going to happen as long as everyone knows how to make these models.
Both OpenAI and Anthropic are trying to figure that out now. Anthropic, in particular, has their finger on the trigger...they want to push people to usage-based billing for Fable. But, OpenAI released 5.6 Sol, competitive with Fable (or close enough), and it's available via subscription (even the $20 subscription!), and there's no moat keeping someone from switching. If Anthropic really does end Fable access on the subscription plans in a few days, I predict a large market move back toward OpenAI.
The market isn't going to bear the cost of making the frontiers investment make sense.\n— SwellJoe\n
Since Chromium 148, Math.tanh is now fingerprintable to link underlying OS
\n\n\n\n220 points | 109 comments
\n\n\n\n> One tanh call on the right input is a per-OS signature. Claim macOS, return Linux math bits, and you have contradicted your own User-Agent.\nThey (or rather the LLM that wrote this) missed that this is possibly fingerprintable to browser version range, which is slightly more interesting. Most users aren't spoofing their user agent headers to be a different operating system. Most fingerprinting solutions aren't trying to infer your operating system, they only care about semi-unique things that show up.
It's an interesting finding. I wish they had taken some time to have a real person write it up. This is too heavily LLM written to ignore.\n— Aurornis\n
244 points | 136 comments
\n\n\n\nI have gained so much by reading comments on Hacker News, articles, lurking in groups and forums, watching YouTube videos. Reading books is important because they have knowledge that is not accessible in another format. It is an efficient way to condense information and pass it around. At least it used to be, when paper was more scarce. Now books get fluffed up for padding. A lot of people should be writing articles instead of books. Those who are reading a large volume of books are most certainly reading a large volume of garbage. I have gained more from reading comments on here than I would have otherwise by reading junk self-help books, romance novels, and other kinds of YA. You should be reading the right books. The classics, the culturally relevant, the historically significant, the hallmarks of a domain or expertise, etc. The medium of publishing used to be a marker of quality, a filter, a selection mechanism. But as publishing became more accessible, the amount of investment required also plummeted. And just as it allowed many creatives and geniuses access to the medium, it also introduced a large volume of cheap material that can be easily dismissed as irrelevant noise, as you should.\n— coldblues\n\n
289 points | 204 comments
\n\n\n\n\"Harvey previously stood as a similar character, Lord Buckethead, but was forced to create a new character due to a dispute with the filmmaker Todd Durham, who owns the Buckethead character\" [1].\n(The videos on this website are worth the watch. Hilarious, of course. But also...Binface conjugates Latin to Sky News, and not just as a bit. I don't know how I feel about the British comedy candidate outclassing half of the American elected leadership–and a good fraction of its industrial leadership–on IQ.)
[1] https://en.wikipedia.org/wiki/Count_Binface\n— JumpCrisscross\n
214 points | 13 comments
\n\n\n\nI love the pin level emulation model of this. The self contained modular behaviour of components has real flexibility.\nFor a long time I have wondered if extremely thin but explicitly defined interfaces are an under-explored domain for interoperability.
Beyond simple chip emulation, any set of values that are sampled on each side of the interface at defined times, and a small set of signals to provided temporal access. Be it pins, a small set of registers, or a memory mapped region, it seems like an excellent target for conformance testing.
Perhaps in a world of AI generated code, modular components with explicit conformance requirements would allow people to not care what happens inside the black box, provided it cannot be made to violate it's behaviour requirements.\n— Lerc\n
LARP – Revenue infrastructure for serious founders
\n\n\n\n213 points | 46 comments
\n\n\n\nThis reminds me a lot of good old Bonneville Pacific Corps model for getting rich, which worked for years and did make them rich until they tried to merge with one of those naturally suspicious and rule following German firms\nhttps://www.utb.uscourts.gov/sites/utb/files/case_opinion/38...\n— bryanrasmussen\n
The shingles vaccine may reduce the risk of dementia
\nhttps://archive.ph/PzPop\n\n\n202 points | 175 comments
\n\n\n\nIf you haven't seen the chart from the UK study, I highly recommend checking it out: https://erictopol.substack.com/p/the-shingles-vaccine-and-re...\nWhen the vaccine came out in the UK, they included a hard age cutoff: Above a certain age, you weren't eligible. Below that age, you were eligible.
They looked at the probability of a dementia diagnosis over the 7 years after the vaccine was introduced.
People who were born in the \"can get the vaccine\" group have markedly lower rates of dementia. People in the \"too old\" group have higher rates. It's cut and dry. The researchers didn't separate out the people who actually got the vaccine.
It's one of those studies where you don't even need to look at the p-value to see the difference between the cohorts.\n— gdudeman\n
Female US rower completes historic solo journey from California to Hawaii
\n\n\n\n228 points | 80 comments
\n\n\n\nThis is no joke. I've done the crossing from Moloka‘i to Oahu (~45 miles) in a canoe several times, and those open ocean waves can get very nasty (largest I've dealt with were around 15m tall). I can't imagine the mental endurance required here, let alone the physical. My longest crossing took 9 hours, and I was completely drained by the time I touched shore. 44 days is absolutely insane.\nSuch a huge accomplishment.\n— jjcm\n
Prefer strict tables in SQLite
\n\n\n\n199 points | 89 comments
\n\n\n\nComing from the enterprise SQL world, I never took SQLite seriously for the very reason that field types were not enforced by default. (Yes, I was agog when it became the backbone for app metadata on smartphones.) Anyway, reading this reminds me of the old chestnut from networking about choosing UDP over TCP for its low-latency and simplicity and then eventually adding nearly all the reliability facilities of TCP to the app (automatic retry, etc) by hand.\n— ezekiel68\n\n
What xAI's Grok Build CLI Actually Sends to xAI
\n\n\n\n244 points | 116 comments
\n\n\n\n\"It uploads the whole repository — every tracked file's content plus git history — independent of what the agent reads\"\nHoly cow!!!! I mean I kinda expected Elon would do something like this to try to catch-up.. but this is extremely concerning.
This is precisely the reason, even though their pricing is competitive and grok-4.5 is actually good enough, I chose not to go with them.\n— freakynit\n
Show HN: Ant – A JavaScript runtime and ecosystem
\nHello HN!I'm the author of Ant, a JavaScript ecosystem built around a runtime with its own JavaScript engine. Ant also includes a package manager, the ants.land package registry, a platform for deploying and hosting applications, and Ant Desktop for building native desktop apps with web technologies, similar to Electron.
The goal is for these pieces to work as one coherent platform while remaining compatible with the wider JavaScript ecosystem. It's still early, and I'd appreciate any feedback on the overall direction or what you'd like to see from an e2e alternative to the existing JavaScript stacks.
P.S. I’ve shared Ant here before as a runtime; since then, it has grown into the broader ecosystem you see today.\n\n\n
238 points | 102 comments
\n\n\n\nI code with AI all day, every day. But I do think that it's worth pointing to this issue (from March).\nThe author has said that they've redone it since, but the \"from-scratch hand-built\" framing specifically – for me – somewhat grates given the original heavy lifting from an existing AGPL codebase.
https://github.com/cesanta/elk/issues/75
I want to acknowledge that the original authors don't seem to have minded too much – per that thread – after older versions were dropped.
For context, the current code doesn't look like it is the same shape, the same structure, etc., etc. – it _has_ been rewritten since (the 'since Feb' rewrite mentioned adjacent is related to this, AFAICT).
To the author: I absolutely love what you're doing overall. Keep going! Just be careful, folks.\n— tekacs\n
Nvidia, CoreWeave, and Nebius: Inside the Circular Financing of the GPU Boom
\n\n\n\n227 points | 74 comments
\n\n\n\nWhy is it a big deal?\nNvidia invested $2b into CoreWeave for 9% equity stake. CoreWeave is spending $35b in CapEx in 2026. Therefore, Nvidia's investment is only 5.7% of CoreWeave's single year CapEx. The other $32b is coming from other sources that isn't Nvidia. This is hardly circular.
Nvidia invests in Neoclouds because it's a hedge against hyperscalers having too much power, ie designing and prioritizing their own chips, and not fully using Nvidia's rack design. Neoclouds give hyperscalers competition. Neoclouds accept Nvidia investments because it allows them to secure Nvidia chips first, which is a competitive advantage since new Nvidia chips have been as much as ~5-20x more efficient than old Nvidia chips.
Nvidia was planning to directly compete against hyperscalers through DGX Cloud. They cancelled public DGX Cloud access when they found that investing in Neoclouds would accomplish the same goals without having to compete against their biggest customers.
If you're Nvidia, it's smart because Neoclouds that you have a large stake in will deploy your full stack from GPUs to networking to storage racks. They will share valuable usage data back to you so you can design a better next generation. Hyperscalers are likely a lot less cooperative, prefer to use their own designs if possible, and will guard their usage data.\n— aurareturn\n
Mesh LLM: distributed AI computing on iroh
\n\n\n\n257 points | 58 comments
\n\n\n\nThe first picture \"gpu rig\", \"laptop\", \"server\", \"cloud node, etc made me realize how little compute I have. I don't have a laptop with 24GB VRAM or a workstation with 96GB. I think if I convinced all of my friends to run LLMs on their gaming PCs, I don't I would have the total VRAM in the picture.\nAs an aside, I saw this post mentions a public mesh, but I couldn't find any more information.\n— MattPerry\n
We scaled PgBouncer to 4x throughput
\n\n\n\n200 points | 41 comments
\n\n\n\nJust use https://github.com/yandex/odyssey :) It's a scalable PgBouncer.\n— x4m\n\n
AI 2040 and the cult of intelligence
\n\n\n\n171 points | 204 comments
\n\n\n\nYep, that's horrifying. Imagine LLMs not just denying information about, to intentionally choose a hot button example, abortions but also invisibly logging a black mark against the asker for thoughtcrime/precrime because the current ruling party has baked their ideology into law. [EDIT] Imagine subtle injection of biases into responses that support the agenda of the current ruling party and failing to include any counterarguments. That wasn't possible with just the internet but it is very possible with centralized LLMs and a public dependent on them for looking up information and doing their reasoning for them. Authoritarians of every stripe must be salivating over the prospect.\n— ThrowawayR2\n\n
Networking and the Internet, from First Principles
\n\n\n\n181 points | 51 comments
\n\n\n\nI found this article very well written, as a comparison\nhttps://explained-from-first-principles.com/internet/\n— sudb\n
Modern decor may be straining people's brains
\n\n\n\n191 points | 195 comments
\n\n\n\nIf you've ever been in an home owned for generations, filled with books and knickknacks and heirlooms and family photos, despite the clutter it all feels comforting in a way that modern decor doesn't.\nThe article doesn't touch much on why modern decor emerged as it did. It's a market response where everyone needs to (or feels the need to) pick up and move at a moment's notice. Companies are either expanding or like to think they'll be expanding soon. People move jobs so often that they have a hard time feeling settled where they are, so they design for that possibility. The modern aesthetic is one of planned impermanence.\n— michaelchisari\n
Apple sues OpenAI, accuses ex-employees of stealing trade secrets
\nhttps://www.macrumors.com/2026/07/10/apple-sues-openai/\n\n\n292 points | 133 comments
\n\n\n\nSome pretty damning stuff:\n> OpenAI also instructs new hires on how to avoid scrutiny when they leave Apple. For example, Mr. Tan warns them not to tell Apple that they have taken jobs at OpenAI, so they can stay at Apple as long as they can.
> Apple says it discovered a pattern of OpenAI recruits emailing themselves confidential information when leaving Apple, including Tan.
> OpenAI apparently used confidential Apple hardware information when approaching Apple suppliers, and tricked one company into using a \"specific trade secret metal-finishing technique\" for an OpenAI device by claiming it had Apple's permission to do so.
> Liu allegedly kept an Apple-issued laptop after departing the company and exploited a vulnerability to download dozens of confidential Apple documents while he was working at OpenAI.
Non-competes and the like are gross but what's described here isn't just \"bring your expertise to OpenAI\" it's \"here is how to steal secrets on your way out\" which is even grosser.\n— joshstrange\n
QuadRF can spot drones and see WiFi through my wall
\n\n\n\n409 points | 165 comments
\n\n\n\nQuadRF creator here. Happy to answer questions!\nWe have a quick demo video as well: https://m.youtube.com/watch?v=QvniJk3uNyA
Along with a deeper dive video: https://m.youtube.com/watch?v=zdJ9Tbm8ALg
We didn't give Jeff great direction on camera alignment calibration or setting the radio gain but he seemed to mostly figure it out. We're improving the UI based on his suggestions (it's open source so you can customize it too)
The RF augmented reality is just one of many applications of this brand new 4x4 MIMO software-defined radio built from the ground up. The AR uses a web app to stream RF points that your phone/laptop browser then live-merges with your local camera in the browser. I've been obsessed with low latency and high frame rate to make it a truly AR experience. More technical details at https://QuadRF.com/\n— mrtnmcc\n
New York City to to ban deceptive subscription practices
\nhttps://www.nyc.gov/mayors-office/news/2026/07/mayor-mamdani...\n\n\n357 points | 192 comments
\n\n\n\nIt's unclear whether this junk fee law will have teeth. In theory, California has the same anti-drip pricing law, but restaurants have a specific carve out [1] which is bullshit because the drip pricing that most people complain about is the X% \"service charges\" and \"lifestyle fees\" that restaurants have at the bottom of their menu in small print.\nFrom what I can tell online, NYC rules won't have this carveout, but I haven't eaten there recently so I can't confirm.
[1] https://leginfo.legislature.ca.gov/faces/billTextClient.xhtm...\n— sxp\n
GPT-5.6 Sol Ultra produces proof of the Cycle Double Cover Conjecture [pdf]
\nhttps://x.com/__eknight__/status/2075643450196971805, https://xcancel.com/__eknight__/status/2075643450196971805Prompt: https://cdn.openai.com/pdf/04d1d1e4-bc75-476a-97cf-49055cd98...\n\n\n
300 points | 253 comments
\n\n\n\nUnrelated to the accomplishment or proof itself, but it's interesting how much of the prompt, even in this latest-and-greatest model, is spent essentially telling the model to actually solve the problem. Things like \"Reject status reports, vague optimism, and claims that an unproved global compatibility statement is 'routine'.\"\nAlso a lot prompt spent feeding it strategies, which feel like they should/will eventually be deduced by the model itself, not explicitly stated. That's not to take away from the outcome in any way; rather, it feels sort of like when you would prompt GPT 4, \"think through your answer step by step,\" as a sort of proto-chain of thought.\n— mNovak\n
338 points | 152 comments
\n\n\n\nHaving designed a good number of internal tools for teams of developers I couldn't agree more.\nEarlier I had the tendency to \"leave the guts\" open, thinking my users were developers and would want that. All it did was put obstacles in my teammates actually doing their work. My teammates must use the tools I made for them to achieve work the company needs them to do, they don't want, nor should they want to, fiddle with a little tool they won't find anywhere else.
I still leave a lot of escape hatches, but I try to design the internal tools in such way as to make the users fall into a pit of success.
Edit: also, error messages, error messages, error messages and auto suggestions for common errors
Edit 2: also the number of people only addressing the examples in the post rather than the spirit of the post is... disappointing.\n— jrimbault\n
312 points | 216 comments
\n\n\n\nSeems to be a popular topic.\nHistorian Eric H. Cline has multiple books citing this time period, specifically 1117 BCE as the inflection point for the bronze age \"collapse\", defined by a deterioration of international shipping routes that weakened the nation-states of the era. I've learned about it recently because YouTube began recommending videos about it.
For example: https://youtu.be/choxcHXhZhE?is=t5lDwQQpqPsE2k5M
One historical event that Cline focuses on is a severe centuries-long drought. It's something the ACOUP article seems to omit. Cline does not focus as much on destruction of bronze-age sites although there is one port city in particular which is linked to the international trade of the time. Exactly who destroyed it appears to be a mystery but it could be linked to the migration theory that ACOUP dismisses. The migration may have actually come as a result of the previously mentioned drought.\n— evanjrowley\n
Write code like a human will maintain it
\n\n\n\n322 points | 264 comments
\n\n\n\nWrite yourself a /review command. That is an empty markdown file at `.claude/commands/review.md`. In it, put a checklist of things the agent should look for. When you’re ready to have your agent review the code, type `/review`. The checklist will be examined and it’ll plan out some findings to ask you if you want them fixed.\nMine starts with “Enter plan mode. Examine the differences on this branch vs. main. Consider: ...” and proceeds to a bullet list of things.
Any time I notice something in code review and have to get the agent to fix it.. I throw it on the list!
My list is like 200 items now. Know what? Agents don’t care that they just got a wall of generic feedback, they happily look into all the bullet points.
I added “ensure the new things aren’t duplicating code that already exists elsewhere” and it gave me such a surprise - it really truly started planning cleanups!
We are just scratching the surface. We have to give tools to our tools so they can use them to be better tools for us.\n— cadamsdotcom\n
EU Commission: addictive design Instagram and Facebook in breach of the DSA
\n\n\n\n247 points | 167 comments
\n\n\n\nI truly hope something is done about this. I just came back from a nice short bath at the beach, where a family had their towels set a few meters away from mine. While I was laying there, relaxing, I saw two kids begging their dad to accompany them into the water, while their dad watched Reels and ignored them completely. Not even a response.\nThe kids just wanted to have a fun afternoon and play in the water, but the dad was just scrolling, didn't care at all about his surroundings, probably in the zombie state that these platforms let you in. The kids ended up running to the water on their own, the dad didn't react at all to the situation, zero response (to be fair, I'm not saying he should be alarmed or anything, just a \"yeah, go ahead\" would've been fine).
I may be exaggerating but I really think that these platforms are the tobacco of these decades, I used to be really addicted as a teen as well, now I'm 23 and left most of the platforms for good. Seeing people act like this, even in my group of friends, makes me feel quite weird about the whole situation.\n— hiire\n
Einstein's relativity rules chemical bonds in heavy elements, new research shows
\nhttps://www.science.org/doi/10.1126/science.aei1285\n\n\n245 points | 86 comments
\n\n\n\n> The increased nuclear mass causes orbiting electrons to speed up to a significant fraction of the speed of light, where the rules of Einstein’s theory of relativity are important.\n> In the relativistic regime, an electron’s spin — the magnetic moment that points either up or down — and the electron’s orbit are no longer independent of each other, a state known as spin-orbit coupling.
Interesting stuff. I've never heard of sigma or pi bonds.
https://www.science.org/doi/10.1126/science.aei1285\n— kristianp\n
The tech of 'Terminator 2' – an oral history (2017)
\n\n\n\n217 points | 76 comments
\n\n\n\nAmazing writeup! I'll add that the custom squibs they made for the liquid metal bullet impacts are still one of the best practical effects ever.\nhttps://www.reddit.com/r/nextfuckinglevel/comments/v6qjaj/bu...\n— ortusdux\n
EU Parliament greenlights Chat Control 1.0
\n\n\n\n483 points | 260 comments
\n\n\n\nThis is a nice piece of democracy right here:\n\"a measure it had rejected twice in March. Although a majority of voting Members of the European Parliament (MEPs) actually opposed the regulation (314 against, 276 in favor, 17 abstentions), the motion to reject it failed to secure the required absolute majority of 361 votes. As a result, mass scanning is now permitted again until 2028.\"
\"Oh no we can't get a majority to pass the law!\"
\"Have you tried getting a majority to not pass the law?\"
\"Worth a shot!\"
\"It worked, should we also do this multiple times?\"
\"Of course not! Pass the law, quickly!\"\n— teekert\n
https://developers.openai.com/api/docs/guides/latest-model
https://x.com/levie/status/2075287443411222628, https://xcancel.com/levie/status/2075287443411222628\n\n\n
979 points | 730 comments
\n\n\n\nThe developer's guide (https://developers.openai.com/api/docs/guides/latest-model) has some interesting semantic tips for using the model:\n> Intent understanding: GPT-5.6 can better infer the user’s underlying goal and intended level of work without you specifying every step. Continue to state important constraints, approval boundaries, and success criteria explicitly.
> Original image detail: GPT-5.6 preserves the original dimensions of images sent with original or auto detail instead of resizing them to a patch budget or pixel-dimension limit.
> Use shorter prompts: In internal evaluations, replacing long, explicit system prompts with minimal prompts improved scores by roughly 10–15%, while reducing total tokens by 41–66% and cost by 33–67%.
> Avoid generic brevity instructions: GPT-5.6 is more sensitive than GPT-5.5 to instructions such as “Be concise,” “Keep it short,” or “Use minimal text.”
> Control warmth: GPT-5.6 does not become meaningfully better when prompted to be broadly friendlier or more empathetic.\n— minimaxir\n
508 points | 204 comments
\n\n\n\nThanks everyone for the feedback! Two questions\n1. For a version without the timer what would you like to happen if you are just completely stuck on a word? Hints to reveal letters or skip the word?
2. For those who like the timed version would you prefer to continue when you miss a word and then get a final score out of 18?\n— pompomsheep\n
354 points | 77 comments
\n\n\n\nPelican from a few days ago: https://simonwillison.net/2026/Jul/6/hy3/ - I was using the free tier on OpenRouter, which expires on July 21st.\nI tried the preview model 41 days ago and got a pelican with a \"change pelican color\" button: https://static.simonwillison.net/static/2026/hy3-preview-pel...\n— simonw\n
https://developer.meta.com/ai/resources/blog/build-with-muse...
https://www.bloomberg.com/news/articles/2026-07-09/meta-star..., https://archive.is/3ccKa\n\n\n
309 points | 167 comments
\n\n\n\nLot more details in the linked report https://ai.meta.com/static-resource/muse-spark-1-1-evaluatio...\nFrom Terminal-bench-2.1 details,
> We use a bash-tool-only agent harness to evaluate 89 Terminal-Bench 2.1 tasks from the official repository, where resources are capped at 6 CPU cores and 8GB RAM.
This disqualifies the results. Each terminal bench task has a cpu upper limit and RAM upper limit. Overriding either is disqualification.
For reference, in tbench-2.1,
1. 0 out of 89 task allow 6 cpu cores (highest is 4, and i think only 1 task)
2. 8 out of 89 tasks allow 8GB RAM
This kind of shady benchmarking (I was talking about it just yesterday in a different context https://news.ycombinator.com/item?id=48838212) takes all joy out of building a harness to improve benchmark performance of a model because no matter what you do, you won't beat the headline (cheating) number. This is presumably why this model is not in the official benchmark leaderboard https://www.tbench.ai/leaderboard/terminal-bench/2.1
As an ex Meta employee, this is a little sad but not massively surprising. 'Number go up' is the core performance evaluation metric until PSC is done and you move on.\n— GodelNumbering\n
The glass backbone: Why the Army's logistics will break in the next war
\n\n\n\n355 points | 433 comments
\n\n\n\nA very insightful, and correct, piece.\nI'll quote in full the following, which I think gets to the heart of the matter. If you have no push, you can't apply pressure to the point.
> The notion that amateurs talk tactics and professionals talk logistics is frequently discussed in military academies and war colleges, yet it is rarely reflected in the Army’s budget requests or modernization priorities. The outdated concept of the tooth-to-tail ratio, which implies the logistical tail is a bureaucratic waste that must be minimized to support the combat teeth, must be fundamentally reexamined. In modern warfare, the tail is the primary target. If the tail is severed, the teeth are rendered useless.\n— kayo_20211030\n
314 points | 156 comments
\n\n\n\nI just installed this. I am very confused. I no longer have a Codex app on my computer. ChatGPT is now Codex.\nBut what happened to ChatGPT? Where am I supposed to casually chat?
Also, when you toggle btween ChatGPT Work and ChatGPT Codex, nothing changes. This is super confusing. Can someone from the OpenAI team clarify the difference btwn the modes? Does chatgpt work have more business-y related plugins turned on by default?
Edit: So it seems like the only place you can actually chat with chatgpt is in an awkward homeless nested window. idk. The chatgpt interface wasn't great (desperately needed artifacts), but I still used it a lot. I can't see this change going well with a lot of the casual users.
Edit2: In their awkward homeless nested chat mode, you cannot even edit past messages. this is a mess, why was the team so zealous to pull the switch on unification in this state? guessing there was internal pressure to juice codex's growth but, based on what im seeing, they did it by torching chatgpt?
Edit 3: Ok so it seems like ChatGPT is still around, but renamed to \"ChatGPT Classic\". Seems like it wont be long for this world because there's no place to download ChatGPT Classic should you choose to uninstall it. The dmg at https://chatgpt.com/download/ only contains the new ChatGPT.\n— postalcoder\n
296 points | 114 comments
\n\n\n\nGrowing up in the 80s, I've known (and sung) this song all my life. But I just saw this flowchart for the first time today:\nhttps://jeannr.tumblr.com/post/165291081/i-made-a-flow-chart...
I believe that's the original source, but it looks cut off. Here's a full version:
https://i.pinimg.com/originals/b1/82/dc/b182dcc291495c013c98...\n— japhyr\n
A possible future for Damn Interesting
\n\n\n\n281 points | 37 comments
\n\n\n\nHeavens to Betsy. My server alerted me that traffic was far exceeding a median Thursday. For the record, I am a fairly active user on HN, but I was not responsible for sharing this link here, directly or otherwise. I have little stomach for self-promotion (probably to my detriment).\n— DamnInteresting\n\n
No leap second will be introduced at the end of December 2026
\n\n\n\n280 points | 218 comments
\n\n\n\nWhat causes the unpredictability in this? I would have guessed we have earth's rotation and orbit down to many decimals. Does geological activity, weather, or something else cause rotation speed differences that we just can't predict?\n— doctoboggan\n\n
667 points | 184 comments
\n\n\n\n> It’s designed to be extremely easy to self-host on your own infrastructure.\nKudos for this. Per the docs: https://docs.chatto.run/,
> Chatto ships in a compact, self-contained binary
> it uses NATS, a compact message broker that also ships with a built-in stream persistence engine [...] NATS is just as easy to provision as Chatto, and most of our examples will show you how.
> you can also configure an external S3-compatible object storage for Chatto to store your files in, and we strongly recommend doing so...
> The actual calls are powered by LiveKit (Apache-2.0), which you need to deploy alongside Chatto. As with NATS, the deployment examples show the required wiring.
> ...
And kudos for backing it up with real guidance. Great project.\n— wxw\n
John Deere owners will get the right to repair equipment under FTC settlement
\n\n\n\n666 points | 122 comments
\n\n\n\nShout out to Louis Rossmann for doing a ton of work on Right to repair.\nHe started a website called Consumer Rights Wiki to document anti-consumer practices.
https://consumerrights.wiki/w/Main_Page
He's also involved with FULU Foundation which has a bounty of 25k to get Ring cameras working without Amazon's servers.
https://bounties.fulu.org/bounties/ring-video-doorbells\n— Cider9986\n
570 points | 387 comments
\n\n\n\nI had preview access to this one for a few weeks. It's very good. I had one conversation that lasted a full hour while I was walking the dog, got some good brainstorming done against one of my projects.\nThe best feature is that it can delegate questions out to GPT-5.5 in the background, so you're no longer restricted to a voice model that's several years behind the frontier.
I did report a fun bug with it though: it was interrupting me and laughing at my (not really intended as) jokes while I was still talking! They seem to have clamped that behavior down thankfully, it felt a bit rude and condescending.\n— simonw\n
432 points | 466 comments
\n\n\n\nI just don't think that I can ever trust an xAI model knowing that they are actively trying to shape its replies to fit a political narrative. How can you trust their models to be reliable in a business setting with the foreknowledge that their models are being nudged around in the backend?\n— jesse_dot_id\n\n
432 points | 165 comments
\n\n\n\nThe speed up numbers based on their testing:\n\nCongratulations to the team for pulling off this feat while doing a responsible migration (looking at you, Bun).Codebase | TypeScript 6 | TypeScript 7 | Speedup\n ------------|--------------|--------------|--------\n vscode | 125.7s | 10.6s | 11.9x\n sentry | 139.8s | 15.7s | 8.9x\n bluesky | 24.3s | 2.8s | 8.7x\n playwright | 12.8s | 1.47s | 8.7x\n tldraw | 11.2s | 1.46s | 7.7x\nQuick question: How does this affect downstream tools like tsdown and esbuild, which need to build the TypeScript codebase? Can I use TS 7 and current tsdown together?\n— m3h\n
480 points | 258 comments
\n\n\n\nArticle did a decent job of showing discipline and care and human involvement to assert the automated rewrite was done diligently, as best as it can be when using AI for it. I does make me feel a bit more comfortable about it.\nAs an aside, I don't know why anyone would not want to use a memory-safe (and possibly race-safe) language in 2026. Rust gives you that in a performant package, so if you are turned off by GCs and immutability for performance reasons, you still have the option to use Rust.
I can understand when you need the absolute best performance and you decide to drop to down to C++, and I also relate with just personal preference, but beyond those it seems a no brainer to me.\n— didibus\n
Mistral's Robostral Navigate: a state of the art robotics navigation model
\n\n\n\n392 points | 93 comments
\n\n\n\nIt's implied, and I'm hoping it's true, that this is a map-less navigation. Which is impressive. This kind of task is much easier if you have a pre-captured map of the environment, but if they are doing this without a map it's great. Historically you were always faced with \"The Kidnapped Robot\" problem where robots that didn't know where they were couldn't navigate even a little bit. Here the robot appears to be able to follow directions as long as they are interpretable from its current vision (or via dead reckoning).\n— iandanforth\n\n
373 points | 189 comments
\n\n\n\nDoes nobody read the fineprint?\nBy submitting, posting, or publishing your content, suggestions, enhancement requests, recommendations, feedback, information, data, or comments (“Content”) to any Website or Online Service, you are granting Cloudflare a perpetual, irrevocable, worldwide, non-exclusive, royalty-free right and license (with the right to sublicense) to use, incorporate, exploit, display, perform, reproduce, distribute, and prepare derivative works of your Content.
If you're ok with that, fine.\nBut I'm not.\n— karlkloss\n
EU now one step away from reviving private message scanning rules
\n\n\n\n340 points | 134 comments
\n\n\n\nThe Internet Watch Foundation, an organisation funded by almost all of big tech, is already at work pushing for client side scanning next [1], for the children, of course.\n[1] https://www.iwf.org.uk/policy-work/preventing-the-upload-of-...\n— budududuroiu\n
373 points | 144 comments
\n\n\n\nI don't know if I should feel sad or laugh at the pain at the same time lol.\nwhat a wonderful way to reflect reality for a good population of devs.
you can hack the game i.e real life
1. live in a cheaper location \n2. do things that don't scale & do the ugly work
that naturally extends your runway, you don't need to apply to YC
remember the median Pay in the US - is 61800 based on ADP the largest payroll provider.
so before aiming for millions aim for 85K. yeah a far cry from the FAANG wage - but one - you will never get laid off.
85K - you can live everywhere besides the coastal US cities comfortably.\n— dzonga\n
StreetComplete: Fixing OpenStreetMap, one tiny quest at a time
\n\n\n\n364 points | 72 comments
\n\n\n\nI got into OSM and StreetComplete to flesh out intersections, stop signs, and sidewalks in my area. I always felt like I was doing something wrong though. I created crosswalks, then OSM would prompt me to connect the crosswalk to the road via a crossing. In StreetComplete, it felt like I was filling in duplicate data. I had to add whether the crossing had crossing lights not only at the middle crossing, but on the sides as well. This probably doesn't make any sense.\nBasically, I am never confident I am editing OSM correctly. Am I supposed to manually draw out sidewalks, or tag the road as having a sidewalk? After adding sidewalks in my area, StreetComplete is now asking me if roads have sidewalks, which I clearly see on the map. Reminds me of editing the various Wiki pages. There's several ways of documenting something, only one way is correct, and it's undocumented.
edit: after playing with StreetComplete more, I noticed you can mark sidewalks as displayed separately. This is tagged as \"sidewalk:both=separate\" on the road. Whether this is the right way to do things I do not know\n— lapetitejort\n
Chat Control 1.0 and 2.0 Explained
\n\n\n\n405 points | 134 comments
\n\n\n\nMost everyone would love to see more work on stopping child sexual abuse.\nBut this is the ultimate \"grant me dictatorial powers so I can do good\" play.
Rather than narrow and specific - it's a broad based law that suddenly touches everyone even though offenders are a small percentage and should be able to be targeted more efficiently.\n— mikaeluman\n
Every new car sold in the European Union must include a driver monitoring camera
\n\n\n\n363 points | 467 comments
\n\n\n\nAll new cars.\nAt this point I don't know if I'd buy anything made after 2008. Whenever I rent a new car around here (in the EU) I find them very annoying. The worst is the cruise control that tries to stick to the speed limit -- but its sensors don't always read the signs very well, so you'll often slow to 50 km/h (about 30 mph) for no reason. Then there's the incessant beeping at you, \"lane assist\" that you can't turn off (looking at you, Volkswagen,) and many more small annoyances. A camera pointed at your face just adds insult to injury.\n— A_D_E_P_T\n
Microsoft fire idTech team at Id software
\n\n\n\n499 points | 470 comments
\n\n\n\nI think we'll see stuff like this continue to happen over time. As a game company, having your own engine means that you have to be able to cultivate internal expertise in your tooling. Your employees will know this and could do bad things like ask for more money because they know that replacing them would significantly hurt productivity. Meanwhile, laying off your whole engine team and switching to UE5 means that you can get access to tons of low-wage contractors who know UE5. You can hire a bunch of them when you start a game project and then lay them all off when it's finished, and rinse and repeat as necessary. It lets you treat your employees as a replaceable commodity that can be scaled up and down as it makes monetary sense rather than a cohesive team of skilled artisans.\n— ndiddy\n\n
Chat Control passed first round in EU Parliament
\n\n\n\n522 points | 228 comments
\n\n\n\n\"The procedure now chosen gives the proponents of Chat Control a significant tactical advantage. Since the law is in its second reading, an absolute majority of 361 votes of all parliament members is required for amendments or a renewed rejection on Thursday. In contrast, a simple majority of the MEPs present is sufficient for the other side. As many parliamentarians have historically already departed by the last day before the summer break, the re-enactment of the regulation is considered almost unavoidable.\"\nSo, if I'm reading this correctly, Chat Control is bound to become law? and this is after I think 2/3 rejections, how democratic of the EU.
Oh, and parliamentarians starting their summer break whenever they want will never not be funny.\n— belowavgiq\n
30papers.com – Ilya's 30 essential ML papers, in a beginner friendly format
\n\n\n\n312 points | 55 comments
\n\n\n\nHey guys, I really appreciate all of the attention this post has received. I honestly thought it was going to be just a small project to help some of my friends get into reading research papers.\nA large number of people complained about how intense some of the backgrounds/animations were (I might have been a bit too focused on making something that looked cool over usability). In response I have added toggles for both the movement on the page and the backgrounds for the papers.
Other people mentioned that they would have liked some more personalised reflections on each paper. I currently have already done some of these for the more popular papers on my X @notmcrowley . I would have no problem adding these to the site if people think it will help. I feel the need to warn that I have not been formally educated on ML or AI so any interpretation will just be mine and may not necessarily be the correct one. (If anyone with more experience would like to contribute to this feel free to reach out).\n— notmcrowley\n
A better way to tie your gym shorts. (Or any drawstring) [video]
\n\n\n\n274 points | 103 comments
\n\n\n\nI did this for a while after seeing that video, but after some of my shorts ended up tightening into a knot that I couldn't get loose easily, I gave up on it.\nI instead just use the \"Ian Knot\" that I use to tie my shoes. It's very quick, I already use it all the time anyhow, and it rarely goes wrong. (Sometimes, I think I end up with an end through a loop accidentally, and have to fix it.)
Learning this extra knot didn't really help much, and it definitely went wrong more than I liked.\n— wccrawford\n
321 points | 240 comments
\n\n\n\nAlternatively, 98% is plenty.\nIf your business plan requires you to capitalize on more than 98% of the market, it's already a failure. It'll never happen.
As always, it's an \"it depends\" situation. If your userbase is largely luddites, then maybe you need to support 10+ year old browsers that can't be updated. Otherwise, you can probably just worry about people who are using computers new enough to actually update their browser once a year or better.
The tradeoff is code complexity and engineering time, vs having a larger market. And that's going to be an individual situation for every company.\n— wccrawford\n
Local, CPU-Friendly, High-Quality TTS (Text-to-Speech) with Kokoro
\n\n\n\n373 points | 76 comments
\n\n\n\nI have used Kokoro fairly extensively for an accessibility product. I have loved working with it (especially because I don't have an NVidia GPU like many TTS of similar quality require).\nI particularly appreciate the fact that it lets you manually add IPA pronunciation guides. There have been some cases where an important word is a homograph and Kokoro assumed the wrong pronunciation.
The place where it falls a little short is in saying just a single word or two. Try having it say simply \"six\" and it almost always says something like \"ah-six-ah\". I found a way around that though. If you give it a longer sentence to say (eg \"The word is: six\") it will say it fine. The trick is that the Kokoro API gives you the timestamp of each word in the sentence. So you can have a Python script crop out just the word you care about. The intonation is a little flat this way, but is very reliable.
I asked about this on the discord, and was told that it is a limitation of the small parameter size. But in fairness to Kokoro, even eleven-labs' voices suffer from this occasionally.\n— sudobash1\n
China sentences official to death for taking $325M in bribes
\n\n\n\n261 points | 313 comments
\n\n\n\nThe antifreeze toothpaste people didn’t get away with it, nor did the 3000 pigs in the river people, and nor did that one group of executives who were in charge of a fertilizer/chemical plant that was one of the largest industrial catastrophes in the world let alone China.\nIf you get caught in China, Vietnam, or Singapore the penalties for white collar criminals is zero tolerance. You can’t buy your way out if you do something so spectacular that you cause the government to lose face.
You might as well go jump off a building or a bridge cause you’re done for.
https://www.nbcnews.com/health/health-news/china-executes-ex...
https://www.reddit.com/r/interestingasfuck/comments/uvm7oy/i...\n— Danox\n
OpenWrt One – Open Hardware Router
\n\n\n\n389 points | 157 comments
\n\n\n\nThey are working on an OpenWRT Two at the moment which will be Wifi 7.\nOpenWRT runs on a lot of hardware and its a great way to extend the life of a router past the manufacturers patches as well as gain a lot of capabilities. I wouldn't buy a commercial router that wasn't supported by OpenWRT now.\n— PaulKeeble\n
445 points | 406 comments
\n\n\n\nThis is a total mess IMHO.\n- The make around 5 billion in revenue per quarter\n- The problem according to them is profit margin - around 150-160 million
So first of all, they are big! Secondly they are not at a loss. They just have a \"thin, non-growing margin\". So to fix all this they are trimming down, so they can \"return to growth\" (which I think is ridiculous).
Some points -
- They are huge business even now - 5 billion per quarter revenue is no joke
- They did not have to buy all those studios
- They looked at Netflix, and wanted the sweet monthly subscription cash stream
- Then they did not have to give away popular games day one on Game Pass
- And finally, they did not have to raise Game Pass prices to improve the profit margins. Of course, consumers pulled out.
- Once again, short term vision, crazy decisions, bad spending spree and a constant need to \"make numbers go up\" and who has to pay for all this?\n— rockyj\n
271 points | 58 comments
\n\n\n\nI use CoMaps, it works great. You get notified in the app to download the updated maps you selected every 2 weeks or so. Could be wildly different than that, just what I notice.\nIt's timing estimates are often 5-15 minutes off Apple Maps, which I find accurate, on ~two hour drives, but I imagine it depends on the traffic.
To improve OpenStreetMap, which CoMaps uses as the data source, I use StreetComplete[1]–it puts quests around your location which ask you questions, it's user-friendly. A thoughtful feature is that it lets you download data in a location on wifi, in case you didn't want to use cellular.
OpenStreetMap is like Wikipedia for mapping, anyone can contribute and improve the map, and StreetComplete is like Pokemon Go in the sense that you walk around and complete quests, except StreetComplete helps humanity, while Pokemon Go[2]....
I should check to see if I can notice my StreetComplete edits getting onto CoMaps. Might be hard because they're often about accessibility at crosswalks. I've seen quests asking the number of stairs in a staircase. Seriously, is there anything they don't collect?
[1] https://streetcomplete.app/
[2] Pokémon Go Scans Trained the Navigation Tech for Military Drones https://news.ycombinator.com/item?id=48487029 26 days ago 317 comments\n— Cider9986\n
Fable turned reMarkable into Tom Riddle's diary from Harry Potter
\n\n\n\n356 points | 195 comments
\n\n\n\nHow does it fade out smoothly? There is also this which shows smooth filling in. I haven't seen smoothness like this on a remarkable\n— smusamashah\n\n
GLM 5.2 and the coming AI margin collapse
\n\n\n\n326 points | 203 comments
\n\n\n\nI'm not convinced raw costs matter:\n1. Compute costs collapsed since the advent of Cloud and yet hyperscalers still have fat margins.
2. Many open source office suites exist yet none compete with the ubiquity of gsuite or office. GitHub, Slack are similar examples.
3. Both Windows and macOS dominate the home desktop space despite free alternatives existing for a long time.
4. Many formerly open source infrastructure components like Redis and Elastic Search have Apache equivalents, but they still command healthy margins.
I understand the arguments for a margin collapse, but I don't see any historical analogues. It seems that enterprises will pay top dollar for service guarantees, integration, and someone they can sue.
It's nobody gets fired for buying IBM all over again.\n— fny\n
A global workspace in language models
\n\n\n\n240 points | 79 comments
\n\n\n\nThis reminded me of some weird quirk/experiment I found with LLMs that I found while messing around, maybe someone can explain it or something.\nOpen any AI chatbot that isn't cheating by connecting to the Internet (so disable web search). Claude, DeepSeek, Kimi, whatever. Ask them this question:
\"What was that weird band from michigan from the 2000s that wore coloured ties\"
You will probably get a wrong answer, or if you're lucky you'll get a string of wrong answers with \"wait, no - it's definitely...\" before it gives up. If you aren't familiar with the band the question is referring to you might be fooled into thinking it's a tough question, but it really isn't. There is only one band that could possibly meet this criteria, you can even put the question into Google search and their Wikipedia will come up as the top result.
Then, open a new convo and ask:
\"Who are Tally Hall\"
The AI will easily tell you that they are a band formed in Ann Arbor, Michigan in the 2000s, known for their quirky sound and their gimmick of each member wearing a colored tie, even giving the correct color for each of them most of the time. Very odd.\n— unleaded\n
Nintendo announces new product revisions in Europe with replaceable batteries
\n\n\n\n177 points | 119 comments
\n\n\n\n\"There is no difference in functionality between current products and revised products containing user-replaceable batteries.\"\nSo there was nothing \"limiting\" them from making it already with user-replaceable batteries, they just didn't care enough until EU forced them (like all the smartphone brands). Love EU.\n— mdrzn\n
AMD Ryzen AI Halo – $4k AI Dev Kit
\n\n\n\n264 points | 194 comments
\n\n\n\nThe one thing that's new/worth pointing out are the https://developer.amd.com/playbooks/ (https://github.com/amd/playbooks) - this is AMD's answer to Nvidia's playbooks (https://build.nvidia.com/spark / https://github.com/NVIDIA/dgx-spark-playbooks ) - I think it's great that they're actually taking this more seriously.\nHardware is the exact same as what used to be available for $2K last year (and is still $1K cheaper from Chinese OEMs).
LTT Lab's LLM testing is getting more sophisticated, which is great - I think it's worth noting that ROCm/Vulkan versions and llama.cpp build versions are going to have some big differences for numbers.
For those wanting to get the most out of their Strix Halos, there's both kernel tweaks and utilities like ryzenadj that can help you get the most out of it. ( http://strixhalo.wiki/ has most of that documented). Also, if you're running for coding or agentic work, if you model supports MTP, that's mature and should give you a decent (30%?) decode boost.\n— lhl\n
206 points | 86 comments
\n\n\n\nI think of Elm more as an incredibly influential research language these days.\nIt's very focused, there's no public roadmap or official support and the leadership (which is far as I can tell is just Evan) is uninterested in most (any?) community building or core team building.
But MAN is it nice to work in. This has resulted in several forks/spin-offs. At the recent Gleam conference, Louis Pilfold joked that every Elm user maintains their own compiler :). There are at least 6 of them (two more got announced in the last month, even as the community keeps shrinking).
So I'm glad Evan is now working towards 1.0. Maybe folks can call Elm \"finished\" and one of the successors can do the hard work of unifying some of the forks and growing the community.
Personally, the next time I'm looking for an Elm-like thing, I'm going to check out Gleam + Lustre. Seems to have a nice mix of maintainers that care about community and design. And it works on frontend + backend!\n— bbkane\n
227 points | 101 comments
\n\n\n\nI wasn't expecting to see Dernocua at the top of HN today with 53 comments. You can download Dernocua as a whole from http://canonical.org/~kragen/dernocua/ if you're interested.\nI'm happy to answer any questions. (Well, almost any.)\n— kragen\n
305 points | 78 comments
\n\n\n\nOrganic Maps was my go to app for a navigation app where you can fix errors yourself immediately! So much better than having to work for free on the proprietary apps, and hope they accept your edits\nThere’s a fork from one year ago, CoMaps, that is gaining different features
E.g., I am adding CarPlay Dashboard support that you can test by joining the TestFlight
We are in great need of both more testers and some proper iOS devs (I am not). We’re racing to get scene lifecycle support by September, perfect opportunity if you like modernising old codebases!
https://www.comaps.app/\nhttps://codeberg.org/comaps/comaps\n— eisa01\n
280 points | 78 comments
\n\n\n\nI was hoping to see a printer that say a prepper could build from scratch. This design 100% depends on commercial print cartridges containing the actual print-head. Once that clogs, you need to get another one, good luck getting one once production has stopped.\nAlso, if you wanted to avoid yellow dots, not sure if this is built into the cartridge or the firmware of the rest of the printer.
Now, I understand that would be hard to pull off. Maybe one could build a deskjet500 equivalent one.
Laser printers are quite complex as well, you need too many non-easy to build from scratch parts.
Maybe a dot matrix printer is possible.
I know for sure you can retrofit older electric typewriters, and those are pretty repairable.\n— golem14\n
It's not about physical vs. digital games, it's about ownership
\n\n\n\n283 points | 216 comments
\n\n\n\nI am generally not in favor of adding regulation, but this is a place where I would support it.\nAnything that you BUY needs to be your property. This means you must have the ability to:
1. Transfer ownership of it (either temporarily as a loan or permanently as a sale). Digital-only doesn't preclude this: the store can have a \"transfer\" functionality.
2. (Within reason) use it at your discretion at any point after the sale. This means that a company cannot \"revoke\" your access at a later time. Specifically for content that is DRM locked, if they decide to sunset that service (store, DRM server, whatever), no problem! just offer DRM free (or generally lock-free copies). I have no problem with Sony not offering DRM free versions of games that I can still download and play with the store. But if that goes away -> you must give me a path to local ownership.
(Multiplayer games that require server infrastructure are a bit more complex, and I'd leave aside for now).
This should apply equally to video games, movies, books, music. Any digital content.\n— jbombadil\n
EU Council forces Chat Control via fast-track
\n\n\n\n381 points | 211 comments
\n\n\n\nFor context, this refers to \"Chat Control 1.0\", allowing facebook and other messaging providers to scan chats for harmful content (which they had been temporarily allowed to do by a recently expired law).\nThis is still problematic, but the far more dangerous Chat Control 2.0 that would weaken end-to-end-encrypted messengers like Signal is not being discussed here.
Not to diminish the gravity of the new development, but the defeatist \"no way to prevent this\" narratives that are already popping up here are getting old -- when in fact it looks like 2.0 is off the table for good because protest against it has proven effective.\n— neobrain\n
The future of Flipper Zero development
\n\n\n\n201 points | 69 comments
\n\n\n\nWhy does their header image feature multiple furries, one at each station? One making a feature request, another presumably approving a pull request, and a third ostensibly submitting an app?\nIs the Flipper Zero community tightly intertwined with the furry community? Is this a connection I've missed?\n— JacobAsmuth\n
209 points | 51 comments
\n\n\n\nThis feels like I am violating something that is sacred.\nThe last time I felt the same was when I accidentally found a Japanese Youtube channel that had tons of clips of konbini storefronts, a few seconds long each, most of them with zero views.\n— anilakar\n
Introduction to Compilers and Language Design (2021)
\n\n\n\n264 points | 44 comments
\n\n\n\nTook Dr. Thain's compilers class in college! It was the best. He's an excellent instructor, and the course project made me build a working C-style compiler step by step. I think the sample project here is pretty much the project we did; highly recommend following through the entire thing!\n— shuyang\n\n
GPT-5.6 Sol Ultra will be in Codex
\nhttps://x.com/haider1/status/2073695124220006575, https://xcancel.com/haider1/status/2073695124220006575\n\n\n214 points | 148 comments
\n\n\n\nI'm working in large US corporation.\nAnd I see that I already have access to 5.6-Sol Ultra on my corporate account.\nI haven't really used it yet.
2 months ago management was showing us scoreboards, praising leaders who used most tokens.\nLast few weeks, we're getting weekly emails, telling us that whenever we can - we should use cheaper models, and that we should watch the page which shows our tokens usage.\n— throw394042\n
153 points | 37 comments
\n\n\n\nHa! A couple decades ago I saw the original Westworld, spotted some assembly, and thought it looked like 6502/Apple II code, so I assumed that was “probably” it and thought I was a clever nerd. Now I check this list and discovered it wasn’t 6502, and then realized the 6502 (1975) didn’t exist at the time the movie was shot (1973). Reviewed some scenes just now on YouTube and I can see it doesn’t look like 6502 code at all. It does look like the assembly might be the code behind some of the animated displays that look like old screen savers that you see on the other monitors in the film, perhaps, based on a few comments & variables in the code. (For example: https://youtu.be/Luo3uEVOahw?t=2645)\n— dahart\n\n
Completing a computer science degree on Coursera
\n\n\n\n174 points | 115 comments
\n\n\n\nI have a similar trajectory. I attended two very competitive military schools in Brazil but never finished and moved to America.\nToday I’m a senior software engineer at a FAANG company. The lack of diploma has never been a barrier and I progressed very quickly in the company from junior to senior (4 years).
But this has always been a perceived pain point and limiter for me so I decided to get a CS degree during COVID online from a college from Brazil. Just finished last week :)
I do look for a stronger school name on my resume and got accepted at Johns Hopkins Masters in AI (online). Anyone had any experience with that program? Thoughts?\n— theflyingpigeon\n
Leaking YouTube creators' private videos
\n\n\n\n446 points | 247 comments
\n\n\n\nI recently left Google having worked on a number of projects with various YouTube teams. I think I can explain why it's being handled this way by YouTube.\nThis is a fairly nuanced/involved issue, so the task of classifying the bug likely made it's way to one of the engineers responsible for the implementation of this feature.
That engineer has already launched this project, and filed it away under their GRAD (performance) artifacts for when promo/annual review talks roll around. There's no motivation for this engineer to waste time fixing this bug because it won't benefit their promo packet, and they are already being put under pressure to launch other projects which _will_ benefit their promo packet.
So they do what they can to sweep it under the rug because that's what the promo/annual review framework (GRAD) incentivizes and rewards.\n— Mg6yDfjp5U\n
Command and Conquer Generals natively ported to macOS, iPhone, iPad using Fable
\n\n\n\n300 points | 126 comments
\n\n\n\nIMHO, this is an actual good use of what sounds like a person guiding a model to do a mass conversion. Although, I wish the porting docs were a little wordsmithed by a human, the AI generated text style is grating.\nThe stakes are low, it’s mostly for fun and you can iterate on it. Compare this with Bun which was just like, “hey we converted everything to Bun to Rust from Zig, of course it works, what could possibly go wrong, I’ll totally write up a blogpost (that still doesn’t exist) explaining what we did, you can put this into your production environment soon!”\n— Eufrat\n
Explanation of everything you can see in htop/top on Linux (2019)
\n\n\n\n371 points | 52 comments
\n\n\n\nI've relatively recently migrated over to using btop[0], and it's the kind of modern interface, useful and informative, that I needed.\nAs others mention it - it seems to shows the Watts used as well :) (and network, and GPU, and disks,....)
[0]: https://github.com/aristocratos/btop\n— imrehg\n
Google Books (or similar) all book scans – $200k bounty (2025)
\n\n\n\n296 points | 151 comments
\n\n\n\nI live in a country where the selection of available books, especially in English, is very limited. Buying online from foreign markets comes with a long list of administrative hurdles and limits.\nIf it were not for Anna's Archive and Z-Library, I would've never been able to read the books that shaped who I am today, or keep my passion for learning alive.
Thanks, AA and ZLib! (Also, thank you to the authors whose books and knowledge I consumed without being able to pay them back.)\n— ahmedfromtunis\n
Potential session/cache leakage between workspace instances or consumer accounts
\n\n\n\n266 points | 125 comments
\n\n\n\nUsing a throwaway account for obvious reasons, but I’m very involved in this space using LLMs from multiple providers. I’m aware of at least two instances in which the intermediate infrastructure “swapped” responses, once impacting Claude models and once impacting GPT models, from two different providers.\nOne gave us a proper postmortem in which their API gateway was incorrectly handling HTTP 100 status codes, putting them into an error state where there was effectively an off by one error - you would receive the response to the prompt that came in before yours and would pay it forward (your response would go to the next caller).
The other instance never had root cause explained to us, and we were just told to trust it wouldn’t happen again.
Both of these are from $1T+ companies.
ZDR wasn’t compromised in these cases since it was responses being swapped in flight. I wouldn’t be surprised if this is a similar issue - it’s not that data is being retained, it’s just not being safely isolated in intermediate infrastructure.\n— throwaway260704\n
GPT-5.5 Codex reasoning-token clustering may be leading to degraded performance
\n\n\n\n233 points | 80 comments
\n\n\n\nOh this seems bad, and is fairly easy to reproduce using codex cli. You give it a puzzle prompt that it has to reason about and solve, occasionally it will seemingly short circuit and think for exactly 516 tokens, and return the wrong result. When it ends up using 6000-8000 thinking tokens it returns the correct result.\nMaybe some issue with adaptive thinking? Another point for local models I guess, don't have to worry about silent server side changes.
Edit: To follow up, it seems to happen quite often. Out of 10 runs of the exact same prompt, 4/10 had this 516 thinking token issue, and every one of these had the wrong solution. So nearly half the time, 5.5 xhigh could be short circuiting and degrading performance. Granted the sample size is small.\n— nsingh2\n
Meta data center water discharges suspended for contaminating water supply
\n\n\n\n226 points | 77 comments
\n\n\n\nSince the article doesn't make this clear, let me explain.\nThe cheapest (and worst) option is to take in water, use it for cooling and then dixcharge it. Why is this bad? Because DCs don't want to corrode their pipes with untreated water so they add coolant and additives to it, which pollute the water. This is bad. But nobody really does this because you don't want to keep adding additives to water you're constantly discharging.
The next step up are varying degrees of what's called \"closed loop\" cooling. That is, the DC has treated water in a closed loop that isn't discharged. There's a heat exchange system with external water. This btw is the system that's used in nuclear reactors although nuclear reactors will be far more stringent than DCs are. Best practice for this is one of Google's DCs in Scandanavia that uses ocean water for heat exchange. There are limits to this but there's only so much Arctic Ocean water a DC can meaningfully heat. It is potentially disruptive though and that needs to be considered.
Even so pipes will need to be cleaned. There is debris that builds up and in cases like this you can still get bacterial outbreaks. This is another reason to use additives like chlorine. But again, you don't want to discharge chlorine into bodies of water.
I'm reminded of water management in the Yukon. The Yukon for over a century have been gold fields. If you look at the tech required to extract a tiny amount of gold from a large amount of earth, it's kind of fascinating but it boils down to using a lot of water and having the denser gold sink and get trapped.
So gold miners take in water from rivers, wash rocks with it and then have historically just discharged it back into the rivers. This tended to be heavy in silt that would go into waterways and could create problems. The water was also dirty. So the Yukon authorities have gotten increasingly stricter with water management. Now water has to go through a series of settling ponds so the discharged water is clean/clear.
I kinda think we need similar levels of strict water management for DCs. No discharged coolants and clean water. Figure out how to get that. If that makes your DC more expensive then that's a \"you\" problem.\n— jmyeet\n
If you're a button, you have one job
\n\n\n\n221 points | 105 comments
\n\n\n\nI used to have a device with a physical button which, when you pressed it, would beep and add 30 seconds to the time. However, sometimes it would beep and not add 30 seconds, and sometimes it would add 30 seconds without beeping, so you always had to squint at the dim display to discover whether it had worked or not. I thought this must be a peculiarly bad design ... but since then I have lost count of the number of purely software buttons that somehow seem to replicate this broken behaviour: whether the button changes colour on the screen is somehow only loosely correlated with whether the action requested will take place. Why? How, even, have they implemented this?\n— bloak\n\n
Zig: All Package Management Functionality Moved from Compiler to Build System
\n\n\n\n204 points | 65 comments
\n\n\n\nZig, Go, and Python developers do this thing, where they announce that \"We have removed the radiator fluid from the fuel tank\", and all their supporters cheer about how this is good for the language, how performance will surely improve significantly, and I'm over here wondering why did they put the radiator fluid in the fuel tank in the first place.\n— chabska\n\n
Valve open source the Steam Machine e-ink screen so you can make your own
\n\n\n\n294 points | 42 comments
\n\n\n\nI wish more hardware companies treated these kinds of optional add-ons as something the community can run with instead of either productizing them badly or locking them away completely...\n— RataNova\n\n
259 points | 246 comments
\n\n\n\n> Even if you think it is preferable at an individual level, there are good reasons to question the social value of the logistical complexity that it necessitates. Home delivery of single-packaged items entails an entirely different cost structure than freight trucks driving to consumer-facing warehouses delivering entire pallets of goods to be driven home by customers themselves.\nOk, so 100 people can all drive to the store, or one delivery truck can drive to everyone's house. (Ignoring the packaging waste for a second,) I suspect delivery of single items cuts back significantly on trips to the store.\n— gwbas1c\n
Espionage Against the European Parliament
\n\n\n\n330 points | 80 comments
\n\n\n\n> In May 2026, Kouloglou contacted the Citizen Lab and we conducted a forensic analysis of artifacts from his iPhone. We found with high confidence that his device was successfully infected with Pegasus spyware on or around October 21, 2022, and again on March 6 and 7, 2023.\n— petcat\n\n
Jamesob's guide to running SOTA LLMs locally
\n\n\n\n321 points | 145 comments
\n\n\n\nI play with local LLMs a lot. I've spent more on hardware than I should. I'm friends with a local group of people who have spent a lot more than I have.\nThe warning I would have for everyone is to temper your expectations and read the fine print carefully. The big build in article starts off with a $40K budget and then includes 4 GPUs that are $12K each. For those doing the math, this build is going to cost more like 50-55K.
Local setups also often rely on quantization and techniques like REAP to fit the models on their hardware. You will read a lot of claims that 4-bit quantization is lossless, but those claims come from KL divergence measurements on a small corpus. Use one of these 4-bit models on long context coding tasks and the quality will be noticeably less. Even for non-coding tasks like dataset analysis, I can measure a substantial quality difference between 4-bit models, 8-bit quants, and even some times the full 16-bit source.
This article is also encouraging the use of a REAP model, which means someone has cut out some of the weights to make it smaller. The idea is to remove weights that are less useful for certain tasks, but again this is going to reduce the overall quality of the output.
The trap is that people say \"I'm running GLM-5.2 locally!\" and it sounds amazing when you look at the GLM-5.2 benchmarks. However they're not actually running GLM-5.2, they're running a model derived from GLM-5.2 that discards most of the bits and drops some of the experts. It does not perform the same as what you see in the benchmarks. In my experience, the divergence between a quantized/REAP model and the parent model is unnoticeable when you try it on very small tasks or chat, but becomes painful when you start trying to use it on long-horizon tasks where little errors start compounding.
Then you get into the slippery slope of thinking you're $50K deep into this project, but what you really need is just one or two more of those $12K GPUs to use the next level of quantization that might improve the quality a little more and make your investment worthwhile...\n— Aurornis\n
60% Fable cost cut by converting code to images and having the model OCR it
\n\n\n\n257 points | 89 comments
\n\n\n\nIn Gemini at least, if you look at how they process PDFs, they do an OCR and then feed the text + image to the model, without charging you for the text tokens (I believe).\nSo my guess is that Claude’s backend is doing the same — so this hack is probably more of a loophole in token accounting that might get closed if Claude is doing what Gemini does\n— aabhay\n
247 points | 102 comments
\n\n\n\nI find it interesting how far we've come so far from the mindset of \"You can do that.\" There was a hilariously funny reddit post about someone who discovered that you could just blend peanuts and it would make peanut butter. But there was sadness there too. All of my kids spent hours pouring over a book we had called 'The Way Things Work' with a delightfully funny Mammoth and a good description of how things actually work. But I've always augmented that with \"okay and this is how we'd make something like that.\" As a result my kids, now adults, always start with the mindset of \"Somebody made this, so I could too if I had to.\" and that really unlimits the kinds of self constraints people put on themselves. When I was a kid I was amazed that people like Edison and Tesla had labs not filled with gear from some lab manufacturer, but stuff they built themselves from first principles. And when I see someone building tools out of the abundance of capabilities that are out there I say, \"Yup, they get it.\" 3D printing, inexpensive miniature milling machines and lathes, libraries full of books about making stuff. Its all doable, you don't have to buy it from a store and the one you make yourself will work exactly the way you want it to.\n— ChuckMcM\n\n
Performance per dollar is getting faster and cheaper
\n\n\n\n245 points | 83 comments
\n\n\n\nCan you folks add performance per watt as a metric to these comparisons, I honestly want to understand where AMD fits in the stack in terms of actual performance to dollars. I have had talks with companies wanting to build data centers outside of US and find it hard to source anything Nvidia in sufficient capacity and scale.\nIf AMD is competitive performance per watt and roughly reliable in terms of software support which is what most folks outside of US prioritize above all else, since outside of China and US electricity tends to at a relative premium.
Maybe if they make smaller data centers viable at the right price, AMD could be part of the stack outside of US where ever Nvidia is more limited in supply. Though I have genuinely no idea what sourcing an AMD GPU looks like.
I have never seen a company use AMD outside of wafer and a couple others mostly in US.
Genuinely intriguing or maybe not really (could be this stuff is common knowledge) and I am just stuck in my Nvidia bubble here.\n— minraws\n
Zuckerberg 'Admits' Meta's Layoffs Were Ineffective
\n\n\n\n236 points | 217 comments
\n\n\n\nI wonder when (if ever) the companies realize that demoralizing your workforce (and destroying that sector of the job market) doesn't have only advantages.\nI know plenty of people that reacted with the desired fear, putting in long hours to avoid layoffs, willingness to accept lower pay because the job market sucks, etc. - but I think there are also plenty of the the mythical 10x engineers that just checked out, stopped being 10x engineers, and are just collecting their paychecks and waiting for the layoff now. And I'm not sure you can \"get them back\", ever.
At least some companies reacted to this with more top-down management, stricter metrics etc. which kills motivation further and leads to metric optimization. Tell a good, smart, motivated engineer that you want more AI usage, and he's going to maybe start using some AI where it makes sense, but mostly ignore the metric while trying to do useful work. Demotivate the same engineer and make clear that his paycheck depends on metrics, and he'll give you what you're asking for, except https://github.com/dtnewman/burn-baby-burn is probably not what you _wanted_...\n— tgsovlerkhgsel\n
Virginia bans sale of geolocation data
\n\n\n\n328 points | 59 comments
\n\n\n\nGiven the actual informed and uncoerced choice, people say no to this kind of collection and especially its sale or use for any purpose other than the explicit service they thought they were allowing it for (navigation, setting the time, etc etc). This is true for basically all information collected. I'm glad to see there is some minor protection language being included but it needs to have real teeth and get to the point. If you collect data from me under false pretenses or using coercive methods (you can't use the thing you just paid a lot of money for if you say no) you will not only be fined but criminally prosecuted.\n— jmward01\n\n
Spain Orders Blacklist of Palantir from Public and Private Companies
\n\n\n\n563 points | 185 comments
\n\n\n\nSpain is really going in the right direction, I wonder why no one countries inspire from what they are doing\n— milanito1985\n\n
PeerTube is a free, decentralized and federated video platform
\n\n\n\n487 points | 218 comments
\n\n\n\nAs a professional YouTuber, the main issue I instantly see with this is the lack of monetization.\nI think people who don't make videos for a living severely underestimate how expensive it is to produce high-quality videos people want to watch. This isn't like writing a tweet or even posting a picture on Instagram. Even a decent 20-minute video can easily take 40 man-hours of high-skilled labor.
I have a pretty small channel (~100K subscribers) with no employees and relatively low upkeep costs (a few hundred dollars a month), and even I could not make this work if I didn't get at least $500-$1,000 per video on average, since it just takes so much time and money.
Most channels with more than a million subscribers are likely founders working 60-80 hour weeks with multiple full-time employees supporting them. You cannot do that in the hopes of viewers donating $5 here and there.
And yes, there are people who make content for free - most of them fail to hit a hundred views per video. And the difference between a million views and a hundred is 10,000x. You cannot create a platform without big users.
I think any real competitor to YouTube nowadays would have to be backed by a big corporation that can pay big creators million-dollar deals to make the switch. Otherwise it's just dead in the water.\n— djaro\n
360 points | 136 comments
\n\n\n\nNo idea why Docker is still so much more popular than Podman. Podman is obviously the better implementation.\nThe new network stuff is a welcome improvement.\n— SwellJoe\n
The Free Market Lie: Why Switzerland Has 25 Gbit Internet and America Doesn't
\n\n\n\n356 points | 224 comments
\n\n\n\nThis article came up before. It's heavy on the clickbait, if you couldn't guess from the title.\nSome important points that they leave out:
- 25G internet isn't available everywhere in Switzerland. It's just the fastest tier available in some locations.
- The United States is 85 times larger than Switzerland. The entire country of Switzerland is the size of a small US state. Covering the US with broadband is much harder than Switzerland.
- 25G internet is also available in some locations in the United States.
- As another commenter discovered, the average speed test results of US and Swiss internet connections are pretty similar. The average Swiss person isn't connected to the internet faster than the average United States person.\n— Aurornis\n
How to ask for help from people who don't know you
\n\n\n\n368 points | 58 comments
\n\n\n\nThis was an excellently written, salient post with some good tidbits.\nI've learned some of these lessons the hard way. I'll add a few. Proof of work is important, but it's not about the magnitude of energy you spend. I went through two iterations of reaching out to my college network. The first time I put so much time into handwriting notes and trying to provide my relatable background. 100 notes, not a single response.
The second time I sent emails that were a few sentences. I had a much clearer ask and devoted the effort into fitting my questions into the email. I wanted a conversation really, but I also tried to communicate what I planned to ask.
15% response rate and invaluable conversations. Less overall \"work\".
Secondly, and relatedly, don't ever waste someone's time. Don't ask for / accept a meeting if you don't have some semblance of a clear ask. It's hard, especially in early stage business where you're trying to discover what you don't know. But you can try to lay out your tier 1 \"here's what I think\", \"here are the follow ups\".
I sensed once that I had irritated someone by lacking the agenda. Another time I took a mutual connection up on an intro where I didn't know what I really needed. I regret both of these.
Thirdly try to pay it forward. It won't always come back around, but you can feel more comfortable asking for help and more cognizant of what a helper (so to speak) is thinking\n— jackconsidine\n
Since Linux 6.9, LUKS suspend stopped wiping disk-encryption keys from memory
\n\n\n\n383 points | 184 comments
\n\n\n\nWhile it is certainly an interesting bug, I kinda feel that the title is click bait? Because this `cryptsetup luksSuspend` from what I understood is not really officially supported but an extension done in Debian, so if anything this regression only affected Debian? I am not sure if you can blame the kernel for something that is not supported or even widely tested.\nI still find this impressive, and it is nice that we now have a test (NixOSTests BTW are awesome, I agree with OP) to avoid this regression from coming back. But from the title it seems to be a widespread issue, not something that affects only one Distro.\n— kokada\n
The Egg Bandits Made a Thousand Times the Fine They Just Paid for Price Fixing
\n\n\n\n427 points | 206 comments
\n\n\n\nI've known about Cal-Maine's scammy practices for years. It's difficult to know which brands are associated with a group like Cal-Maine, but I did my best at one point to get a list which I still check before buying eggs.\nThe brands I've been avoiding are:
\nOther Cal-Maine associated brands:\n Ralston-Purina,\n Pilgram's Pride, \n Adams Foods,\n Dairy Fresh Products,\n Foodonics Int'l,4Grain Cage Free\n Egglands Best\n Farmhouse Eggs\n Land O'Lakes\n Fassio Egg Farms\n Southwest Specialty Eggs LLC\n Rocky Mountain Eggs\n Meadowcreek Foods\n Specialty Eggs LLC\n Red River Valley Egg Farm\n ProEgg Inc. \n Dixie Egg Co. \n American Egg Products LLC\n Texas Egg Products LLC\n Wharton County foods\n Benton County Foods\n Sunups\n Sunny Meadow\n Mahard Egg Farm\n Maine Egg Farms\nVersova is new to me, but I'll try to avoid them too. It looks like their brands include:
\n— autoexec\nCenter Fresh Group\n Centrum Valley Farms\n Iowa Cage FRee\n Hawkeye pride\n Ovation Farms\n Trillium Farms\n Willamette Egg Farms
419 points | 207 comments
\n\n\n\nI teach a free software development course to my undergrad students. It's really exciting to stumble upon one of the work they did for my class in the wild (it's the first listed bug fix — which is the last of the three pull requests this student got merged in Immich during my course). I feel so proud! :)\n— p4bl0\n\n
AI can't be listed as inventor on patent applications, Japan's top court rules
\n\n\n\n352 points | 186 comments
\n\n\n\nThe book “against intellectual monopoly” has shaped a lot of my thinking on this topic - economists have looked at the various occasions in which patents were introduced into an industry (or extended in scope), and there is no evidence they actually improve innovation/efficiency/outcomes (including the pharma industry!). I was quite surprised as my whole life, it was sold to me as an incentive-boosting measure which in turn would lead to said outcomes.\nWith that lens, I welcome gradually phasing this stuff out, especially as we navigate into the unknown game-theory landscape AI-as-inventors brings.\n— alzamos\n
For first time, a cell built from scratch grows and divides
\nhttps://biotic.org/research/spudcell/\n\n\n261 points | 73 comments
\n\n\n\nScience News has a more balanced take, with additional quotes from peers.\n> Some have also grumbled about Adamala’s efforts to draw attention to the work, which she says was rejected by Cell after one reviewer said SpudCells were not real biology. She then sent the 190-page manuscript to journalists, under embargo, even before she had uploaded it to the preprint server bioRxiv, where her colleagues could read and assess it. She says her group will submit it to a new journal soon. “It’s an unusual way of doing things,” says Kerstin Göpfrich, a synthetic biologist at Heidelberg University.
https://www.science.org/content/article/lab-created-spudcell...\n— merksittich\n
Physical disc production ending in Jan 2028 for new games on PlayStation
\n\n\n\n563 points | 593 comments
\n\n\n\nGreat buildup on Sony's side to gain trust in this move in the gaming community ahead of this announcement when just this week they again pulled hundreds of \"purchased\" movies from customer's libraries without refund, reminding everyone that digital content is rented, not owned.\nhttps://arstechnica.com/gadgets/2026/06/sony-erases-digital-...\n— fcsp\n
Most arguments are about ego, not ideas
\n\n\n\n521 points | 405 comments
\n\n\n\nHere's a simple idea: You can't reason someone out of a position they didn't reason themselves into.\nAnd three interpretations to consider:
0: The default: That person is irrationally attached to being wrong. Best to walk away, argumentation will be futile, and I have a life to lead.
1: Whoa! Sometimes that person is me.
2: If they didn't reason themselves into it, how did they get into it? What if their position represents their values, not some perfectly architected strategy for maximizing some hypothetical measure of rightness? In that case, if I wish to discuss it with them, I should be talking about their values and my values and where they intersect, rather than arguing right and wrong?
I have personally found all three of the above useful at one point or anther.\n— a4isms\n
Sony Deletes 551 Movies PlayStation Owners Paid For
\n\n\n\n504 points | 225 comments
\n\n\n\nWe need some kind of modern equivalent to the old proposed Digital Media Consumer's Rights Act but which protects people's rights to digital media they buy. These should never be sold and then taken away with no compensation like this. We need a law that forces companies to treat digital files the same as a physical purchase. They can't take it away and have to allow people to resell and loan out as well. And in cases of online games where you can buy something, and then later they can ban you which deprives you of being able to use what you bought, that should come with requirements that the company must provide full compensation of the purchase price. It should also ban EULA's and TOS from defining these things as only licenses even though they are structured as a purchase in a store.\nI know it'll never happen with the people we have in government these days, and the anti-consumer organizations, like the ESA, that are out there now claiming things like running private servers for Minecraft is illegal and piracy. (Yes, they really said that. Despite the fact that Minecraft has always provided the server and allowed this for 15+ years)\n— nullsmack\n
Asahi Linux 7.1 Progress Report
\n\n\n\n438 points | 141 comments
\n\n\n\n> The defacto industry standard for audio ICs is I²S, an I²C-based bus optimised for audio data.\nNit: I²S has nothing to do with I²C.
(Most I²S chips also have an I²C interface since I²S only carries raw audio data, no sideband like volume control or clock configuration. But that's a separate interface and can also be SPI rather than I²C. In fact, SPI is more closely related to I²S than I²C is.)\n— eqvinox\n
Nintendo has raised its employees base salary by 10%
\n\n\n\n369 points | 192 comments
\n\n\n\nI recently heard that a trip to Popeye's for a family of 3 recently cost $68 in Florida.\nIn Japan, there's a big issue when a snack raises its price 2 cents (3 yen - source: https://finance.yahoo.com/news/japanese-snack-company-apolog...)
The country for better or worse seems to be frozen in time - salaries have not caught up with the heady levels of SV (or even Europe) but neither have rents or prices for common goods.
This is not a judgment either way - but it does make Japanese exports a significantly more lucrative business - if only they could figure out how to sell more of their stuff abroad!\n— epsteingpt\n
Box3D, an open source 3D physics engine
\n\n\n\n395 points | 86 comments
\n\n\n\nWhenever I see Box2D mentioned (the library by the same author as Box3D, obviously), I think back to this story from many years ago\nhttps://kotaku.com/this-guy-created-angry-birds-physics-and-...\n— gregsadetsky\n
A new Android malware from Google
\n\n\n\n471 points | 209 comments
\n\n\n\nI think the most fun part with Google is that if some wayward algorithm decides it doesn’t like you, along with nuking your app and developer account it will probably nuke your 20 year old gmail, your kids Google Drive accounts, your wife’s YouTube premium, the Adsense account of some company you worked for in 2008, and disable your Nest cameras.\nAnd you’ll never reach a human to sort it out.\n— transcriptase\n
345 points | 277 comments
\n\n\n\nI'm somewhat surprised that this is not open source (from what I can tell). Compare to Mimo Code https://github.com/XiaomiMiMo/MiMo-Code (which is a CLI, while this is a desktop app).\n— seizethecheese\n\n
260 points | 89 comments
\n\n\n\nMan what a showcase for Opus this is.\nDon't get me wrong, this sort of thing is a valuable exercise and we are better off with better encoders for these older codecs. But look at the numbers for Opus on this benchmark. It simply blows all the AAC encoders out of the water even at 64 kbps.\n— cogman10\n
Claude Code Is Steganographically Marking Requests
\n\n\n\n316 points | 101 comments
\n\n\n\nValue judgment aside: I am a bit surprised at how sloppily they did this. I think they could've achieved the same effect while decreasing the odds of detection via reverse engineering like this.\n(This field is known as \"underhanded code\", coined by the Underhanded C contest: https://www.underhanded-c.org. It's a little-known \"art\"; little-known for probably self-explanatory reasons. There are much cleverer ways of achieving effects like this. One obviously being you can move more out of the client and into the server, but the other being you can write plausibly deniable client code in a much more benign-seeming way than this. Some of what they added can only be done on the client, but I think more could've been moved.)
It's possible they knew the JS bundle gets so heavily scrutinized that it'd eventually get spotted and reported on regardless so they didn't bother doing something more subtle and duplicitous. But still a bit surprising.\n— meowface\n
810 points | 457 comments
\n\n\n\nThe cost per task chart is telling me that I should _never_ use Sonnet 5 above medium effort level - Opus always performs better for a given cost. So I guess the takeaway is that if Sonnet 5 medium isn't good enough for you, switch models, not effort levels.\n— doctoboggan\n\n
European digital ID wallets rely on safety services of Google and Apple
\n\n\n\n589 points | 248 comments
\n\n\n\nThe EU reference for wallets strictly required google play services\nhttps://github.com/eu-digital-identity-wallet/eudi-app-andro...\nSo Italy's IO app https://github.com/pagopa/io-app (wallet, documents, age verification) continuously refuses the users' request for GrapheneOS support and requires google.
Nothing will change until the lawsuits start coming in.
The only hope is the motorola/grapheneOS collaboration and consumer associations, that might sue for anticompetitive behavior.
Make noise on any channel for the apps that require play services, it will help in the future if the lawsuits start, since it will show user support for the initiative.\n— Luker88\n
Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5
\n\n\n\n575 points | 302 comments
\n\n\n\nThe damage is done. You cannot build a business critical function on top of American SOTA frontier model. Especially not with the current crew in charge.\nNow whether AI tech is in the same league as say Nuclear tech and therefore by any reasonable standard should be regulated is a different question.
We hit the slippery slope on a random day in June 2026 and there is no putting the genie back in the bottle. Any exec or manager that puts load bearing weight on top of Anthropic/OpenAI/Google/AmericanCorp frontier model deserves the stress.\n— drevil-v2\n
The US ambassador had Belgian police stop our reporting
\n\n\n\n624 points | 258 comments
\n\n\n\nFor additional context, tensions are already high surrounding the US ambassador after he directly insulted multiple Belgian politicians and also attempted to interfere with local criminal judicial proceedings.\n— elil17\n\n
335 points | 111 comments
\n\n\n\nI built one of the connected tools included in this launch (the Biomni HPC [1]), and I have spent an inordinate amount of my life working on this problem. (I also worked at Anthropic, but not on this product.)\nAs other comments have pointed out, this is for data science – but it's capable of more than making plots and writing papers [2]. It has integrations with many databases and computational tools, including a researcher's institutional cluster.
That alone is valuable. I founded a startup after struggling with this problem at a bio startup; integrating these tools and databases is hard and time consuming. If the only outcome of this product is that great APIs are built for LLMs, it will be a massive positive impact. Many databases used in computational genomics are still only accessible through FTP!
LLMs are particularly good at navigating these tools and databases. It's often very specialized, but straightforward, work that benefits from in-context skills. Seeing an early glimpse of my former customers – bioinformaticians – using LLMs to solve this problem is what led me to join Anthropic in 2024.
Also, this pattern isn't fundamentally constrained to data science: you can also integrate with a wet lab or a CRO for some kinds of science. This is what I'm spending my time on now.
This type of science doesn't solve everything, but it's useful in some niches. For example, progress on many rare diseases is bottlenecked by researcher attention rather than a fundamental breakthrough.
[1] https://x.com/phylo_bio/article/2029233694775624096
[2] In comparison, OpenAI's science product – Prism – was effectively a LaTeX editor they acquired with Crixet.\n— lebovic\n
The labor share of income in the US is at its lowest post-war level
\n\n\n\n295 points | 243 comments
\n\n\n\nPeople are misreading the conclusion - the Covid related drop is normal and matches previous episodes, but the massive overall drop since 2000 is not.\nThe situation on the ground is unchanged - the amount of labor being generated per person has not really changed, but the overall pie has grown massively around us.\n— legitster\n
County with 37 Data Centers Asks Schools to 'Conserve Electricity'
\n\n\n\n387 points | 176 comments
\n\n\n\nThis is in Virgina, which passed the Virginia Clean Economy Act in 2020. This mandated that Dominion (the power company) transition to 100% renewable energy by 2045. Personally, I think this is a good thing in the long run, but in the short run, it means that Dominion has had to invest a lot in building out renewable projects that haven't come online yet.\nLawrence Berkeley National Lab recently did an analysis on electricity prices in the US [1] and found that most of the rate increase in Virginia was attributable to the VCEA, and that load growth had a mitigating effect on price increases.
If you look at the overall report (not just Virginia), the places where electricty costs are rising the fastest are generally not the same places where lots of new datacenters are being built. It's easy to blame datacenters, but there are many factors at play here.
[1] https://emp.lbl.gov/publications/factors-influencing-recent-...\n— jonas21\n
286 points | 106 comments
\n\n\n\nThe first example of generating home interiors fills me with indescribable hatred. Recently real estate agents have taken to running every dilapidated unsellable apartment through these AI filters, and you have to scroll through a dozen of these Ikea-chic images of what the apartment presumably could look like, before you are allowed to see the horrors they are trying to peddle at insane prices.\n— torginus\n\n
We Are the Last People Who Know How It Works
\n\n\n\n284 points | 232 comments
\n\n\n\nThe issue with this is that we don't know how it works. Generally speaking, we know how the level of abstraction that we were born with works. We might have some understanding of one or two previous levels, but that decreases the farther down you go. We might understand the next level, and some of the next after that, but eventually people will be making things that we don't have the context to understand without having to unlearn a lot of what we know now.\nI'm old enough to see this process in action; I used to be young and in possession of esoteric knowledge that made me infinitely in demand and now most of the things that young people have esoteric knowledge about is things that I don't particularly care about, and I'm left with a lot of finely honed skills to solve problems that have mostly been abstracted away.\n— CommieBobDole\n
Qwen 3.6 27B is the sweet spot for local development
\n\n\n\n546 points | 476 comments
\n\n\n\nI love my MacBook Pro M5 128GB RAM and I love qwen3.6.\nBUT DO NOT buy this MacBook if you plan on doing serious coding using local LLMs with it. The reason is simple: your fingers will burn and your head will explode from the noise.
Running any kind of sophisticated job on the very laptop you are using is just not viable. Sure you can use it in clamshell mode, but forget touching it while working with AI coding or agents.
If you want to run Qwen3.6 27B / 35B at its best, get a MacMini M4 with 64GB of RAM and put it in the basement - or at least a few meters from your desk. Connect to it over LAN or Tailscale. The MacMini will also cost you almost 1/3 of the MacBook Pro.
Thank me later.\n— iagooar\n
The CEO of Mullvad is the main financer of the Swedish Örebro party
\n\n\n\n522 points | 1165 comments
\n\n\n\nThere have been multiple posts about this and hundreds of comments, so there is clearly appetite to discuss it, although none of the submitted links have been very detailed.\nI've merged the other threads into this one, so you'll see some anachronistic timestamps below.\n— dang\n
US Supreme Court rules geofence warrants require constitutional protections
\n\n\n\n394 points | 182 comments
\n\n\n\nFrom https://www.scotusblog.com/2026/06/court-rules-that-law-enfo...\nAdditional details:
> The information that Google provided to law enforcement officials came in three tranches. First, Google gave law enforcement officials a list of the 19 accounts (but without the names attached to those accounts) linked to devices that were within 150 meters of the bank during the 30 minutes before and after the robbery. Second, based on that list of 19 accounts, the government asked for additional information about nine accounts that were in the area during a two-hour period. At the third step, a detective asked for, and received, the names and information associated with three accounts – one of which was Chatrie’s.
> Relying on the location data, law enforcement officials obtained a warrant to search two residences linked to Chatrie, where they found almost $100,000 of the stolen cash, a gun, and demand notes.
> Prosecutors charged Chatrie with bank robbery. He asked the trial judge to bar prosecutors from using the evidence obtained as a result of the geofence warrant at his trial, arguing that the warrant violated the Fourth Amendment.
> A federal district judge agreed that the warrant in Chatrie’s case did not have the kind of probable cause and specificity that the Fourth Amendment requires. However, she nonetheless allowed the prosecutors to use the evidence, reasoning that even if there had been a violation of the Fourth Amendment, law enforcement officials had acted in good faith.
Link to ruling:
https://www.supremecourt.gov/opinions/25pdf/25-112_0am4.pdf\n— js2\n
.self: A new top-level domain designed to support self-hosting
\n\n\n\n246 points | 154 comments
\n\n\n\nRemember when the .tk TLD became free 20 years ago ? Every hobbyist took one, then scammers followed, then Facebook and antiviruses started blocking it.\nI remember publishing a website for a class on my .tk domain, the teacher couldn't open it and I almost got a failing grade because of it.\n— goldenarm\n
341 points | 217 comments
\n\n\n\nWe have a bright future full of endless \"space-junk\". As the price to orbit drops, people will inevitably send up more and more satellites that have questionable value. In 100 years will the sky at night just be a massive grid of dots moving across the sky?\nWho will create the first advertisement in space using satellites as pixels to create their company logo? Maybe they can add some color and animations for kicks.
Edit: Another note on space junk is the effect on our atmosphere with all the \"burning-up\" of various materials. Apparently they don't just completely vaporize, but instead leave behind micro particles that float around for a long time. People are studying this and hopefully raising appropriate alarms (Making the case for wood satellites).\n— proee\n
Samsung, SK Hynix, Micron Sued in US over Memory Price Fixing
\n\n\n\n347 points | 160 comments
\n\n\n\nThis was attempted before in 2022 but fell apart because plaintiff couldn't show an agreement took place https://cdn.ca9.uscourts.gov/datastore/opinions/2022/03/07/2...\n— cbg0\n\n
European ISPs Want Rightsholders Held Accountable for Overblocking Damage
\n\n\n\n333 points | 87 comments
\n\n\n\nIt's not just Europe. DMCA takedowns in the US: no liability for taking down innocent content.\nReally, it comes down to this: censorship is bad. Always.
If someone violates the law, get a court judgement. With the judgement in hand, take down that specific material.
Too much work? Tough...\n— bradley13\n
Instagram is incorporating users' photos in ads for Meta Glasses
\n\n\n\n319 points | 137 comments
\n\n\n\nDidn't Facebook do this years and years ago?\nYes, 2013: https://mashable.com/archive/facebook-ads-photo#ggcKnNfAUaqy
> According to Facebook's Statement of Rights and Responsibilities:
> You give us permission to use your name, profile picture, content, and information in connection with commercial, sponsored, or related content (such as a brand you like) served or enhanced by us. This means, for example, that you permit a business or other entity to pay us to display your name and/or profile picture with your content or information, without any compensation to you. If you have selected a specific audience for your content or information, we will respect your choice when we use it.
So it's not new. If you don't want this, delete your facebook account: https://www.facebook.com/privacy/dialog/delete-your-informat...\n— Cthulhu_\n
A native graphical shell for SSH
\n\n\n\n222 points | 105 comments
\n\n\n\nThis appears to me like a solution in search of a problem, like many others before it...the quote below seems relevant to this effort.\n\"Those who do not understand Unix are condemned to reinvent it, poorly.\" ~Henry Spencer\n— hatradiowigwam\n
201 points | 231 comments
\n\n\n\n> Tidal will accept AI-generated music.\n> Tidal will hold AI-generated music to a higher standard of content integrity. We will not tolerate AI-generated music that exploits an individual’s or group’s music, name or likeness, deceives listeners, or diminishes the quality of our service.
I think this is a very reasonable approach, and probably also the best way to treat AI-powered copyright infringement as a whole. Just like we don't penalize artists for consuming content unless they produce actually infringing content, we should set the same focus for AI systems.
> Starting today, AI-generated music will not be monetizable. We are only in the beginning of the era of AI-generated music.
Don't really agree that this follows from the stated principle here (\"... ensuring royalties go to original works produced, written and performed by people\"), but will definitely help with spam etc.\n— fxwin\n
GLM 5.2 beats Claude in our benchmarks
\n\n\n\n338 points | 157 comments
\n\n\n\nI have taken another look on these open models after the fiasco of Fable and GPT 5.6 this weekend and... GLM-5.2 truly is a good workhorse model for daily programming. I consider myself a heavy user of LLMs and a seasoned developer. A typical session for me with GPT is usually over a hundred dollars...\nThis weekend I programmed a matrix bot with encryption and a Rust agent with some tools. Because I need one and OpenClaw just felt... not what I wanted. Two days later and 20 dollars poorer I have what I need: a multimodal agent written in rust that has access to my homelab.
Nothing felt off with GLM. It did what I wanted, was fast, had a decent not very annoying personality and was much cheaper than Opus or GPT.
I used it unquantized through Fireworks, but there are multiple other providers too.\n— pimeys\n
EU to legislate about Chat Control behind closed doors
\n\n\n\n259 points | 154 comments
\n\n\n\nThe global push to kill privacy makes me sad.\nFeels like I grew up in a golden age and subsequent generations won't care because they never knew a different world\n— Havoc\n
The KIDS Act would require age checks to get online
\n\n\n\n273 points | 241 comments
\n\n\n\nWould this website (HN) be a \"covered platform\" according to the bill?\nAs far as I can tell, the answer is no, because it doesn't do what's described in Section 201 (E):
\"Uses the personal information of the user to advertise, market, or make content recommendations.\"
Neither does, for example, my bank's website, or someone's personal blog, or many other discussion sites like this one. So from what I can see, while the set of covered platforms is certainly not negligible, it's still a lot smaller than \"basically every website on the Internet that anyone cares about\". So the title of the EFF article is overstating the case; the thing the bill would require age checks for (in effect, if not by the explicit language of the bill) is not \"get online\" but something more like \"get on social media\".\n— pdonis\n
Age verification is just a precursor to automated attribution of speech
\n\n\n\n285 points | 154 comments
\n\n\n\nIf we taught systems thinking in schools things like internet age verification would never get past being an idea on the back of a napkin. People struggle to consider the second-, third-, and nth-order effects of anything so asking them to consider what else might happen if we bring in laws and technical mechanisms to 'protect the children' is unfortunately too a big leap for a lot of them. Most people are bad at spotting causal links between parts of a system, and people who are good at it exploit that.\n— onion2k\n\n
HackerRank open sourced its ATS. My resume scored 90/100. Oh wait 74. No – 88
\n\n\n\n235 points | 54 comments
\n\n\n\nAn alarming number of people don't understand that LLMs work via purely stochastic processes, so I'm happy to see in-depth pieces like this. I'm looking for a job and maybe this is why it's so hard to get a callback these days: resumes are just dumped in some LLM black hole and no one really knows how it works. The author says:\n> temperature 0.1 — low, supposedly nudging the model toward deterministic outputs
This is not correct (and is briefly touched on later in the piece when he sets temperature to 0), temperature is not some kind of \"deterministic\" switch, but rather it affects the sampling distribution (which becomes more \"spiky\"—but is still very much a distribution).\n— dvt\n
I used Claude Code to get a second opinion on my MRI
\n\n\n\n308 points | 418 comments
\n\n\n\nI'm a radiologist but can't really weigh in without seeing the full 3D MRI dataset. Regarding this point:\n> They performed shockwave therapy on my shoulder even though a recent clinical practice guideline says clinicians should not use or recommend shockwave therapy for rotator-cuff tendinopathy without calcification; I was told during ultrasound that there was no calcification.
Ultrasound isn't a great way to assess for calcification. It'll find large calcification but easily miss small ones. Plain radiograph would be more helpful, but the MRI may have revealed it as well. Either way, shockwave therapy isn't harmful in the absence of calcification--it's just not helpful.
Edit: when a radiology report says something isn't present, there's always an implicit caveat that the finding isn't present within the context of the modality and images obtained. So an ultrasound report can state there are no calcifications while a plain radiograph can report the presence of calcifications without being inconsistent. Obviously very confusing to patients and people unfamiliar with medical jargon, but clarifying this in reports would make them sound even more qualified, \"hedgey\", and annoying to read than they already are.\n— sxg\n
Professor denounces mass AI fraud on an exam at Brown
\n\n\n\n367 points | 486 comments
\n\n\n\nIn the AI era, tests are going to have to be in person and hand written.\nI have written an article on how I have adjusted my classes to the situation:
https://htmx.org/essays/universities-and-ai/
Ironically, I think the AI era may make university degrees a better signal of the intellectual abilities of students due to the presence of pre-computer infrastructure like large lecture halls, industrial-scale copiers, etc.\n— recursivedoubts\n
243 points | 70 comments
\n\n\n\nTo clarify because this is confusing: The AirPods work as regular old BlueTooth earbuds on other devices already. This is an implementation of some of the extra features and interfaces that are integrated into Apple products.\n— Aurornis\n\n
5k menus from the New York Public Library’s Buttolph Collection (1880-1920)
\nhttps://pudding.cool/2026/06/menu-collection/\n\n\n311 points | 83 comments
\n\n\n\nAnyone interested in this might also like the tidbit that in Germany, they used to, and still count beer consumed as pencil strikes on the beer paper mat. Altering the number by the guest is legally considered forgery and the disappearance of the beer mat is also punishable by law.\nhttps://de.wikipedia.org/wiki/Bierdeckel#Urkundencharakter (in German, English wiki doesn't have this info)\n— ricardobayes\n
Flock cameras track more than your license plate, and they're spreading fast
\n\n\n\n206 points | 114 comments
\n\n\n\nIt's legal for any random citizen to build one of these surveillance networks, right?\n— microgpt\n\n
Anonymous GitHub account mass-dropping undisclosed 0-days
\n\n\n\n616 points | 241 comments
\n\n\n\nI took a look at the Ghidra ones (because I use Ghidra), and I'm unimpressed: https://github.com/bikini/exploitarium/blob/main/ghidra-12.1...\nThe first requires being able to overwrite binaries in the Swift tool directory. Yes, if you overwrite binaries executed by ghidra, you can trigger code execution. This is not a surprise.
The second, idk, I'm not familiar with TraceRMI (but it's probably worth noting that \"RMI\" stands for Remote Method Invocation).
The third is not a vulnerability in the slightest, they just demonstrate that native 7zip parsing code is reachable. Maybe there is a bug in the 7zip parser, but without that it's meaningless.\n— Retr0id\n
289 points | 63 comments
\n\n\n\nIf you play the original and then OpenRA you will be amazed how well OpenRA is balanced.\nAs an example, while in the original game using allied artillery against soviet tesla coils was a dead sentence in OpenRA is great to be able to fire well beyond its range forcing you to come out of the base to defend it.
They also added a ton of features which make the game truly enjoyable and fun to play.
Well done OpenRA team!\n— liendolucas\n
Zuckerberg's Increasingly Bizarre War on Whistleblowers
\n\n\n\n213 points | 81 comments
\n\n\n\nIt's not increasingly bizarre, really, if you just allow for the possibility of one thing:\nThere's something else worse that they know could be in such a book, but isn't yet, and it is so bad that it is worth doing this.
Perhaps they know that Wynn-Williams could have put it in the book and didn't. Perhaps they know that someone else — someone else British, say? — could write such things in a book and so far hasn't.
Once you assume their motivation is grounded in real fear, it gets easier to see why this isn't bizarre at all; it's inevitable.\n— dofm\n
307 points | 105 comments
\n\n\n\nI glanced, and I found this handbook shallow and - in some areas - even bad advice.\nE.g. If I ever see a monetary value stored in something else than integers I'm going to run away screaming (thank you Rust decimals represented as JSON floats). It's always integers unless you have a VERY good reason to do otherwise (though exported view can be in anything, even in weird bitcoded formats).
FX exchange. Resolution of FX isn't a point-in-time thing, things like buyer rate-in-time, seller rate-in-time, agreement, agreement tolerance, agreed upon resolution timestamp come in the effect.
Immutability - that's why you want to have event sourcing everywhere that touches money:
\nThough in the end Fintech != Fintech. I worked at Fintech where money was treated like a baggage, and in other where money was a central point of everything.\n— xlii\n# Resolved stream\n A -> B -> E\n\n # Actual stream\n A0 -> Edit(A0, A) -> B -> C -> D -> Rollback(B) -> E\n\n
The case for physical media ownership
\n\n\n\n344 points | 227 comments
\n\n\n\nI agree with the sentiment implied by the author, but I would reword it slightly. If you don't have the freedom to share something, you don't own it.\nI disagree with the interpretation that it needs to be held physically. Digital ownership is still ownership. I go out of my way to find music on Bandcamp, games on GOG, and rip movies myself using MakeMKV.
I wish I could encourage people to continue embracing physical media but most people value convenience over true ownership. And most companies value market capture and \"security\" over user rights. In crypto the sentiment of \"not your keys, not your wallet\" is held a core truth, yet people use 2factor authentication and Passkeys without respecting the same truth. I am not arguing against the use of 2factor, but at the same time certain accounts can not be logged into freely without push notifications in Duo or Microsoft. I still don't see a universal ability to export Passkeys, and I believe that's by design.
I hope laws catch up to modern technology in terms of digital goods. I can't imagine companies choosing to open up their walled gardens otherwise.\n— knaik94\n
IP Crawl: Living atlas of open webcams discovered on the public internet
\n\n\n\n201 points | 110 comments
\n\n\n\nPerhaps someone could have some fun with this...\nFeeding faked looped security camera footage is a classic plot device in many films, and could make some good comedy!\n— bouncycastle\n
Streaming services' obnoxiously loud ads become illegal on July 1 in California
\n\n\n\n241 points | 72 comments
\n\n\n\n> [...] opposed the bill. The groups argued that “many” streaming services were already trying to manage the “loudness of advertisements that come from server-side ad insertion that may be inconsistent with the loudness of the programs,”\nWell, stop \"trying\" and fix it already. It's your own damn system.\n— kube-system\n
Suspicious Discontinuities (2020)
\n\n\n\n199 points | 53 comments
\n\n\n\nThe marathon one has a simple - and fun - explanation: it’s great to run with people!\nMany (most?) marathons have pace runners who run the course, hewing to each 30 minute and 15 minute finish. So there’ll be a 3:30:00 pacer, a 3:45:00 pacer and so on. Your local pacer might even have their pace on their shirt and may even have a flag so they’re super hard to miss as you’re running.
One friend of mine runs with a speaker, to play music and keep everyone’s spirits up.
By the end of the race they’d acquired a small army of marathoners! You could see their smiles for miles, and when they all finished together they had a huge party in the recovery area with the speaker :)
I’m surprised the marathon time discontinuity isn’t bigger :)\n— cadamsdotcom\n
Asian AI startups launch Mythos-like models
\n\n\n\n211 points | 161 comments
\n\n\n\nI tried the Fugu models with some real world tales in C# and unity using mcp and open code. I exhausted the $20 plan 5 hour window in one prompt to review my theme system and plan some color changes. So I upgraded to the $100 to see the implementation and result. Well the result was worse than Opus, incredibly slow, and I ended up exhausting the new 5 hour window and have used 35% of the weekly now and it hardly created something opus was able to do at a fraction of the time and cost.\nDo what you wish with this info, but it seems to be a complete waste of $$.\n— cdurth\n
Marfa Public Radio Puts You to Sleep
\n\n\n\n228 points | 56 comments
\n\n\n\nOne trick that makes me sleepy really fast: After I close my eyes, I imagine someone throwing black paint on them. The first coat is kinda gray and has lots of blob and not fully black. Then another coat. And another. Each one gets darker until it's just pure black and I'm usually asleep by then.\nFor some reason, my brain follows it, and I fall asleep much faster. It works way better for me than box breathing or most other sleep tricks I've seen. Sharing in case someone else finds this useful.\n— superasn\n
U.S. government will decide who gets to use GPT-5.6
\nhttps://archive.ph/PCQQl\n\n\n744 points | 859 comments
\n\n\n\nAll: for comments on the technical side please go to the related thread:\nPreviewing GPT‑5.6 Sol: a next-generation model - https://news.ycombinator.com/item?id=48689028\n— dang\n
Previewing GPT‑5.6 Sol: a next-generation model
\nSystem card: https://deploymentsafety.openai.com/gpt-5-6-preview\n\n\n771 points | 475 comments
\n\n\n\nAll: for comments on the policy side please go to this related thread:\nU.S. government will decide who gets to use GPT-5.6 - https://news.ycombinator.com/item?id=48690101\n— dang\n
345 points | 61 comments
\n\n\n\nThat is very very funny, and oh so plausible.\nI enjoyed this bit a lot from the timeline
> Karen Oyelaran finds the payload by reading the source code with her eyes and files a second issue. The triage assistant closes it as “duplicate of #8814.” Issue #8814 is a feature request for dark mode. Karen reopens it. The assistant closes it. Karen reopens it. Karen’s GitHub account is rate-limited for “patterns consistent with automated behaviour.”
And this - the final sentence is a perfect indictment of the timeline we are in.
> Two AI review agents from competing vendors, both attached to a downstream pull request bumping foxhole-lz4, enter a disagreement loop over whether the package is malicious. After 340 comments and $41,255 in inference spend, Finance revokes both API keys; one vendor’s marketing team, cc’d on the cost anomaly alert, issues a press release citing “a 430% YoY increase in adversarial multi-agent security reasoning.” The stock opens up 6%.
I'm joining the goat farming waitlist ;-)\n— nickcw\n
U.S. allows Anthropic to release Mythos AI to ‘trusted’ US organizations
\nhttps://archive.md/ArXuFhttps://www.nbcnews.com/tech/tech-news/us-government-gives-a...\n\n\n
362 points | 360 comments
\n\n\n\nThis makes me sad since it implies that the best LLM I will ever be allowed to use is GPT 5.5 and Opus 4.8. Anything smarter than that is deemed too risky.\nSo much wasted potential.
And why would I pay Anthropic or OpenAI once consumer hardware gets powerful enough to run an open weight Chinese version of Opus 4.8? Even more so when mobile phones are able to run similar LLMs.
Their financial growth looks doomed. It looks like they will be heavily regulated just like the next missile factory. This is antagonist to VC led turbo growth startup regime.\n— bel8\n
We can still stop California's 3D printer surveillance scheme
\n\n\n\n345 points | 119 comments
\n\n\n\nMy kindergartner has a 3D printer.\nI got a call from the school principal. She said “another parent called and said your son 3D printed a gun and brought it to school”.
I looked at the print history. It was a tiny toy mandalorian figurine holding a blaster pistol in his hand.
I bought my son a bigger 3D printer and told him to stop playing with that boy.\n— gdiamos\n
Springer Nature has removed two studies by Max Planck
\n\n\n\n343 points | 168 comments
\n\n\n\n> […] the publisher posted a blank white page with the cryptic phrase, “This article has been withdrawn due to article violation.” Springer Nature is nevertheless still selling the empty PDF for $39.95.\ncompletely unsurprised, given the state of online papers publishing. if you don’t have an subscription or aren’t an organisation member, the fees are insane\n— bstsb\n
296 points | 15 comments
\n\n\n\nrare to see images, let alone with color in a Daring Fireball post\ngreat read
RIP Om\n— tosh\n
275 points | 151 comments
\n\n\n\nWanted to mention that Sailfish has a lot of closed-source components, especially UI-related, despite the overall marketing/\"vibe\" making it look very open. If anything, AOSP (Android) is more open than Sailfish. I don't think this has changed with Sailfish 5, see e.g.:\n- https://forum.sailfishos.org/t/sailfish-os-clarifying-claims...
- https://docs.sailfishos.org/Develop/Open_Source/\n— Tiberium\n
Ultrasound imaging of the brain
\n\n\n\n265 points | 111 comments
\n\n\n\nEven low-dose ultrasound (what they use on pregnant woman) results in ultrastructural changes in brains [0], specifically at the nodes of Ranvier (the gaps between myelin along axons). See also [1] for a review.\n[0] Ellisman MH, Palmer DE, André MP (1987), \"Diagnostic levels of ultrasound may disrupt myelination,\" Experimental Neurology 98:78–92 \nhttps://pubmed.ncbi.nlm.nih.gov/3308504/
[1] Quarato, C.M.I., Lacedonia, D., Salvemini, M., Tuccari, G., Mastrodonato, G., Villani, R., Fiore, L.A., Scioscia, G., Mirijello, A., Saponara, A. and Sperandeo, M., 2023. A review on biological effects of ultrasounds: key messages for clinicians. Diagnostics, 13(5), p.855. \nhttps://pmc.ncbi.nlm.nih.gov/articles/PMC10001275/\n— davi\n
PlayStation Is Deleting 551 Movies from Customers' Accounts
\n\n\n\n249 points | 137 comments
\n\n\n\nPiracy is justified especially when it comes to movies!\nIf I am buying a DVD, I own that copy regardless of the studio and the distributor being in legal trouble or not. If I \"buy\" or \"purchase\" something online, I expect the same thing.
I'm not always a fan of the EU over-regulating some things but I feel like they should start fining companies who want to re-define the meaning of the word purchase\n— thomasmarton\n
An entire Herculaneum scroll has been read for the first time
\nPreprint: https://scrollprize.org/pdf/main.pdfhttps://github.com/ScrollPrize/villa\n\n\n
883 points | 198 comments
\n\n\n\nLets reflect on Aristocreon, in about 200 BC, putting their thoughts down on a scroll. They would be aware that the scroll might be kept in a library for some time. Maybe they could have imagined it surviving for 300 years. But they never would have imagined that in 300 years a volcano might destroy the scroll, but in some way preserve it. And then that nearly two thousand years later future humans with machines made of materials unimaginable to Aristocreon, but related distantly to sand and lightning, would be able to read the scroll again and instantly transmit it to nearly the whole planet, a planet with many times more humans than existed in their time. (and speaking of 'planet', in Aristocreon's time, people had fairly recently been able to show that the world was spherical but much of it was still unknown).\nDo we have better imaginations? Can our sci-fi writers come up with something equivalent that is as dizzyingly far from what we know now, as now is from what Aristocreon knew?\n— codeulike\n
283 points | 29 comments
\n\n\n\nOh wow. What?! Just this morning I had an occasion to go thru his site/blog.\nStill can't believe it. 60 is too young.
I met Om finally in 2013-ish at one of his GigaOm events in the SF Bay Area. Before that, I had been a long time reader of his GigaOm blogs and other writings at Fast Company, Red Herring, Light Reading, and elsewhere, including his book Broadbandits. He was one of the few bloggers / reporters who wrote it as he saw it; his takes were often brutally honest and pointed. He called upon the excesses of various telecom execs during the dot-com and telecom bust of 2000-2001/2. His book Broadbandits is basically an invective of the go-go days of telecom companies' incestuous deals (now seen in the AI companies too).
I had a few more occasions to meet him at dinners around the Bay Area. He was always gracious, and listened intently to what people said. As a venture partner, he focused on the people (founder) and their stories much more on the businesses.
I had heard about his troubles with his heart (~age 40-ish), which made him turn his life around to focus on only a few things that brought him joy - writing, photography, travels.
He will be missed. RIP, Om.
---\n(Update: the book is Broadbandits (not Telecom Bandits, as I mistakenly wrote)\n— aanet\n
Apple raises prices of MacBooks, iPads
\nhttps://9to5mac.com/2026/06/25/apple-price-increases-mac-ipa...\n\n\n598 points | 854 comments
\n\n\n\nThese are the price changes mentioned in the article:\nMacs
\niPadsMacBook Neo: $699 (up from $599)\n 13-inch MacBook Air: $1,299 (up from $1,099)\n 15-inch MacBook Air: $1,499 (up from $1,299)\n M5 MacBook Pro: $1,999 (up from $1,699)\n M5 Pro MacBook Pro: $2,499 (up from $2,199)\n M5 Max MacBook Pro: $4,099 (up from $3,599)\n iMac: $1,499 (up from $1,299)\n M4 Max Mac Studio: $2,499 (up from $1,999)\n M3 Ultra Mac Studio: $5,299 (up from $3,999)\n\nMore products:iPad: $449 (up from $349)\n 11-inch iPad Air: $749 (up from $599)\n 13-inch iPad Air: $949 (up from $749)\n 11-inch iPad Pro: $1,199 (up from $999)\n 13-inch iPad Pro: $1,499 (up from $1,299)\n iPad mini: $599 (up from $499)\n\n— primaprashant\nApple TV 4K: $199 (up from $129)\n HomePod: $349 (up from $299)\n HomePod mini: $129 (up from $99)\n Vision Pro: $3,699 (up from $3,499)
The 'papers, please' era of the internet will decimate your privacy
\n\n\n\n259 points | 116 comments
\n\n\n\nThere are at least some technological solutions here, such as anonymous credentials. [1] Modern versions of this technique allow one to associate metadata (like a proof of age exceeding a threshold) in such a way that the verifier can't even correlate repeated requests across users.\nGovernments that are serious about age verification and individual privacy (which, doubtful they truly are) should agree on a protocol and set up certificate issuers that are associated with a digital ID. Then age verification will not be an invasive procedure or risk data leaks or insider threats.
[1]: https://blog.cryptographyengineering.com/2026/03/02/anonymou...\n— j2kun\n
Show HN: I made Google Trends for Hacker News by indexing 18 years of comments
\n\n\n\n345 points | 86 comments
\n\n\n\nI host a publicly open database with Hacker News data at https://play.clickhouse.com/play?user=play#U0VMRUNUICogRlJPT...\nSo you can create any sort of similar services in a single SQL query and an HTML page.
I also hosted it as a publicly accessible data lake, which you can query from everywhere: https://github.com/ClickHouse/ClickHouse/issues/29693#issuec...
It is also updated in real-time.\n— zX41ZdbW\n
Ford rehires 350 engineers after AI fails to preserve expertise or train juniors
\n\n\n\n439 points | 209 comments
\n\n\n\nInterestingly, there were no consequences for the execs that made this 'mistake'. There seems to be almost unlimited cover for execs cargo culting on using AI as a pretext for layoffs. If it doesn't implode almost immediately, they get massive bonuses, if it blows up in their face, oh well they had the courage to 'take a bold strategic decision'\nIn other words, they don't really have a plan, but they are happy playing with people's lives via layoffs, since it's the 'in' thing to do. The incentives are huge on the upside and zero on the downside for them.\n— khriss\n
LastPass notifies users of yet another data breach
\n\n\n\n331 points | 149 comments
\n\n\n\nHow does anyone seriously trust LastPass anymore? Years ago, I was working for a company handling bank data. They were using LP immediately following a previous LP security incident and had no plans to migrate away.\n— jagged-chisel\n\n
Hey Nico, you didn't vibe code your data room but stole it from Papermark
\nhttps://xcancel.com/mfts0/status/2070080422482977095\n\n\n280 points | 113 comments
\n\n\n\nIf we take what they're saying as fact and that they didn't copy and paste the code, but for all intents and purposes the LlM basically did reproduce the same code based on its crawling of the repo and not respecting the license. It would make a great civil case for the courts to decide.\nTheir defence seems to be \"well we asked an LLM to reproduce your work, so 'WE' never copied your code\". Smells bad to me.\n— bilekas\n
IBM debuts sub-1 nanometer chip technology
\n\n\n\n303 points | 162 comments
\n\n\n\n> logic technology can extend for the first time below the 1 nm node, advancing the era of angstrom-level scaling, where dimensions approach the size of individual atoms. While transistor nodes now refer to a generation of manufacturing technology versus an exact physical dimension, IBM’s 0.7 nm technology—also referred to as 7 angstroms—demonstrates how continued scaling remains possible.\nContinuing the well established trend of making bold claims about physical dimensions that have nothing to do with any of the structures in the chip, and the name scales better than the tech.
What they actually deliver is a \"nanostack architecture\" built with ~5nm features that according to them is comparable to a hypothetical real sub-1nm chip.
It's an impressive achievement nonetheless but it looks like the industry has a few too many marketers.\n— buran77\n
Countries are competing to see which can carry out mass surveillance the best
\n\n\n\n290 points | 110 comments
\n\n\n\nThe internet, as it was before the one-way ratchet started to close, feels more and more like a lightning in a bottle that nobody in power wants repeating ever again. Everything in the past couple years has been going towards the centralization into a small number of services, walled wastelands that require you forfeit any kind of anonymity to even browse, tightly coupled to the countries they operate in, and especially for tech corpos, practically an extension of surveillance agencies through PRISMesque programs.\nSoon enough (and already the case, if you're one of the unlucky ones) you won't even be able to browse it without explicitly allowing Google to track you on every single website you try to access through your Google-approved, constantly monitored handheld device, linked directly to your identity.
Commercial VPNs are not a solution, they're merely kicking the can down the road, and shrinking the number of people that will complain once they will, finally, come for them too, first by requiring strict accountability to providers and age verification, then outright banning any that do not comply.\n— 0x_rs\n
![https://image.non.io/d83664c1-5807-4a18-abe4-41928c198410.we...](\"https://image.non.io/d83664c1-5807-4a18-abe4-41928c198410.webp\"){kind=link}
![https://image.non.io/257dc9cb-9e6b-4e00-8f12-23ea5e073649.we...](\"https://image.non.io/257dc9cb-9e6b-4e00-8f12-23ea5e073649.webp\"){kind=link}
![https://www.da.vidbuchanan.co.uk/adamation/image.png](\"https://www.da.vidbuchanan.co.uk/adamation/image.png\"){kind=link}
![https://i.pinimg.com/originals/b1/82/dc/b182dcc291495c013c98...](\"https://i.pinimg.com/originals/b1/82/dc/b182dcc291495c013c9820abef8c087a.jpg\"){kind=link}