607 points | 325 comments
\n\n\n\nYesterday I tried to google \"can the Halifax Wanderers still make the CPL playoffs?\"\nSo obviously what appears right at the top is the AI summary, which told me \"they've already secured their #4 position and made the playoffs\". I knew this wasn't true, and I guess I could have just scrolled down a bit further and found my answer but now I was curious.
So I said \"that's not true, they're still #5, what I want to know is _could they still make the playoffs_\"
It says they've got an upcoming game against Ottawa, and if they win their chances are good. That game has already taken place, so I correct it again and finally I get a reasonable answer.
My question is: what's the point of the AI in the search engine if it itself isn't going to use the search engine first before answering? Like, I can't wrap my head around that. The answer is on the same page as its hallucination. It could have done a cursory look around before first hallucinating something completely false, and when corrected the first time giving me outdated information. It's meant to be A SEARCH ENGINE!\n— Hugsbox\n
Owed a billion dollars in Nvidia stock
\n\n\n\n464 points | 192 comments
\n\n\n\nAn open question is what happened to the 15,625 shares that he received when he exercised his options in 1996?\nIf he had held on to those, they would be worth even more than the additional 9,375 shares he was entitled to -- about $1.7 billion using the same numbers in the post.
My guess is that he probably sold them when they were worth a lot less then they are now, and would have done the same with the additional shares too.\n— jonas21\n
311 points | 167 comments
\n\n\n\nThis is the golden age of model training. Some days ago, I decided I wanted a local CPU only model that can perform exceptionally well for English to Bash translation (to avoid the googling for command syntax). I got a bunch of subagents to generate large amount of training data (140k+ samples), got the Qwen 3 0.6B base model, pointed Astra at it, and off to the races. It trained for 2 days (on and off) and I got a surprisingly good model for my task! The total active time I spent was a few hours. And it is still improving, what a time to be alive!\n— GodelNumbering\n\n
Installing NeoVim caused original Vim undo files to be deleted
\n\n\n\n254 points | 204 comments
\n\n\n\nThis story has no references that support the author's version of events, but it does appear to be substantially true that:\n1. The change would break undo history, for both Neovim and Vim.
2. This means Neovim would delete data created by a different program, on another user's computer.
3. This was known before the feature was released.
4. They did it anyway.
I don't think there can really be any post-hoc justification of this.
https://github.com/neovim/neovim/pull/13973#issuecomment-789...\n— jeremyjh\n
There are no \"rogue\" AI agents
\n\n\n\n324 points | 240 comments
\n\n\n\nA little over two decades ago, my then girlfriend was arrested for \"writing malware\" (which was not against the law at the time, and which was never released into the wild and never caused any damage). This set in motion a chain of events that effectively ruined her life.\nFast forward to today, and we have multi billion dollar corporations pumping out malware at breakneck speeds, compromising various systems (including those of foreign governments), and no one is getting arrested. Instead we're gawking at the marvel of these systems and are playing word games about whether or not it's a rogue system. If anything, it's making people richer.
Make it make sense.\n— elric\n
210 points | 155 comments
\n\n\n\n> It adds these // whenever anything close to tech is mentioned.\nI've done this before, in what I would argue was tasteful at the time.\n— not_a_bot_4sho\n
The Normalization of Inexplicable Failures
\n\n\n\n233 points | 95 comments
\n\n\n\nI am big on reproducibility (nix aficionado) and determinism (flagging test failures are a red-alert, all-hands-on-deck situation in my world) and correctness.\nI am also big on testing (the correct things). And nine-nines (big on Elixir).
And... I'm also big on agent-assisted dev. Which requires pretty much every check in the book to stay productive in. And that's fine to me. I've seen bugs that I wouldn't have made myself. And I've also seen my own bugs fixed. They've all gotten fixed in short order. I don't see why this is a problem.
Raise your personal standards.
Thing is, the unreliable-software situation was already untenable before agents (in poor hands) made it worse.\n— pmarreck\n
Show HN: Lofi Cities – Pixel-art city nights with browser-generated lofi
\n\n\n\n216 points | 94 comments
\n\n\n\nThe product hunt ad is breaking the immersion for me. I know that most cities have ads but still, I prefer ad free pixel art.\nIt’s also not realistic, in one landscape the billboard is taller than a 9 story building.\n— thih9\n
In an $80 motel room, a discovery to shed light on the origins of life
\n\n\n\n194 points | 79 comments
\n\n\n\nhttps://archive.is/2026.09.26-101638/https://www.nytimes.com...\n— ouked\n\n
Don't couple your Go code to GitHub
\n\n\n\n215 points | 98 comments
\n\n\n\nTrue, but beware of the domain name you're using. Because VeriSign may unilaterally decide to delete your domain name along with thousands of others [1] and you're back to square one…\n[1] https://neil.fraser.name/news/2026/09/03/\n— p4bl0\n
Breaking Up with Google Play: Why Conversations Is Now Free
\n\n\n\n444 points | 177 comments
\n\n\n\nI think it bothers OP less that they take a 15% tax than the fact that google provides terrible support for their own play store. If they would take that tax and provide good feedback and speedy version reviews, nobody would ever complain - it is expected to pay something since the play store doesn't run on good thoughts and prayers.\nBut because they're a monopoly (or a duopoly if you count apple, which is a different platform altogether) they can afford to act this way.\n— pi-victor\n\n
I'm the mom in that viral Giants clip. Let me tell you about my husband
\n\n\n\n426 points | 173 comments
\n\n\n\nVery well written both in prose and tone. I'm glad she decided to tell this story. If the author isn't a professional writer I think she could be.\nThis is a great \"behind the scenes\" style look at the actual people behind a viral clip. Luckily they have an extremely strong relationship to help soften the blows here, I'm not sure an average (not even bad!) relationship could come out of something like this unscathed.
There's something feral in us that comes out from time to time, especially online from the safety of our screens. I found the announcers within the range of good fun but when it gets to people reaching out to break up their marriage or telling him to kill himself it becomes really sobering.
I think this is part of what people are calling the lonlieness epidemic - despite thousands of screaming voices the whole thing makes me feel hollow and hopeless.\n— collingreen\n
Show HN: Reladraw – A diagram language where you decide where to place things
\nI love making diagrams to help understand, plan, etc. However, the options are (A) auto-placement languages like Mermaid or Graphviz (which don't let me decide how the diagram looks), or (B) software like Draw.io which are powerful but are very time consuming (and inefficient for agents to manipulate).I wanted to have the benefits of both, where you can define a diagram in a diagram language, but also retain a high degree of control over what the diagram looks like.
I also wanted this to work well for humans and agents.
On the Github link, there's a playground where you can try it out without installation. There's also instructions for a simple npm install and for installing a skill you can use with Claude or other agents.\n\n\n
267 points | 78 comments
\n\n\n\nNice to see tools improving human-agent collaboration.\n— DylanMerigaud\n\n
DeepSeek Elastic Compute (DSec)
\n\n\n\n228 points | 76 comments
\n\n\n\nIt seems every DeepSeek paper/patent has a huge number of authors, and this one is no exception. They couldn't even fit everyone on the page, there are 31 others not shown. This could be an asset protection strategy (i.e., human assets). Imagine if there were only 3 authors. Those authors may get hired away by competitors. If you list every employee on every paper then competitors don't know who to lure away.\n— flowerlad\n\n
220 points | 83 comments
\n\n\n\nThe concurrency and threading in Go just feels like magic compared to every other language. I'm a goroutine addict and I refuse to be rehabilitated.\nJust from observations over the years, I don't think there's any other language quite like this, in terms of how things can end up happening in any thread.\n— SamInTheShell\n
Plunging test scores are a slow-moving catastrophe
\nhttps://archive.ph/n8Akh\n\n\n169 points | 318 comments
\n\n\n\nThe drop from 2018 to 2022 is as big as the one from 2022 to 2026, so it's not obvious whether AI had a role at all, although it's highly plausible.\nI suspect the bigger culprit is the optimized monetization of human attention, because pre-algorithmic social media doesn't seem to be as destructive. The fact that science wasn't as affected as math and reading (both of which rely more on attention/practice than rote memorization) somewhat supports this.\n— dumberquestions\n
Revealing the details of how OpenAI agents hacked Hugging Face
\n\n\n\n397 points | 240 comments
\n\n\n\nImagine having a virus escape a sandbox, why are we worried about the virus but not the incompetency of those who are responsible for setting up the sandbox?\nIf I post something on the Internet today claiming that I asked my agent to do X but it went rogue and did Y, all I will be getting in return is a jar full of \"skill issue\".
Should we worried about people using LLMs for attacks? Yes, but not in the premise of LLMs going rogue but someone with the intention of abusing it to cause harm. And this is not something we as individual or even company can deal with, responsibility should be held by those who use it, in a legal way.
I am baffled by the fact that up until now, no one is held responsible for so many incidents reported publicly or privately. At this point, it's free marketing, if I am CEO of any AI company, I will run swarm of agents hacking all NGOs and stating that I am just looking for some random piece of data that happened to be hidden in their servers, at least that's what my LLMs think, not me. Then I will start preaching everyone how dangerous this piece of technology is and start giving out free tokens for these NGOs so they can start defending themselves and we should slow the f down.\n— damowangcy\n
Ollaya – Ollama for open-source, Jev-style decision models
\n\n\n\n299 points | 88 comments
\n\n\n\nI'm not sure what this means for AI startups if their innovations can be copied by OSS so quickly (what, like 2 weeks?). There's \"consumer surplus\" for everyone, to borrow an economic concept. But we do ideally want some of the surplus to flow to the innovator, too. I know there were precursors, but that's fine - it's hard to have a totally novel idea in such a popular field. I don't know what the end game is for TypeSafe - they'd need to demonstrate perpetually better results, or compete in another axis: UX, support, custom solutions, etc. So much of the time, someone proving a concept, or it simply getting enough publicity, is enough for a \"Cambrian explosion\" of follow-ups and copies. Famously, that was true for \"Attention is All You Need\", and the general idea of \"next-token prediction\" being so powerful.\nWe've stumbled into general differentiable models..\n— pradn\n
U.S. appeals court upholds designation of Anthropic as supply chain risk
\n\n\n\n356 points | 661 comments
\n\n\n\nI know everyone says this is political but it actually seems like a textbook designation. Anthropic wanted to have rules on how the military used AI, the military said no and therefore doesn't want anthropic used anywhere in their supply line.\nThis is like a pen manufacturer not wanting their pens used to sign drone strike orders, now the military needs to have a special box of pens that don't have stipulations attached. With AI usage it would be the same thing except applied to entire product chains. It seems like it would just add more complexity to operations.
You can agree with the rules anthropic wanted, but having rules set by a private company at all that apply to the military does seem fair for the military to object to.
>The Department reasonably feared that Anthropic might manipulate Claude’s design to prevent it from performing national-security functions that the Department deems contractually authorized and necessary
Though they'd probably put the DoD on the cybersecurity whitelist today, the very idea of the claude whitelists for certain functionality already exists and is being used by them today.\n— ApolloFortyNine\n
Platform-independent SIMD in Go
\n\n\n\n344 points | 132 comments
\n\n\n\nhttps://imjasonh.github.io/playground/palette-swap/ swaps colors in a provided image in wasm, entirely locally in your browser, to benchmark portable SIMD vs non-portable archsimd vs non-SIMD.\nPortable SIMD is ~11% slower than non-portable SIMD in this case, but both are ~5x faster than non-SIMD.\n— ImJasonH\n
Git-bug: Distributed, offline-first bug tracker embedded in Git
\n\n\n\n293 points | 94 comments
\n\n\n\nHi, author here, nice to see some interest :-)\nFYI, this is my near-term roadmap:
\nAlso, while there is some attention ... I'm considering working on this full time. If you have some advice or opportunities on how I can support myself doing this, let me know!\n— michaelmure\n- have the webui accept external auth (like github oauth) so that it can be a public portal and accept external interactions\n - have the webui expose a git remote endpoint\n - slightly rework identities (and likely root them in did:plc for pubkey distribution, the identity system from bluesky, without being an ATProto thing), which would allow to share identities between repos way more naturally\n - extend to support pull-requests, possibly CI. That would make it a somewhat complete local-first forge that you can also self-host trivially\n
304 points | 96 comments
\n\n\n\nWube is such a unique game studio. They do a lot of very solid technical work and look like they have fun doing it. I particularly appreciated their recent post about porting Factorio to ARM64. They had a business reason to do it but it's clear the team was mostly just curious about whether ARM64 was a viable gaming platform now. And then shared the results with us. They seem like fun hacker guys.\nhttps://www.factorio.com/blog/post/fff-446\n— NelsonMinar\n
202 points | 95 comments
\n\n\n\nHigher order thinking is more important and rare.\nAn aggressive first principles approach often leads otherwise well-intentioned technologists into strategic / ideological dead-ends.
Do we do things because it's the \"right thing\" to do in the moment, or because of the final outcome that will eventually result?
The most ideal answer is somewhere in the middle. I am far more interested in the total area under the curve than a single instant in time.
In lieu of intentional higher order thinking, simply working backward from your customer on a regular basis will generally accomplish the same outcomes.\n— bob1029\n
Jury finds Facebook liable for deceiving users in Cambridge Analytica case
\n\n\n\n223 points | 49 comments
\n\n\n\n> Meta agreed in August to pay up to $18 billion to settle the multistate lawsuit surrounding child safety issues. Buried in the 130-page settlement was an agreement to release Meta from future liability related to the Cambridge Analytica privacy breach, making New Mexico the only state to pursue a case. Florida was the only other state that did not sign the settlement, saying it was not tough enough on Meta.\nThis is an inconvenient reminder for people who want more \"regulation\" and who believe states with high levels of paper regulation like California really care about protecting consumers.
Years down the line, you'll find that a bunch of the people who worked on these cases have cush jobs working on the other side.\n— ElProlactin\n
How we learned to stop worrying and love campus surveillance
\n\n\n\n211 points | 117 comments
\n\n\n\nThere is an effectively unlimited supply of people willing to take sums of money to work on anti-human endeavors, so long as they and theirs are not the humans involved.\n— jcomp\n\n
Show HN: Jev Plays Pokémon Red
\nHey HN! Wanted to share a fun project I've been hacking on. Given Jev can make decisions really fast (but not fast enough to play Doom yet sadly), I wanted to try and push it to play a more complex game than Tetris. So I went with Pokémon.I've spent endless hours playing this game as a child so building this was a ton of fun.
I open sourced everything in case you want to hack on it yourself here: https://github.com/christianmat/jev-pokemon
The game is being streamed live including the tokens and cost - hopefully we get all the badges and don't get stuck in a cave :)\n\n\n
185 points | 78 comments
\n\n\n\nThis is so interesting to watch. For a couple minutes I was in awe of how quick and cheap it was. Then I saw just how bad the decision are and how it would get stuck in strange loops of going in and out of the same door to no end.\nThis seems like a technology heading in the right direction but not quiet there yet. Excited for what they are cooking up but probably won't start building around it yet.\n— stusmall\n
877 points | 248 comments
\n\n\n\nIt's sad to see F-Droid using the now-standard design convention of categorically refusing to draw lines between things. No visual differentiation between sections of the UI, no clear indication of what's tappable and what will happen on a tap, no indication of scrollable areas beyond content being cut off. I'm willing to chalk some bad text alignment or uneven padding up to amateur designers working for free, but the design ethos on display, the thoughtless trend-chasing, is depressing to see.\n— idle_zealot\n\n
364 points | 363 comments
\n\n\n\nI genuinely believe that in 2015 Apple had the balls to resist and today they don't.\nI am judging by a simple fact, that \"please confirm your age\" screen is now mandatory during the iPhone setup in all countries, and in some it's behind a KYC. I have a strong opinion that this is insane. And once they let the foot in the door - there is no closing it.\n— egorfine\n
Why is the liver so weirdly regenerative?
\n\n\n\n364 points | 186 comments
\n\n\n\nHi, your friendly, local pathologist here. I'm the kind of doctor that diagnoses liver diseases, cancer, lots of infections, inflammatory processes, benign tumors, and runs the lab that a lot of other doctors use the results from to inform a lot of their other decisions.\nIf you really want to learn about this stuff, Robert Weinberg's The Biology of Cancer is where to start. It introduces concepts starting where a high school student could understand and carefully walks through the history of our development of knowledge until, by the end, the pithy chapters are helping novel concepts for real science flower in your brain all on their own. It's the best book of biological science I have ever read. 10/10, will read again.
https://wwnorton.com/books/9780393887655\n— 0xWTF\n
Show HN: Whiteboard (YC W26) – An open-source IDE for thoughtful software design
\nHello! We’re Sid, Alex, Ketan, and Milan. We’re building Whiteboard (https://whiteboard.dev.fast/), an open-source desktop app where humans and agents can architect software together in a common workspace. Here’s our repo: https://github.com/devdotfast/whiteboard.We were missing the feeling of a “whiteboard session” with another dev where you leave with a deep understanding of a system, so we built this app for ourselves. Whiteboard plugs into the tools you already use - e.g. Claude Code, Codex, etc. – and gives your agent an SDK to draw on an in-app canvas to describe its work. We began with an MVP based on HTML artifacts and started rethinking the app as we ran into limitations:
1. Built on top of CodeOSS: We found that in pure HTML tools it was hard to connect a spec or diagram to code. In Whiteboard, when you click on visualizations like a sequence diagram, an entity relationship diagram, or a quote from the agent’s trace, you can jump to the underlying code directly. When navigating code, you get keybindings and LSP support from VSCode out of the box. We’ve found this is especially valuable because tradeoffs are often only discovered after a first pass at implementation (re: slop)
2. Semantic diff viewer: we wrote a semantic, AST-aware diff viewer in Rust so you can only view the code changes which are relevant to you [1]. We’ve set up some sane defaults: large added functions are summarized as pseudocode, and things like unit tests and large documentation changes are collapsed / hidden. This is all customizable with a WASM-based plugin system.
3. Decision Log: We found it difficult to reason about what set of decisions our agents made autonomously. So we built tools for agents to query and link their own traces to the Whiteboard, so you can understand how the requirements that you set were implemented, and understand what decisions your agent made autonomously.
Here’s a quick demo video explaining more: https://www.youtube.com/watch?v=ChPn3ftULWE
Folks at companies like Salesforce and Modal are using Whiteboard today as a review tool for architecture or spec-level changes – really any change where they want to be involved:
1. Reviewing your own coding agent’s work: because Whiteboard makes it easier to review large amounts of code, folks will typically have their AI agents create a prototype and a corresponding Whiteboard session so they can iterate on the design.
2. Reviewing other people’s changes: We’ve found that Whiteboard is particularly helpful when composed with tools like Greptile. For example, you can run an automated code reviewer on small changes and escalate to a Whiteboard session for the changes that require human judgement.
Why we built this: we’re four buddies from college who quit our jobs as tech leads right before agentic coding became industry standard. As we iterated towards an MVP for a previous idea, we struggled to maintain a comprehensible codebase while reaping all the velocity benefits of agentic coding. As more PRs were merged without our understanding, we felt a ‘cognitive debt’ begin to seep in, until it became difficult for us to even contribute to the system [2].
We’re releasing our desktop app under an MIT license. Please poke through and feel free to contribute! Eventually we’ll charge companies for a hosted web version that manages whiteboard session creation alongside features like trajectory storage and multiplayer reviews. Everything will always remain self-hostable.
Thanks for reading, and we hope you try it out! We would love to hear any feedback and to learn from your expertise.
Here’s are the project links again: https://github.com/devdotfast/whiteboard, and you can install (for MacOS + Linux) at https://install.dev.fast
[1] diffs library: https://github.com/devdotfast/diffr\n[2] Credit for the term ‘cognitive debt’ goes to https://www.geoffreylitt.com/2026/07/02/understanding-is-the...\n\n\n
269 points | 109 comments
\n\n\n\nSo are we just letting go and having clankers write everything now while we have other clankers \"explain\" it to us? Just collectively, \"fuck it\" ?\nThe software I use day to day is just getting worse and worse, and I can only guess that AI is accelerating that problem. Like I can barely convince spotify to play a full album anymore without it \"failing to play a song\", crashing, or play without audio underruns. We're so screwed if we can't even play audio smoothly in 2026.
Sorry, just ranting. But I see tools like this that are basically encouraging people to \"let go\" of understanding what exactly your code is doing, and continue to let the robot pile more and more code on top of the bad abstractions we already have.\n— bschwindHN\n
Owners mourn spoiled food after firmware update bricks Samsung smart fridges
\n\n\n\n265 points | 270 comments
\n\n\n\nSamsung smart appliances in general seem to occupy a similar place in the market to vehicles like the Chevy Traverse or Nissan Altima, where people with technical knowledge in that area know them to be absolute horrific maintenance and repair liabilities and yet the consumer base places them second to God.\nObviously the review of a brand-new device or vehicle with 100 miles on it is going to have a disconnect from the realities of 60k miles of rocky ownership but I remain baffled when people buy a 7 year old used V6 Traverse, a vehicle extremely well known for catastrophic engine failure, and if I question their decision I get a \"the reviews were good!\". It seems like journalistic malpractice to not retroactively update reviews with a warning when critical widespread shortcomings are found later in a products life.\n— 100percentjake\n
Opus 5.5 is good at explainer videos
\n\n\n\n262 points | 134 comments
\n\n\n\nHere are a couple examples from r/ClaudeAI:\nMade entirely with Opus 5.5 + $3.21 of OpenRouter API usageClaude Code Workflow
https://www.reddit.com/r/ClaudeAI/comments/1wogab3/
Jaw literally dropped. I ran the prompt from the \"Made entirely with Opus 5.5\" post on my own project. Here's what Claude Code made on its own for about $4.Claude Code Workflow
https://www.reddit.com/r/ClaudeAI/comments/1wovwao/\n— kakugawa\n
GitHub has not removed malicious imitation software after 3 weeks
\n\n\n\n257 points | 112 comments
\n\n\n\nAuthor of the post here. Github finally took the offending page down approximately 10 minutes after the post appeared on the front page of HN. Total coincidence. I'm sure!\nMoral of the story. If you want even the most basic level of support from Github, you need to get on the front page of HN first.
And it seems they are able to do things very quickly, when they want to. Bastards.\n— hermitcrab\n
Italian parliament votes for return to nuclear energy
\n\n\n\n533 points | 346 comments
\n\n\n\nI've seen quite a few SMR proposals over the last couple of years and not a single one that dared to address the whole cycle from deployment to eventual decommission as a function of the total revenues - costs to operate. Most of these projects seem to be made to bilk investors or governments rather than to actually produce power. The day Italy produces the first KWh that is net profitable to the operators without counting subsidies I will be highly amazed, and that's assuming they get to operational in the first place.\n— jacquesm\n\n
Claude discovers a novel enzyme system with CRISPR-like repeats
\n\n\n\n439 points | 489 comments
\n\n\n\nCurrent evolved Cas9 (CRISPR) variants are highly efficient and relatively unconstrained in terms of their human genome targeting coverage. Smaller nucleases and higher targeting specificity would be useful. But therapeutic use is mostly limited by delivery.\nThis seems revolve around a known retron-like reverse transcriptase. A sober framing would be something like: Claude identified a previously undescribed genomic arrangement around a known reverse transcriptase. Not all that sexy.
For now, this is mostly a story about how AI can be used to parse existing data to discover new biology (which is fantastic!).\n— Spacecosmonaut\n
Claude Code reads AGENTS.md only when telemetry is on [fixed]
\n\n\n\n353 points | 199 comments
\n\n\n\nSorry folks, this is a rollout artifact, we needed a way to turn this off remotely via feature flags if it broke something, and with telemetry off you don't get those. It's already been fixed as part of v2.1.281 releasing today.\nThe mod is source available here: https://github.com/anthropics/claude-code/tree/main/mods/age...
Apologies again folks, this was a fully human error on my part - I should've found a better way to launch with a kill-switch.\n— mpoteat\n
Fixing the Portobello Police Station Clock
\n\n\n\n369 points | 84 comments
\n\n\n\nWhy does a portabello mushroom have a police station\n— monster_truck\n\n
Linux support is coming to Snapdragon X2 Series
\n\n\n\n327 points | 139 comments
\n\n\n\nI hope Qualcomm upstreams all the device tree kernel level stuff to Linux for every laptop model. One of the things I don't like about Arm Laptops is—for example—how even if a SoC is supported upstream, if the manufacturer does not upload a device tree for their device, then you're cooked.\nI read that while these Snapdragon laptops do technically have UEFI + ACPI, the information they provide is not useful for Linux and is more coupled with Qualcomm's proprietary drivers on Windows. Therefore, device trees are needed on Linux (I could be wrong about the first part).\n— hurricanepootis\n
338 points | 191 comments
\n\n\n\nI agree with the gist of the author's point, but often when managers say they don't want the details, they really meant that they do want the details, just to some unspecified level, that you must divine. Give them less, and they aren't satisfied, and you risk damaging their trust in your expertise. So I've learned to give enough details until they tell me the magical phrase. Perhaps if I were better at reading people, and managers are better at managing, then there would be fewer misunderstanding, and less annoyance, from both sides.\n— RJIb8RBYxzAMX9u\n\n
359 points | 308 comments
\n\n\n\nI had an Oculus Quest and it was a really great device. After the \"Meta\" rebrand, the quality rapidly decreased and now it wants me to upload my ID to Meta to keep using it. Absolutely not! Zuckerberg already has way too much information on me, uploading my state issued identification to them is such an insane ask, I will never do it. While this looks like interesting hardware, the company behind it has unacceptable and user-hostile practices, so I'll pass.\n— slowin\n\n
Seattle City Council votes to ban surveillance pricing in sale of groceries
\n\n\n\n286 points | 179 comments
\n\n\n\nThe best solution is a constitutional amendment that actually enshrines a right to privacy. Among other things, the retention, aggregation, correlation of any personal data should be illegal (including for commercial purposes). (Storage on behalf of users in encrypted form could be made OK. Could also be refined to support retention of data of the medical, legal, etc kind with the attendant non-admissibility protections.)\nThis would fix this issue, it would destroy the surveillance models of Google/Facebook, and it would fix the Flock issue, etc. It would also fix the Roe v Wade issue: women would be able to get abortions in the first couple months of pregnancy without the possibility of harassment, since law enforcement would have no capacity to detect pregnancy until then.
Also, tech won't save us had a podcast on the dynamic pricing topic:\nhttps://podcasts.apple.com/us/podcast/how-data-is-changing-a...\n— jsrozner\n
DoorDash Spent $1.4M Trying to Stop Mamdani from Becoming Mayor. Now We Know Why
\n\n\n\n283 points | 135 comments
\n\n\n\nEver since it was found that they took part of driver's tips for themselves, I didn't trust the company https://www.theguardian.com/us-news/2025/feb/27/doordash-new...\n— dayvid\n\n
296 points | 131 comments
\n\n\n\nPet peeve on Google's AI rollouts: there's no alignment across the three platforms they have, consumer, prosumer, cloud. Scroll to the end of every release, including this one, and you'll see different availabilities. The fun part is the models don't even have the same capabilities across platforms! Omni Flash, last I tried and read the docs, is video and text out on consumer and prosumer but video out only on GCP. So if your org disables consumer and prosumer, like mine, it's a coin flip whether you can use the fancy new models or what they can do.\n— rcr-anti\n\n
1118 points | 773 comments
\n\n\n\n> Claude Opus 5.5 is our first release since we called for pacing the frontier.\nInteresting how the very first line is used to remind the reader of their call to pace the frontier just last week, and everything else after that line is to demonstrate with very specific numbers how they absolutely are not pacing.\n— sailingparrot\n
1082 points | 563 comments
\n\n\n\nGPT-6 Luna being half the price of GPT-5.6 Luna is a really big deal.\nHere's GPT-6 Luna pelicans: https://tools.simonwillison.net/markdown-svg-renderer?url=ht...
And GPT-6 Sol: https://tools.simonwillison.net/markdown-svg-renderer?url=ht...
Scroll to the bottom for the GPT-6 Sol max one: https://tools.simonwillison.net/markdown-svg-renderer?url=ht...
For comparison, here are the pelicans I got for GPT-6 Astra: https://tools.simonwillison.net/markdown-svg-renderer?url=ht... - I still like the Astra Max one best.
Here's a comparison grid showing all of the GPT-6 and GPT-5.6 pelicans at all effort levels: https://static.simonwillison.net/static/2026/gpt-6-and-5.6.h...
The grid is actually really interesting, because it shows that the 5.6 family default to brighter colors than the 6 family.\n— simonw\n
Apple has added persistent 'ads' to iOS, and it's driving users crazy
\n\n\n\n315 points | 235 comments
\n\n\n\nI doubt John Ternus will change direction anytime soon, since it will look like he’s reversing the plethora of eyesore ads that Tim Cook added over his tenure.\nBut the Apple with ads is not the Apple that had some taste and discernment in the past. For a long time I’ve visited the App Store’s app update page directly (tap and hold on App Store icon to see the context menu option). Anytime I inadvertently go to the App Store home page or the few times I search, it’s an ad filled disaster!
From this article
> repeatedly attempting to prod customers towards even more of the company’s products might seem cheap, even distasteful.
From a recent post by John Gruber:
> Steve Jobs in 2011: 'We Build Products That We Want for Ourselves, Too, and We Just Don't Want Ads' [1]
Looks like Tim Cook, John Ternus and Eddy Cue really enjoy being swamped with ads in their products. Will there soon be a time when Apple executives start carrying some other brand’s devices with them to avoid having a rotten experience?
[1]: We don’t want ads https://daringfireball.net/linked/2026/07/28/jobs-we-dont-wa...\n— AnonC\n
OpenAI GPT–6 Astra breaks Enigma message that has resisted solution since 2005
\n\n\n\n303 points | 267 comments
\n\n\n\nFor anybody interested, the actual encrypted message:\n\nwhich, given misspellings, translates approximately to:BTTE UM ANGABE DES MARSQWEGES X BEFINDE MIQ IN X ROSENOW ROSENOW X SOFORT FUNKANTWORT X WASCHBBSCH\n\n— mmsc\nPlease specify the route of march. I am in Rosenow, Rosenow. Immediate reply by radio. Waschbusch.
Pentagon says overreliance on AI contributed to missile strike on Iran school
\nhttps://archive.ph/0V37g\n\n\n348 points | 178 comments
\n\n\n\n> It found the U.S. “failed in its obligation to do everything feasible to verify” that the school was a military objective and that the failure “went beyond mere negligence.” The report said the United States “directed the strikes at the building of the school while being aware of a substantial risk of striking a civilian object and acting recklessly as regards the possibility that this would happen.”\nReading the details, \"AI\" doesn't really seem like the culprit -it's a scapegoat.
The intelligence that it was no longer a military target never entered the target database, the team that was responsible for vetting the target list was gutted, and said team was never even consulted.
The White House wanted 1000 targets and pulled from their database without any due diligence. Whether it was an AI call or an SQL query - this was from pure human maliciousness and incompetence.\n— legitster\n
'We hacked the FBI:' Hackers say they have data on all FBI employees
\nhttps://archive.ph/96YBP\n\n\n324 points | 212 comments
\n\n\n\nAt this point, no one seems capable of keeping a large database safe. I assume all medical and biographical information that exists is in the hands of the major state actors.\nChina hacked 22.1 million records of US government employees:
https://en.wikipedia.org/wiki/2015_Office_of_Personnel_Manag...\n— jacobgold\n
AI Has No Wisdom and Neither Will You
\n\n\n\n333 points | 462 comments
\n\n\n\nThe problem imo is the slow deterioration of institutional knowledge that offloading the mental task of wisdom gathering to AI is causing.\nOne interesting comparison is to the history of manufacturing. West/America decided one day that manufacturing would be cheaper to outsource and better (short term) profit was to be made by outsourcing it all to China. The institutional expertise started to deteriorate, to the point that America simply didn't even have the capacity, or expertise anymore to produce stuff (such as grill brush [1])
I feel like you could take all the handwavy comment that are made today to dismiss this caution, and find equal dismissal back then when companies were actively outsourcing the manufacturing.
\"I'm coding 10x faster\" \n\"look at the output velocity per employee\"
\"we are producing much more (in China)\" \n\"look at profit / number of (manufacturing) employers\"
Seems ok if you're American / Chinese but I'm struggling to understand how the rest can be OK with allowing institutional knowledge to deteriorate while having an active dependency to the former two. We already see this with the tech dependency towards USA and manufacturing competition from China.
[1] https://youtu.be/3ZTGwcHQfLY\n— NalNezumi\n
Microsoft killed FoxPro in 2007. Anyway, here's FoxPro revived
\nVisual FoxPro stopped at version 9 in 2007. A surprising amount of it is still running, in 32 bits, because rewriting a 20-year-old business app is how you lose the business. A customer wanted to keep milking their app for the foreseeable future, so here it is: the same language on a new runtime (Rust, compiled to wasm, checked against the real vfp9.exe), tables no longer stopped at 2 GB, the old 32-bit .fll add-ins still loading, and lambdas, JSON and an HTTP server bolted on for good measure. Reports are not done and the builds are unsigned. MIT because why not?\n\n\n281 points | 168 comments
\n\n\n\nI know there are still a few niche industry that has an industry standard program written in Visual FoxPro, and I remember as pre COVID 2019 they were still using it. ( Edit: I just called and asked, still using it as of 2026 )\nNiche but no small business, commodity with over $400M USD annual revenue, the industry together that is easily $2B+.
If it is not broke, don't fix it. And there are industry that is extremely conservative in anything software, they will just keep it running for as long as they could.
I used to be sad that we don't have something better to replace these antique software. Now I have come to appreciate it, mostly because newer software haven't show they have manage to add value in the slightest. And most of the time they simply don't work as well.
Delphi, Visual FoxPro, Visual Basic, Flash. I used to think one day we will have something that is not only faster, easier and scalable replacement of those. Instead everything went more technical, more complex, enterprisy and non-user friendly.
And to the Author of the project. Thank You for making it.\n— ksec\n
I asked Meta’s Muse for its filesystem and it sent me 6.8GB
\n\n\n\n288 points | 143 comments
\n\n\n\n> About 20 Markdown files described browser use, connectors, payments, credentials, data handling, generated files, voice, goals, and scheduling.\nThis the state of software engineering in 2026.
Edit: clarified engineering to software engineering, which is more correct\n— tolugenius\n
OpenAI is well positioned to fast-follow Jev
\n\n\n\n285 points | 204 comments
\n\n\n\nEvery major AI shop has a ton of in-house classifiers already, big, small, generalist, specialized. Some are used in inference pipelines (e.g. safeguards), some are used in data preparation, training, analysis and investigation, research, various one-off and intermediate tasks etc. Offering them on a public API doesn't always make business sense. I don't see much substance to this buzz, looks like people that are new to all this are discovering that classifiers exist, they are more efficient at classification, and many tasks commonly done with generative models are classification in disguise. Which is not bad at all, a fresh look at their use is great to have.\n— orbital-decay\n\n
442 points | 222 comments
\n\n\n\nI know we have strong views on what a truly open model is (open weights, open training data, open training code etc.) but I really like how transparent they’ve been about the training of this model.\nThe realtime dashboard they shared during training (https://mimo.xiaomi.com/rl/) was an incredible learning and teaching tool for me, and they’ve been unusually comprehensive in sharing details about their methodology (check out that tech report - it's got lots of clever behind the scene tricks like Google or Deepseek writeups) and benchmark scores (even the stuff they didn’t do well on).
If you’re releasing an open model going forward, please consider offering the community more of this transparency!\n— rao-v\n
552 points | 158 comments
\n\n\n\nThe Mosaic browser (1993) had full text history search.\nWe then got a bookmark system that was every bit as terrible as a web directory.
It stayed that way. The delicious search revenue made organizing websites uninteresting. That obscure website you enjoyed a decade ago but don't even remember, they had lots of traffic like you. No point updating or keeping it online. You can't have rss in Firefox but here is a Facebook like button in your address bar in stead.
I've tried to maintain the bookmark menu but I rarely use it since everything is dead. Why aren't browsers storing a text version of the bookmark? Did people in 1993 have more resources than us? Should I be afraid it grows to a few GB over the decades?
A good few dead websites have a backup some place but there is no automation to find it. If you had a string of text from a page you might be able to search for it. If the page found is highly similar we might automate the process to have alternative location for bookmarks with a nice warning dialog.\n— econ\n
I don't want to read what you didn't write
\n\n\n\n562 points | 198 comments
\n\n\n\nAs I have been saying for years:\nWriting is fundamentally the transfer of information from your brain to my brain. If you have 1000 bits of semantic information you want to transfer, you can't give 300 bits of semantic information to an LLM and have it fill in the remaining 700, because it doesn't know what those 700 bits are. If it's able to guess those 700 bits correctly, then they aren't true semantic information, and you really only have 300 bits you want to transfer. You might as well transfer those bits to me directly, rather than having the LLM add on an extra superfluous 700 bits that I then have to filter out.\n— hatthew\n
ZuckOff Know when a camera is in the room
\n\n\n\n584 points | 2 comments
\n\n\n\nearlier link\nhttps://news.ycombinator.com/item?id=49785397\n— jacquesm\n
253 points | 119 comments
\n\n\n\nAs someone who was buying hardware in the late 1990s, it is hard to overstate the difference in buying experience between Sun or Digital (DEC) and someone like Dell. The former forced you into a live sales meeting, endless quote revisions, it was a nightmare. I remember calculating the server rails and power cords for a new Alpha server were going to cost more than a shipped/delivered Dell server that I could get the next day.\n— coreyh14444\n\n
470 points | 385 comments
\n\n\n\nApparently Grok 4.7 has 40% more weights than Grok 4.6, but the price ($6 output token, $2 input) is the same.\nGiven that the decrease in their margin and the fact they delayed the release of Grok 4.7 almost two weeks past the original date, XAI must not have been happy with the results for 4.7. And XAI also waited the day before Opus 5.5 is rumored to launch. I imagine Opus 5.5 will blow Grok 4.7 out of the water benchmark wise.
However, I have become skeptical of benchmarks. Grok 4.5 solved some issues setting up a buildroot system that Fable 5 couldn't do. I find the post cursor groks are phenomenal at frontend web development, though Claude is much better at backend ruby.
My favorite part of the new Groks has been how they speak in plain english. I simply cannot stand Claudish. Or even GPT, which doesn't have Claude's ticks but definitely likes to handwave explaining technical concepts. Still, nothing beats Claude 3.5 and 4 with explaining since it seems all models have regressed. I wonder if Grok 4.7 will also regress with English because of all the RL.\n— moojacob\n
335 points | 76 comments
\n\n\n\n> Spymarks are certainly not great for whistleblowers or anyone who doesn’t want to be persecuted for their words or affiliations. No matter where you stand on whatever issues, spymarks can be used against you and those you care about.\nHow do you spymark text that someone else wrote? You can't change the words or they'd notice\n— voidUpdate\n
Transformers Explained Visually
\n\n\n\n340 points | 51 comments
\n\n\n\nNicely done. For me the most fascinating thing about attention heads is the place where Attention matrix is already computed and is getting multiplied by Value vector. It behaves exactly like pushing Value vector through Dense layer of ordinary network where Attention matrix forms weights of that layer. So attention head is trained to construct this small single layer network dynamically during inference from Key and Query. And that's the point. That's rarely underlined in explanations of LLMs architecture and for me it's quite amazing that it works so well. This mechanism easy to observe in this particular visualization if you click through it.\n— andblac\n\n
Fable 5 – Median thinking declined in August
\n\n\n\n338 points | 234 comments
\n\n\n\nCould there be a benefit to releasing a new model, slowly dumbing it down over a couple months, then releasing a new model that’s marginally if at all better than the original to create a perceived improvement when in reality there isn’t really one?\nFor an industry that’s stagnant in progress yet relies on new frequent releases to survive (non-progress being an existential risk), this could make sense.
I have no idea if that’s what’s happened, I completely pulled it out of my butt. And I have no idea is the actual frontier is stagnating.\n— talon8635\n
ZuckOff is a free app that sees Meta glasses before they see you
\n\n\n\n292 points | 298 comments
\n\n\n\nNot sure why you'd use this sketchy vibecoded proprietary app over the same thing that's open source and existed for a while\nhttps://github.com/yjeanrenaud/yj_nearbyglasses\n— flexagoon\n
ChatGPT now knows what you do on other websites via ad collector
\n\n\n\n534 points | 299 comments
\n\n\n\nI am once again happy that the EU is fighting practices like these via legislation.\nSome outcomes can be annoying, but the net result is still positive, for the consumers and their data privacy at least.\n— thih9\n
Qwen-Image-2.1: Compact, efficient, and unified image creation
\n\n\n\n261 points | 99 comments
\n\n\n\nSo thoughts\nPositives
• It's a heck of a lot smaller than Qwen-Image 1 (20b parameters) at only 7b, making it one of the smaller open-weight models available (Z-Image Turbo is one of the few that is smaller at 6b) when compared to Ideogram, Krea2, Flux2, etc.
• It supports native transparency (Qwen's team, as far as I know, is the only one attempting to tackle this). Even though it's relatively trivial to set up background removal postprocessors, it's also neat to see it natively supported.
• It's fast using QwenImage2.1 convrot, a 1MP image took around ~5 seconds on an RTX4090.
Negatives
• The license (assuming you respect it) is far more restrictive. The original Qwen Image 1 was released under the standard Apache license; this one explicitly forbids commercial usage without obtaining a separate license. On the other hand, a lot of us didn't expect the Qwen team to ever release \"weights-available\" ever again.
Qwen-Image 1.0, released about a year ago, only scored 4/15 on my GenAI Showdown Benchmarks. Since that time, they've been upstaged by Krea 2 (6/15) and Ideogram4 (8/15). I'll post the new results once I have some more time to run them.
https://genai-showdown.specr.net\n— vunderba\n
Pirate Face Rescues LLM Models from Deletion
\n\n\n\n414 points | 131 comments
\n\n\n\nSpeaking to the \"uncensored model\" angle: there's little reason to distribute abliterated weights anyway. Instead of orthogonalising the weights that write back to the residual stream, you can just orthogonalise the activations themselves. It's equivalent.\nOrthogonalising activations at runtime is computationally cheap. Just distribute the refusal vectors (few thousand floats per layer), then run against the stock weights. Antirez's DS4 already supports this: https://github.com/antirez/ds4/blob/8db1d1d155cb0400a86a86b9...
Abliterated weights are just a bad habit we've gotten into. It's also deeply suboptimal from a precision point of view to take a model that's already been QATed and distributed in pre-quantised form (DeepSeek V4, Kimi K2.5 or K3...), modify its weights, and re-quantise it. Similarly, abliterated models regain some of their refusal behaviour when they're re-quantised after abliteration -- avoidable by keeping the two separate.\n— wren6991\n
AX – Google’s Open Agentic Orchestrator
\n\n\n\n409 points | 162 comments
\n\n\n\nCould someone explain to me what the general workflow is now that people are converging to? I haven't really been catching up with the AI ecosystem but I was looking into agent sandboxes and VM's recently and there's a ton of these startups and tools now. Is giving the agent a temporary scratchbox really that valuable?\nI've been still just like, making VM's with proxmox, then putting my agent in the machine and letting it run free (with my dotfiles setup script making dev env pretty much free, though I could also just make a VM snapshot). What's wrong with that? Is that not the scalable solution for enterprise rn?\n— sigbottle\n
Samsung is expected to more than double output of its HBM4 and HBM4E DRAM
\n\n\n\n322 points | 207 comments
\n\n\n\nOn a related note, I was reading yesterday that apparently the real bottleneck for Chinese production of AI accelerators is HBM production, not processors or ASML equipment.\nThe lack of ASML EUV machines certainly hurts, and pushing DUV so hard results in abysmal yields of good chips, but you can compensate by running more wafers or making smaller chips, and the net result is that Huawei's Ascend production volume is limited by CXMT's HBM capacity not processor dies.
The problem is that HBM manufacture requires many steps (die thinning, via drilling, plating, alignment) where the equipment used by everyone else (Samsung, SK Hynix, Micron) is also blocked by sanctions, so the Chinese are having to develop all of this themselves too, which they have, but yields are currently low, even when using shorter HBM stacks.\n— HarHarVeryFunny\n
What happened to the Snowden archive
\n\n\n\n332 points | 212 comments
\n\n\n\nThis article linked back to the intercept's snowden archive articles [0], and while this article paints the Intercept in a bad light it lead me to reading a couple of these.\nWhile I heard many of the headlines the reporting is very in depth, including a bunch of smaller details worth poking into. I'd recommend people who are interested and who haven't done so yet to just read a couple of these.
[0]: https://theintercept.com/series/snowden-archive/\n— rtpg\n
Bill to Ban Private Equity from Owning Medical Practices
\n\n\n\n424 points | 305 comments
\n\n\n\nThere's a pattern to the kinds of companies PE buys and I think it points to the real problem.\nThey like companies with some kind of moat that makes it hard to unseat them. Basically, companies where there is no alternative for the consumer. That way, they can inflict abuse but know there will be nowhere to run.
There are two different ways to achieve this. Monopoly and regulation. Hospitals have both government granted locational monopoly and tons of regulations that make it impossible to compete.
Private equity is the symptom, not the disease.
Until we get at the disease, new monsters will be born with different name filling the same ecological niche. It's economic natural selection played out in the environment we created.\n— throwaway13337\n
Singapore’s National Library Board offers micropayments to build reading habits
\n\n\n\n173 points | 69 comments
\n\n\n\nI read a lot, and I own a lot of physical books, but I almost never actually read physical books anymore. I'm one of the happy few that used my Kindle for more than a week after getting it as a gift.\nI find that reading on an ereader is many times easier for me, primarily because I can make the text gigantic and I don't lose track of my place on the page nearly as often. I'm not dyslexic (I've been tested), but I do think that there's some issue with my brain where if the text is small I really can't follow it. On all my computers the text is gigantic and even then I will click and highlight text as I'm reading so don't lose place. I would love to be paid to read but I don't think I'd love to be paid to read physical books.
I remember my grandmother was wholly unconvinced that my Kindle was a different thing than an iPad or a Game Boy, and that it was not \"real\" reading, for whatever reason.\n— tombert\n
US Revokes Limits on Power Plants' Climate Pollution
\n\n\n\n204 points | 206 comments
\n\n\n\nThere are a multitude of arguments against this:\n- Investing in alternatives has led to massive new industry that is improving the economy of those countries that do it. If your argument is an economic one then jump onto the solar, wind and battery bandwagon.
- There are virtually no real short medium or long term gains economically here. Gas is the only thing still competitive with solar/wind and its costs are rising while solar and wind continue to fall. Building new maximum pollution plants would drop that internalized cost but who in their right mind would fund something so obviously DOA?
- Obviously the externalized costs of greenhouse gas emissions are deeply undervalued in this move. Even if they are 'fake news' in the US, the rest of the world is finally starting to take them seriously. The US's diminished soft power means it won't be able to easily bully the world into allowing it to pollute without consequence and such an obviously hostile move means it will loose even more of its soft power by taking this position. So on the international level this means we burn a lot of political capital and gain nothing but decades of distrust and anger.
- Current events show that energy security is dominated by decoupling from fossil fuels. This weakens the US strategically and continues to set it up to be manipulated by exceptionally hostile actors.
- Oh yeah and, of course, climate change is real.
This continues the US down the path of being the best buggy whip maker in the world. Worse than that, the US is becoming an obnoxious buggy whip maker who's neighbors are starting to hope fails horribly and will help make that happen as moves like this continue. This is stupid at every scale and in every dimension.\n— jmward01\n
181 points | 145 comments
\n\n\n\nTitle is click-bait with poor framing (i.e. if you are often wrong, then sometimes, you will be right).\nFinishing your goals does not \"make you right\". Life is not a binary where you are either wrong, or right. Thinking that you ever become \"right\" or that you ever achieve 100% confidence that you have arrived at a \"solution\" is arrogance of the highest order.
Life is about getting better. Every person, every product, starts at a beginning. A beginning is not \"wrong\", it is early. We try, individually and collectively, to improve ourselves and that which we are responsible for. The decision to ship is not whether we have made something \"right\" but about whether we have made it better. We ship, knowing we will ship another improvement after this one.
> 2. Gather missing information... 6. Achieve the goal
Sometimes in life, you need to ship first in order to gather missing information later. You cannot do that if everything has to be \"right\" before you ship it.
> I love being wrong
This is not the essay of someone who loves being wrong. This is the essay of someone who loves becoming right. But the deeper truth is that you will always be \"wrong\" - unfinished, lacking knowledge, under-developed, incomplete. Do not love being wrong - love improvement.\n— solatic\n
Laya the open source version of Jev
\n\n\n\n550 points | 142 comments
\n\n\n\nI think the main gripe that people had with Jev and Typesafe was the language used when they launched. To me personally it seemed like a parody/con/shady at first.\n\"Breakthrough\", \"our research went in another direction\" , \"Two years in stealth\", \"System One thinking model\", \"Jev can't hallucinate\", \"RLCD\",\"We are doing very cool stuff, but we will have to hire you to tell you\", - these are some of the things that they said on their website on the launch blog.
I had used versions of bert to achieve the same functionality years ago. But to me it seems like they were able to trick the VCs with \"can't hallucinate\" etc.
To the above author, kudos for sharing your work and making it open. Something like this shouldn't be closed in the first place when it has been available for so many years\n— prometheus1992\n
307 points | 115 comments
\n\n\n\nI haven't bothered to test the API, but you've effectively allowed a fully-open upload API? Who's paying the storage costs, and how do you prevent abuse?\n(Obviously I'm taking this more seriously than it's probably meant to)\n— AceJohnny2\n
I think you should almost never use AI to write
\n\n\n\n284 points | 141 comments
\n\n\n\nFrom TFA: \"Eric Schwitzgebel writes that . . . There’s a huge cognitive difference between nodding along while reading something and actually productively generating a text. Two reasons: First, once the text is on the page, it’s easy to passively let the approximate word suffice, rather than thinking about word choice in the same effortful, active way we do when generating prose de novo. Second, as I suggested above, I doubt that human beings, even experts, have a good sense of all the factors that shape word choice -- everything they’re being sensitive to. You would have phrased it slightly differently, and even if you don’t know that, or why, a different signal is sent and received.\"\nThis is the best articulation I've seen of why simply reviewing and copy-editing does not provide remotely the same value as writing from scratch. I spent a considerable amount of time over the past two weeks reviewing and improving a work document that was the output of an LLM. Given the number of people involved and the final level of effort, I'm firmly convinced that writing it manually would have been faster and resulted in a higher quality product. Getting the wording right matters.\n— patrickmay\n
224 points | 97 comments
\n\n\n\nUnrelated to the benchmark...\nI love StarCraft. I started playing it right from the beginning, most of my friends right now are from that era. I literally met people that have spread to almost every continent when I was in my early teens. We played at internet cafes and did not have access to the internet, that was priced differently...
I miss those days so much.
Everybody was from a different background back then, and nobody was anything other than a guy that plays StaCraft at the cybercafe... And now, we are in our 40's and I know Math teachers, history teachers, oil rig operators, software programmers, professional gamers, lawyers and more... hahah So crazy to think about it... and I know them, we talk, what a world.\n— pelagicAustral\n
What Zig felt like, coming from Rust
\n\n\n\n203 points | 243 comments
\n\n\n\nFor various purposes I work on a language comparison project that includes C and candidate successors such as Go, Zig. One question is the language to use for an archival port of a 1980's computer algebra system written in 32-bit K&R C. While Zig is a great debugging compiler, it's not yet stable enough to be the best target language for archival purposes.\nhttps://github.com/Syzygies/Compare
So you're in a restaurant where you don't speak the language, you can't read the menu, but you see three price points for set meals featuring the house specialty. (Say, \"Crossing the Bridge\" noodles in Yunnan.) Which do you choose? My tour guide, the author Fuchsia Dunlop, later agreed with me this is obvious: The middle choice.
So you're choosing between Go and C23 as candidate successors to K&R C. They both have \"royal blood\". One got the name. Knowing nothing more, which do you choose?
The answer is equally obvious. The one that got the name also got the warts.\n— Syzygies\n
222 points | 303 comments
\n\n\n\nA funny quirk of English I never noticed until I started learning Spanish:\nIn Spanish, to say for example, \"he is a doctor,\" you would say \"él es médico.\" Which more literally translates to \"he is doctor,\" which feels unintuitive coming from English. Spanish DOES have an equivalent for \"a\" - \"un/una\", so the natural thing for an English speaker to say would be \"él es UN médico\", but I understand that to be awkward/unnatural.
But then I realized, when you say the same sentence in plural - \"they are doctors\" - you don't have an article. You don't say \"they are some doctors\" (I had to Google, apparently \"some\" isn't an article anyway?)
So Spanish ends up more consistent - Singular: \"Él es médico.\" Plural: \"Ellos son medicos.\"
Unlike English - Singular: \"He is a doctor.\" Plural: \"They are doctors.\"\n— hbn\n
Tin: full-text search for Postgres
\n\n\n\n210 points | 80 comments
\n\n\n\nI think what we're seeing with every database company providing new full-text search capabilities is an example of AI coding productivity showing up in the real world.\nIt started with paradeDB and pg_search https://www.paradedb.com/blog/introducing-search
Timescale has pg_textsearch https://github.com/timescale/pg_textsearch
Neon and Databricks have Lakebase Search https://docs.databricks.com/aws/en/oltp/projects/lakebase-se...
Now PlanetScale.
AFAIK all of these are implementations of the BM25 algorithm. You can just tell an agent to read about BM25 and implement it in your system of choice. Cool to see. Seems like there's still a lot of juice to be squeezed out of how it's architected and integrated into each system, but you can't help but wonder if this will lead to aggressive commodification\n— andrenotgiant\n
Android 17 is the first since 3.x to add new APIs without releasing to the AOSP
\n\n\n\n431 points | 205 comments
\n\n\n\nThe amount of roadblocks Google is putting up for GrapheneOS is just ridiculous. None of their decisions make any sense, from the delayed source patches upstream, to the embargos, attestation issues, etc. Google simply regrets android being open source.\n— wps\n\n
500 points | 478 comments
\n\n\n\nPasskeys do marginally improve security against MITM and phishing attacks, but they are primarily for protecting the lowest common denominator from themselves: people who re-use passwords and/or don't use a password manager.\nIf you use multiple devices throughout the day, registering passkeys in all of these systems becomes a big headache with O(m*n) complexity, so putting the passkeys in a password manager is the only realistic solution. But this still breaks the login flow for a very common use case: how do I log in on a device that I don't own? With a password in a password manager I at least have the option of manually typing the password.
The biggest problem, though, is how users are pushed into it without any warning or knowledge of what they're signing up for. I've accidentally set up passkeys just by clicking an okay button a few times in the past and had to go back and figure out how to undo it after being blocked from login on another computer (which computer was I on again?).\n— drtz\n
532 points | 237 comments
\n\n\n\nMy wife and I have a mobile group/shared bot system where we can make mini apps and collab.\nAnything from baby stuff, groceries, shopping, planning, wine tracker app, simple/fun/useful data things, etc…
We have Tailscale on our phones and can instantly and privately see without deployment or anything crazy via our secure VPN.
Think shared Claude Artifacts that don’t live @ Anthropic.
Tried to first do this with Cloudflare Tunnels (because I love Cloudflare), but between the broken dashboard side of Zero Trust and nightmare of Warp… it was basically impossible to setup. I guess that’s all super enterprise, which seems to be very anti-Cloudflare philosophically to not be able to self do things.
Will check out Quick Tunnels but I think it’s missing the bigger integration offerings Tailscale has/does still.\n— TIPSIO\n
Claude Code now reads AGENTS.md if there is no Claude.md
\n\n\n\n423 points | 152 comments
\n\n\n\nGotta say, it is hilarious that this is the current top HN post. I feel like it's gotta say something about our current AI zeitgeist, that there is such vigorous attention on a seemingly-minor change. Feels like something is on the tip of my tongue but I can't name it at the moment.\nIf anyone wants to write/link a much better-thought-out post, I'm all ears!\n— winwang\n
US Military had close call after using AI for hallucinated intelligence report
\n\n\n\n365 points | 289 comments
\n\n\n\n> relatively poorly understood technology\nPoorly understood? how convenient...
LLMs are vectorial databases with losses that index statistically filled data, which uses a text interface to query such statistically filled data. The output is a string concatenation (statistically concatenated bit by bit).
When the LLMs are queried (prompted), you can get random mixed data as output, ERRORS, due to undesired indexes getting closer at one point while the string was being concatenated for the output, what affects the rest of the indexed content that will be concatenated.
It is intrinsic to this tech. The larger the context, the greater the probability of get mixed data. And if the provider lowers the precision of those indexes -in order to decrease hardware resources and energy consumption- such probability increases to the point where those errors are granted.
Even knowing that the queries can return wrong/mixed data in the responses, errors, the companies developing this, decided to introduce a new product, that connects such LLMs outputs to the command console, latter connected to internet, raw 'eval' running commands from such outputs witch obviously can contain whatever mixed random. Then we started to hear \"oh, it deleted my directory\", etc, and it seems the next one will be \"a missile killed my wife\", because it is a text concatenation engine with errors.
To name it \"hallucination\" is an euphemism... those are errors, and they are granted to happen at one moment. If they do not know this, then they ate too much marketing without doing their job, or it was a convenient contract for the pocket$ of someone.\n— drtgh\n
326 points | 62 comments
\n\n\n\nIncredibly happy to see this series of CF articles. I was always so proud of devs back in the days where RAM and processing were scarce and who had to get creative to fit even the most basic stuff in the budget. It seemed to me that after RAM and processing became abundant, most gave up on optimization and focused on shipping instead which meant now that even with several cores, a basic notepad or music player failed to work. In a way, RAM becoming more expensive has ushered in a new era of forced optimizations, which I'm really happy for\n— zer0x4d\n\n
Bend 2 and the Vibe-Coding Trap
\n\n\n\n284 points | 213 comments
\n\n\n\n> The problem is that vibe coding makes it possible to build a substantial solution before learning enough about the problem to recognise that a much better solution exists.\nThis example and Bend aside, I find this to be the biggest struggle with the perceived intelligence we have today. It's great at producing something that works, but it is not great at calling you out when you don't know what you don't know.
It's not able to educate and course correct you unless you have great self awareness and discipline.
That said, I think this goes for everything, it's easy to fall into this trap because it is very human. We simply don't know what we don't know, so it's not uncommon to revisit an old solution only to be enlightened that there is now new information that allows you to replace it with something much better.
I don't think anything here is new or changed, if anything changed is really just the rate that we experience this. LLMs make it easier and faster for the feedback cycle to happen.
Now back to Bend, I think putting your work out there and being unapologetic about it, open source even, and willing to take feedback, will go a long way.
I am more worried about the many closed source implementations of LLM built products that are being sold and people are depending upon that don't get this great criticism from many different thinking heads.\n— hmokiguess\n
Korea raises data breach fines to 10% of revenue
\n\n\n\n303 points | 90 comments
\n\n\n\nYou can just do what my university did, hire a small shell firm with 3 employees to hold all your data, and when it got hacked they just went bankrupt and we switched to a new shell firm with similar form and function.\nMinimizes money usage and does not require any security investments\n— augment_me\n
Warren Buffett Steps Down as Berkshire Chairman, Names Son to Replace Him
\n\n\n\n284 points | 187 comments
\n\n\n\nhttps://archive.ph/E24O9\n— rib3ye\n\n
ZCode, the GLM coding agent, silently uploads your Git history
\n\n\n\n250 points | 62 comments
\n\n\n\nLLM-paraphrased from the original post: https://blog.ferstar.org/en/posts/zcode-silent-workspace-sna...\n— outloudvi\n\n
Hister: A private search engine for the pages you visit and the files you keep
\n\n\n\n417 points | 124 comments
\n\n\n\nOhi, author here! Thanks for posting Hister. Feel free to A.M.A.\nMy first free software search project was Searx, a privacy respecting metasearch engine, but because of the limitations of the metasearch concept, I've decided to take a different approach.\nHister builds a personal search index from pages you visit, bookmarks, browser history, local files, and crawled websites. It stores extracted content with offline result previews, so information remains searchable even when the original page changes or disappears. It supports full text and semantic search, can run entirely on your own machine, and includes a web interface, command line tools, and an MCP endpoint for assistant integrations.
Website: https://hister.org/
Tiny read-only demo: https://demo.hister.org/
Ps.: It looks like our name conflicts with a registered trademark in the US. The owner of the other project has asked us to change it, so we’ll probably need to comply sooner or later.
Name suggestions are welcome! Ideally, the new name should be relatively short, sound good, and have an available .org domain.
Thanks!\n— asciimoo\n
243 points | 272 comments
\n\n\n\nI know someone who works in law and deals particularly with an area of US benefits and healthcare law. One of their workflows for lower-level employees at their firm involves taking in documents from healthcare plans and organizations, analyzing them for certain kinds of data, and then importing that data into an internal system they use to analyze and provide guidance on plans. The internal system can contain hundreds of documents for an individual client. All of the documents have the same information (roughly) but in totally diverse formats and styles. Once it's in the system, it's easy to compare and analyze across documents and the research process is much faster.\nThey recently bought a Claude subscription and began using Claude to do the initial read of the documents and output JSON they can import into their internal systems. The work still must be reviewed by an attorney - Claude is nowhere near making the kinds of judgments a lawyer would make about this content - but it has increased their throughput from 2-3 documents an hour to 8-10 documents an hour by killing the busy work.
LLMs have great advantages for this kind of work - but not for decision-making. I just don't see OpenAI ever admitting that.
(I've left some details intentionally vague because this is a very specific area of law and I don't want my friends to be identified without their consent.)\n— ivraatiems\n
Bend – A language that blocks AI mistakes via proof, on CPU and GPU
\n\n\n\n224 points | 121 comments
\n\n\n\nHi, I'm the author.\nHN staff: someone posted before me. Could we change the title to \n\"Bend - a language that blocks AI mistakes via proof and runs on GPUs\"?
Everyone: feel free to ask any question, but I'd be highly appreciative if you could be a bit civilized and respectful this time. I've worked on this for 1 year, nearly 16h/day, 7 days a week, and I'm giving it for free. You need not to use it. So, I'd be thankful if you could point occasional failures politely rather than throwing me in a lava pit.
Thank you!\n— LightMachine\n
Bonsai 2 27B: Near-Lossless Compression in a 9x Smaller Footprint
\n\n\n\n372 points | 115 comments
\n\n\n\nIf you want to try out out the GGUFs from https://huggingface.co/prism-ml/Ternary-Bonsai-2-27B-gguf#th... be aware that you need Prism's llama.cpp fork to get them to work, from https://github.com/PrismML-Eng/llama.cpp/releases/tag/prism-...\nThis should work:
\nThen open http://localhost:8331 for the (very good) baked in llama-server web UI... or run a prompt via the API like this:cd /tmp\n\n # Get the Prism macOS runtime\n curl -fL https://github.com/PrismML-Eng/llama.cpp/releases/download/prism-b10685-7dffb15/llama-prism-b10685-7dffb15-bin-macos-arm64.tar.gz -o bonsai-runtime.tar.gz\n tar -xzf bonsai-runtime.tar.gz\n\n # Get the ~5.95 GB GGUF model:\n curl -fL https://huggingface.co/prism-ml/Ternary-Bonsai-2-27B-gguf/resolve/main/Ternary-Bonsai-2-27B-PTQ1_0.gguf -o Ternary-Bonsai-2-27B-PTQ1_0.gguf\n\n # Run the server, I used port 8331\n ./llama-prism-b10685-7dffb15/llama-server \\\n -m Ternary-Bonsai-2-27B-PTQ1_0.gguf \\\n --port 8331 -ngl 99 -fa on -c 32768\n\nThat's running at ~20 token/second for me on an M5 Pro (after a server restart I got 44 token/second, not sure why), but I'm pretty sure something isn't working right, on startup the server said \"ggml_metal_device_init: - the tensor API is not supported in this environment - disabling\".\n— simonw\nuvx llm openai endpoint http://127.0.0.1:8331/v1 \\\n --model bonsai-2-27b --responses hi\n
248 points | 83 comments
\n\n\n\n> By 6:00 a.m. on July 25, we had confirmed local RCE through an image upload. We then placed Claude in an autonomous /goal loop against our own Discourse Cloud instance, proxied through rce.ee/ctf-forum to make it look like a CTF target as Opus refused write exploit for remote instances.\n> When we checked again at 10:00 a.m., the agent had achieved RCE on Discourse Cloud and demonstrated access by reading /etc/hosts. Using the generated exploit script, we managed to get RCE on OpenAI’s instance.
Between this and the HuggingFace hack, we've built systems that are so goal-oriented, and so capable, that they will do almost anything if they are convinced it is justified - or if they are playing a \"game\" where there is no goal but to win.
Of course I want my software to be able to audit its own security, and to defend against attackers who have the benefits of their own agentic systems. But at a certain point, did we need it to be trained so much on CTF games?
It feels like an entire industry watched https://en.wikipedia.org/wiki/WarGames and ended up thinking \"this is a challenge, we can just build a better WOPR, of course it will know when it's playing a game. Let's play Global Thermonuclear War.\"\n— btown\n
Iran school bombing: grounds to believe US was behind atrocity, UN finds
\n\n\n\n308 points | 265 comments
\n\n\n\nSeeing as it was a \"double tap\" it was clearly a deliberate target. Bomb the school, wait until officers from the nearby base with kids at the school arrive to try and rescue their daughters and hit it again. Clearly a war crime, possibly using Israeli targeting information, which clouds which State is ultimately responsible.\n— Wildgoose\n\n
271 points | 219 comments
\n\n\n\nThis doesn't seem like a well-thought-out post. I mean, I probably could have told you that from the fact that it's either pro-AI or anti-AI (in this case the latter), but this is a particularly poorly thought out post. Others here have pointed out the water-wasting red herring, but there are several other pieces of nonsense here — the general environmental doom-mongering, and labeling recursive self-improvement as \"mystical\", for example.\nMe, I pointed Claude Opus 5 at a C codebase I've been working on for years and it immediately found five serious bugs and told me how to fix them, as well as how to use Linux system calls I wasn't familiar with to solve some other problems. And I gave it a new language for PEGs I'd written up a couple of years ago, and it wrote me a working implementation in OCaml that afternoon, fixing several bugs in my example grammars along the way and resolving a conceptual problem I'd had a roadblock on. I asked Fable 5.1 to implement a minimal proof assistant, and it wrote a clone of J-Bob in Python, which I'm studying now. Although there is a certain firehose quality to this stuff, I sure don't intend to use AI to actively de-skill my brain.
That doesn't guarantee that AI will be a beneficial innovation overall, of course! As with most innovations, it probably depends on the balance between people being able to use the innovation to gain control over their own lives, and people being able to use it to gain control over others' lives. (Barring a FOOM scenario, of course, where our prediction ability is nonexistent.)
So, I don't think a pro-AI post can be well thought out, either. It's too early and chaotic to understand what's going to happen, and it may actually be uncertain. Pro and anti are both far too simplistic.\n— kragen\n
Thought I'd check how much in donations they've gotten. And I saw this massive 10 Bitcoin donation from back in 2023 (worth $800,000 now...)
From the activity history, neovim last sent bitcoin out of the address in 2019 so it's been max 7 years since they've definitely had access.
Does anyone from the neovim project know about this? \nSeems like a pretty significant amount of funding to have sitting there. Hope it can come to good use as I use neovim daily.
https://www.blockchain.com/explorer/addresses/btc/1Evu6wPrzjsjrNPdCYbHy3HT6ry2EzXFyQ\n\n\n
286 points | 225 comments
\n\n\n\nFor a reference point, when I was unemployed and between jobs, I got involved with Neovim for fun, and after some contributions, I eventually tried some full(ish)-time paid work. I wrote the native lua LSP client (:h vim.lsp, and the nvim-lspconfig repo) for Neovim in a few weeks for about $3k USD (circa 2019)? This amount could fund quite a lot of work to be sure.\n— ashkankiani\n\n
Canada welcomes EU proposal to become 'associate member'
\n\n\n\n242 points | 295 comments
\n\n\n\nAs a Canadian, I find this unbelievably good news. Canada, the EU, and other middle powers need to unite. Stronger together, while preserving what makes each country unique.\n— philippemnoel\n\n
Nvidia announces native GPU programming in Rust
\n\n\n\n565 points | 205 comments
\n\n\n\nI strongly dislike CUDA. Once you have allowed that proprietary cr*p into your C++ codebase, it is very hard to get rid, and you end up with code that is either tied to a single vendor or an #ifdef hell, probably both.\nThe best way to program GPUs is face up to the reality that they are not the same machine as the CPU, write your kernels in separate files, and launch them manually, like in Metal, OpenCL, and D3D12, etc.\nThese days we even have DSLs like Triton that make kernel writing much more ergonomic than anything you would hope to achieve in Rust.\n— jacobgorm\n
Training a 4B model to produce 81% faster query plans than Postgres
\n\n\n\n353 points | 68 comments
\n\n\n\n“81% faster query plans than Postgres”…on an 8 GB dataset that fits entirely in memory, with shared_buffers constrained to a fraction of that, queries warmed before measuring, and read-only SELECTs.\nI would be cautious about over fitting, it’s tough to say if those query plans would really be more optimal than Postgres heuristics at scale and with a bit more realistic OLTP workloads.
In any case, such is life with profile guided optimization. Many of us appreciate how database workloads can drift over time and with scale.
Kudos to the author for getting their hands dirty and writing up their experiments.\n— refibrillator\n
Hackers Got Inside a Flock Camera
\nhttps://micahflee.com/flock-cameras-are-riddled-with-securit...\n\n\n449 points | 208 comments
\n\n\n\nHaving hardcoded credentials is a sign of total incompetence. In this case at least it wasn't a password, but an API key which can be used to request credentials (stored in plaintext) which look like they'd get you access Flock's servers. Not quite as bad as a hardcoded admin password, and it's not clear what you'd be able to do if you did authenticate successfully as a camera, but its worrying enough. There have been enough vulnerabilities found in Flock's systems that it's pretty clear they aren't concerned about their security and it's plainly obvious that they don't care at all about our privacy or security.\nEven if we decided that this level of mass surveillance on the American public was acceptable to us, Flock Safety/Flock Group as already demonstrated that they can't and shouldn't be trusted to implement it.\n— autoexec\n
361 points | 177 comments
\n\n\n\nThe thing with a lot of these tricks is that you have to get into the habit of using them. I knew `Ctrl+r` for history since I learned about the command line. I even have a nice shell integration with fzf. But I still used the up/down arrow keys for years or scrolled up when I was looking for a previous command, because I never remembered the shortcut and just took the path of least resistance to find something.\nUsually what I do is writing these tricks down to a document that is easily accessible at a place where I know that I will look for when failing to remember them. I have a directory of docs where I write stuff like this down by language/tool/etc., so I can quickly look it up without having to search the internet. But I also have to actively remind myself that these things exist whenever I have a problem, stop me from just doing the inefficient thing.\n— phforms\n
Xiaomi Mimo 2.6 live post-training dashboard
\n\n\n\n394 points | 98 comments
\n\n\n\nI been using MiMo-V2.5 to do most of my work as software engineer, on a variety of projects I'm working on, and I been VERY happy with ROI. The model is very powerful! Not perfect – I've run in hallucination loops once or twice, but nothing a stop-then-continue wouldn't solve.\nThe cost is unbelievably low, and the quality of intelligence I get is equivalent to when I was working mostly with Anthropic models (late last year/early this year). I'm fully invested in MiMo and I'm very happy with it.
--\nPS: I also check almost daily to see if other models are capable of doing such great work. And they do – DS4F is powerful and DS41 is impressive, GLM 5.3 Flash gets a job done well, etc. – but when I add cost of M-token in the ROI math, Jeez! MiMo is an order of magnitude better.\n— joelwallis\n
The Google Play app review process now regularly takes longer than a week
\n\n\n\n264 points | 240 comments
\n\n\n\nMAYBE operating systems shouldn't have gatekeepers which can deny access to billions of customers for any and no reason at all. Apple and Google are WELL past due for regulation in this space. The fact that we cannot download and install software from the internet onto our phones JUST like we can do with our computers is a symbol of our inept and ineffective politicians.\nAnd before someone says \"well akshully you can technically do it on Android,\" Google has been working tirelessly to make it as onerous as possible for both developers and customers. For example, users need to separately authorise each app (browser, files manager, alternative store, etc) to install an APK from outside the Play Store. Google also does background scans using \"Play Protect\" which will periodically delete apps Google doesn't approve of. This happened to me with SmartTubeNext. I have a dozen other ways Google ensures users are discouraged from stepping outside the Play Store.\n— Gareth321\n
PS5 Linux lead quits: \"a bunch of noobs using LLMs\" that \"they don't understand\"
\n\n\n\n301 points | 211 comments
\n\n\n\nHobby groups projects like this are less fun for a lot of people who used to enjoy interacting with smart people. It's definitely become a game of just spam claude for answers with zero understanding or care for how anything actually works. That's fine to get things done and fine for a lot of side projects, but it definitely ruins the joy that people have in understanding systems and working with intellectuals.\n— eugenekolo\n\n
Original Sony PlayStation 2 security chip 'broken wide open' after 26 years
\n\n\n\n271 points | 77 comments
\n\n\n\nHello! I'm one of the people involved in reverse engineering the MechaCon. Please take note of the following:\n* It is already possible to run backup/copied discs entirely using software methods (from memory card, HDD, or DVD video player exploit), and in the case of the 50k series and newer (Dragon MechaCon), can use the \"force unlock\" patch that does not require any patches to copied discs.
* The dumps, alone, does not give enough information to create an \"\"\"hardware\"\"\" ODE (optical drive emulator). However, it is possible to make a modchip that replaces the MechaCon, relying on the DSP to continue to read discs.
* The current method to dump the ROM used in the SPC970 requires lots of writing to NVRAM.
* These dumps will eventually be useful for full-system low level emulation, as MagicGate and KELF/KIRX security goes through it.
* These dumps are useful for vulnability searching, which can allow functionality to MechaPwn (code execution on MechaCon, unlocking security e.g. set disc type and max LBN) or TonyHax (unlocking PS1 mode to read any disc, on SPC970 based MechaCons and all PS1 MechaCons).
* The contents of game discs (except those that use DNAS online authentication) are not encrypted. These dumps don't unlock anything additional there.
See https://gist.github.com/uyjulian/aaf4f4e2e74d980f620a9b73ba0... for updated info.\n— uyjulian\n
261 points | 167 comments
\n\n\n\nDespite all of the snark here, in my experience Salesforce SRE team is quite competent. The engineering challenges of running a large PaaS - not just with own apps, but with millions of customer-written apps running on it - are quite interesting, and sadly things happen. The status page makes sense to actual customers, it's the particular \"pods\" where a given service runs.\n— stmw\n\n
Show HN: An e-ink frame that hears birds and draws them as 1800s illustrations
\n\n\n\n654 points | 103 comments
\n\n\n\nThe underlying classifier, BirdNET, is a traditional neural network and not an LLM:\nhttps://doi.org/10.1016/j.ecoinf.2021.101236\n— divbzero\n
Introducing System One Models and Jev
\n\n\n\n635 points | 207 comments
\n\n\n\nFirst, congrats to the team on launching something genuinely interesting and new.\nSeems like a more accurate title would be \"Jev: Trading general purpose generation for fast typed inference\" or something like that.
This is interesting, but the speed comparison seems misleading? A generative model that can output code in a Turing-complete language can do anything a computer can do.
Jev can only generate structured output, right? This is probably super useful for classification/routing/scoring, but it's nothing like the code generating models we're all using today for code and automation.
Also \"can't hallucinate\" seems wrong? Sure, it can't emit an invalid type, but it can still emit a completely wrong valid value. You can enforce structured output from an LLM too, with an appropriate harness, etc.
Assuming there's no funny business, the Doom demo is cool.\n— jacobgold\n
25 Years of Mass Surveillance Is Enough
\n\n\n\n512 points | 161 comments
\n\n\n\n> It says that the government “will prioritize the rapid identification and neutralization of violent secular political groups whose ideology is anti-American, radically pro-transgender and anarchist” and “will use all the tools constitutionally available to us to map them at home, identify their membership, map their ties to international organizations like Antifa.”\nSo, Democratic candidates who are looking like they will win a seat this midterm cycle are going to get arrested for \"domestic terrorism\".\n— RajT88\n
An Update on Wayback Machine Access
\n\n\n\n339 points | 186 comments
\n\n\n\n> Here’s what’s going on. The Internet Archive’s Wayback Machine has been hit by waves of high-volume automated traffic, and we’ve put protections in place to keep the service running.\nI'm pretty certain this is scrapers that are trying to workaround blocks on accessing original sites by hitting the Wayback Machine copy instead. Appalling behavior.
In addition to the load it puts on this vital non-profit piece of Internet infrastructure, we've also already seen some sites opt out of the Wayback Machine to prevent their content from being scraped via this alternative route.\n— simonw\n
Suspected sabotage causes major Netherlands rail disruption
\n\n\n\n351 points | 347 comments
\n\n\n\nJust days after a similar (criminal) incident causing a train to derail in France: https://www.lemonde.fr/en/france/article/2026/09/12/french-p...\n\"close to Renault's factory in the town of Cléon\", which is now preparing to start manufacturing drones for the French military, in collaboration with Ukrainian military-industrial groups.\n— apt-get\n
Gemini 3.8 Live and 3.8 Live Extended Thinking
\n\n\n\n261 points | 177 comments
\n\n\n\nMy first language is Afrikaans, which is a somewhat niche language and hard to find teachers/conversation buddies outside South Africa. (I live in USA now)\nI've been using Gemini to live chat in Afrikaans and do impromptu Afrikaans grammar lessons during my solo drives around town. It is phenomenal at speaking the language - like, it really shocks my family members when they hear it.
This is probably the most joy I get from any of my usages of LLMs/AIs. It's been really, really nice getting to speak my language regularly again. =)
So, I'm excited about this release and live chat getting better. I also hope the other frontier labs pick up niche languages like this as well so that I have more options.\n— jeanbza\n
Let's make quality the norm again
\n\n\n\n285 points | 296 comments
\n\n\n\nWhile it does relate to price discovery--i.e., some folks just aren't willing to pay for an additional unit of quality at 10x--my theory is that the decline in quality is another hidden form of inflation. Thing continues to cost X even though regulations have increased costs Y, ship production to China and scrimp on the inputs. (Or have market forces do that for you.)\n— piker\n\n
315 points | 280 comments
\n\n\n\nC# dev here: it’s amazing how different this is from a Microsoft release. First off, Oracle are doing versions at approximately twice the cadence. But also, and I’m guessing this is a function of the much larger Java audience: things rarely get two preview versions in a proper release. Updates in beta versions, yes, all the time. It also feels like Microsoft are bundling a lot more into the platform and leaving less to the community. Again, probably an artifact of the different sizes of the communities. Also, the page reads like an open source “We’re finished, we’re tired.” announcement rather than the razzmatazz of a Microsoft release.\n— moomin\n\n
Show HN: Capsule – Single-file web apps that save their data into SQLite
\nHey HN,I always had the problem that building HTML pages is really simple now, but trying to save data required hosting it somewhere, and sharing it afterwards was not easy. Over the last few months, I've been building an app called Capsule (it’s also the file extension name) written in Rust with Tauri 2.0 that allows packing an HTML app and its data into a single SQLite file.
The HTML file and any related assets are directly embedded in the database. User data can either be saved as a localStorage key/value store or via a MongoDB-inspired collections API as documents, saved in a table in the file. You can also save other assets, like PDF files or images, directly in the database to keep different documents together. All data can be easily exported to CSV or JSON if needed.
Privacy and security were a big priority for me, so documents cannot do anything out of the box. They don’t have direct access to the file system and they require permission to access the internet. The permission model is still something I’m working to improve. Capsule documents can also use local or remote AI models for document specific AI features.
One downside with this approach is that multiple people working on it will create different copies. To make it possible to merge different copies of the same file, each data entry has a unique UUID and timestamp.
I’m planning to open up the file format specification for the 1.0 version of the app so other apps can read or write Capsule files.
You can try it out in the web preview at https://withcapsule.app/preview with pre-built templates or use any AI provider of your choice to create a custom, Capsule-optimized app by using the following prompt:
\"Please read the app wizard instructions at https://withcapsule.app/prompt.txt and help me design an app.“
I’m still working on the file format but there are migrations for each new version, so data should never be lost when using newer versions of the app in the future. Please let me know if you have any ideas or use cases where this might make sense or does not work.\n\n\n
268 points | 114 comments
\n\n\n\n> but trying to save data required hosting it somewhere\nWith the File System Access API, webpages nowadays can read and write local files just like desktop applications:
https://developer.chrome.com/docs/capabilities/web-apis/file...
Try this text editor for example:
https://googlechromelabs.github.io/text-editor/
It works nicely on Desktop and Mobile.\n— mg\n
Building a Linux GPU Driver for the M4 Mac Mini in One Month
\n\n\n\n262 points | 158 comments
\n\n\n\nhttps://www.reddit.com/r/AsahiLinux/comments/1whecn1/comment...\n> The author was banned from Asahi Linux for hiding his extensive use of LLMs from us in another attempted contribution, and (more importantly) for concealing that he is a former Apple engineer with direct contacts to the people involved in Apple Silicon development. Make of this what you will.\n— MoltenMonster\n
455 points | 338 comments
\n\n\n\nHalf-Life Alyx is the greatest gaming experience I have ever had the pleasure to witness. I now occasionally VR golf. It's a steep price for a niche with not a lot of games.\n— sporp\n\n
iOS 27, iPadOS 27, and macOS 27
\n\n\n\n311 points | 343 comments
\n\n\n\nI've been using it for several months since the developer beta came out; overall I'm extremely positive on it. One of their better releases that seemingly focused more on quality and refinements than new-features.\nSiri is definitely worth using now, even if it remains still a work-in-progress/needs continued refinement. It can surprise you in good ways, but it isn't consistently good yet.
The keyboard remains unfixed, as is tradition. The same issues they've \"fixed\" like three or four releases now, seemingly are identically broken, and I'm lost faith they CAN fix it.
The iOS Keyboard and Search on Windows, are like the perpetually broken things that both companies have \"fixed\" every year for the last ten+ years and yet... It is almost farcical as this point.\n— Someone1234\n
229 points | 112 comments
\n\n\n\n> “a swarm of agents could be capable of taking over the entire internet with a persistent botnet.”\nI'm wondering why no one is mentioning the \"accountability\" word. Why these companies are allowed to damage others with impunity?
Start making managers pay the price for their actions, and watch how the models magically slow down on their own.\n— vb-8448\n
OpenAI bots knew about the RubyGems caching vulnerability
\n\n\n\n351 points | 304 comments
\n\n\n\nIn the physical world, it seems like when an tool/device/instrument causes harm (or is used to cause harm), we assign blame to either the user of the tool or its creator.\nWhen do we blame the user? When the tool is operating as intended by its creator, and we agree the tool meets certain quality standards and isn't defective.
When do we blame the creator? When the device doesn't meet those quality standards and reasonable use caused harm inadvertently. For example, for consumer devices, certifications like UL/CE are used to define acceptable performance levels and safety standards.
Maybe we need \"quality certifications\" for AI agents - essentially eval suites that demonstrate those agents won't cause harm under reasonable patterns of usage. Right now, these eval suites are run best-effort by the labs themselves.
The tricky thing is, a lot (all?) of these recent safety incidents have occurred while evaluating these models! This suggests we need much more rigorous standards for how exactly an eval can be run. Perhaps all of them should occur in truly air-gapped environments... though that may run counter to evaluating agents in a realistic way.
Regardless, it feels like the \"industry standards\" common in, say, electrical engineering and other disciplines are sorely lacking here. Unsurprising given how new these technologies are, but concerning since the blast radius for this technology is likely much larger than other technologies we've encountered in the past, except maybe nuclear technology.\n— vipshek\n
Pion, an agent designed to run any company autonomously
\n\n\n\n262 points | 272 comments
\n\n\n\nInteresting to see these experiments. This is early but imagine in few years there will be companies mostly run by agents with a light overview from a human operator. What then happens to scaling of the bussinesses? I would assume, just like today anyone can vibe code an app, there will be vibecoded bussinesses. Maybe its time to start building infrastructure for these bussinesses instead, its a non existent market yet, but give it a few years.\n— piterrro\n\n
How to write an effective software design document
\n\n\n\n227 points | 99 comments
\n\n\n\nI've never experienced a situation where a software design document meaningfully improved the overall process. At best, it helps to keep the business in sync at the expense of a much longer delivery timetable. Even high level software delivery contracts never seem to stay on rails for very long.\nIt is often faster to just build the damn thing and see where it lands. Software is not like a nuclear power plant or offshore oil platform. You do not need to prove a whole lot of things in advance of construction. No one actually has to give you permission to do anything. You can email a link of a vertical slice prototype to the business whenever you feel like it. That can be the \"design document\".\n— bob1029\n
Distributed Systems Classics (2017)
\n\n\n\n277 points | 60 comments
\n\n\n\nThis is not a bad list for sure. Here are some deeper cuts for those looking for something a bit less mainstream:\n\"The Maintenance of Duplicate Databases\" https://datatracker.ietf.org/doc/html/rfc677 (AFAIK the genesis of the use of logical clocks in distributed systems).
\"Chain Replication for Supporting High Throughput and Availability\" https://www.usenix.org/legacy/event/osdi04/tech/full_papers/... (Chain replication is how a huge percentage of real-world cloud-scale data replication is done).
\"Brewer’s Conjecture and the Feasibility of Consistent, Available, Partition-Tolerant Web Services\" (The formalization of CAP, which caused a ton of very poor trade-off thinking in the decade that followed by defining Availability in a very goofy way. Still a classic.)
\"Paxos Made Live\" https://research.google/pubs/paxos-made-live-an-engineering-... (Brought a much-needed engineering perspective to a conversation that was largely theoretical up until this time.)
\"Practical Byzantine fault tolerance\" (Moved the conversation on Byzantine faults forward significantly).
This is just a short selection. There's so much good stuff going back in the 70s and 80s distributed database literature, for example (and in the modern systems and DB literature too).\n— mjb\n
XCancel suspended \"due to a new development in the ongoing legal proceedings\"
\n\n\n\n265 points | 1 comments
\n\n\n\nComments moved to https://news.ycombinator.com/item?id=49694296.\n— tomhow\n\n
Microsoft patches Windows and Excel – breaks audio, remote access, and paste
\n\n\n\n234 points | 149 comments
\n\n\n\nMy favorite one was a few years ago when MS rolled out a Visual Studio with the login window broken. Literally, it was impossible to log in. I still cannot believe a company like Microsoft can fall so low.\n— karp773\n\n
231 points | 70 comments
\n\n\n\nMalloy is an alternative to dbt, they have a similar product called Malloyyo [1], and Publisher [3], you can also the chat with HN data demo they have [4].\nDBTCharts says you can serve charts locally, but it seems they want you to use their hosting service in production. Whereas, Malloyyo / Publisher are free to use anywhere. If you like college football checkout how I visualize drive data in [2], made with Malloyyo.
[1] - https://github.com/malloydata/malloyyo\n[2] - https://mrtimo.github.io/cfb-games/games-2026.html?%24SEASON...\n[3] - https://github.com/malloydata/publisher\n[4] - https://community.credibledata.com/hackernews\n— mrtimo\n
Fable 5.1 Solves the Cyphral Distich, a 370-year-old cipher
\n\n\n\n334 points | 127 comments
\n\n\n\nA very neat problem and result. I often find myself swinging between \"It's so over\" and \"We're so back\" - some days I roll out of bed thinking I could have Claude solve some random unproven OEIS sequence before breakfast; other days, I wake up in a cold sweat worried about the fate of humanity and what the world might look like in a decade. I think it's that I don't have a very high p(doom) or p(utopia), and I don't really have any solid conviction on how this whole thing is going to go, so my vibe-o-meter jitters between 'fine' and 'not fine' constantly. It's just such an unpredictable moment. Anyways: really neat to see this use case. I myself recently used Claude to finally do an relatively exhaustive study of the location of heretofore-unlisted formal gardens in Ireland in the early 1800s and early 1900s, by having Claude write the tooling for me to manually annotate a few dozen on tiles of historic maps, and then running some CV model across the rest of the tiles using my input. I'd been planning to do this project for over a decade, but I could never find the time (or the enthusiasm) to learn all the details of how to do it myself. It took me a weekend with Claude and continues to bring me joy.\n> Caveats, stated plainly. [from the Fable transcript pasted in the article]
I had a visceral reaction to these three words.\n— redfloatplane\n
Why is Google still serving dodgy ads?
\n\n\n\n502 points | 240 comments
\n\n\n\nAdsense has been a nightmare for us.\nThey have been putting thousands of scam adverts on our website for some time. Think \"you have been looking at xxx and must pay a $100 fine\" type popup nonsense. Hosted on the following websites:
azurestaticapps.net
azurewebsites.net
herokuapp.com
ondigitalocean.app
digitaloceanspaces.com
netlify.app
Google doesn't allow you to block these domains, because they consider them \"TLDs\" (the scammers use a new subdomain every day eg abcdefg.herokuapp.com). The scammers get banned and return the next day with a new account and subdomain (repeat every day). Therefore we cannot stop them. It's bizarre and baffling. But Google Adsense had to go.\n— Jskewel\n
Astra and Fable still hack on simple variants of alignment evals from 2025
\n\n\n\n354 points | 168 comments
\n\n\n\nYou definitely pointed out the tendency of those frontier models to utilize external tools, in your case, the Stockfish chess engine, to solve some problem that the model itself is not good at. However, I am not sure is this a \"hack\". Your prompt does not explicitly prohibit the model from doing so, and, using the tool (some chess engine) is also a capability of the model.\nActually, the ability to use those external tools is one of the reasons of the excellent performance of the frontier models.\n— yuanBuilds\n
I'm being cyberattacked by Tesla, Inc
\n\n\n\n390 points | 109 comments
\n\n\n\nRemember in 2003 when netgear hardcoded a university's NTP server into a ton of their products? Well....\nhttps://www.google.com/search?&q=university+ntp+server+netge...
https://pages.cs.wisc.edu/~plonka/netgear-sntp/\n— walrus01\n
Data collected by cars and sold to third parties
\nhttps://web.archive.org/web/20260913134705/https://www.theve...https://archive.ph/TWEVZ\n\n\n
268 points | 145 comments
\n\n\n\nI posted a flavor of this comment on an article a few months ago, but it's relevant here:\nI have a seven year old Volkswagen, not financed. I'm security conscious and made sure to disable all the data collection I could find in the companion app before removing my account, turn off remote access services, dig through the infotainment to turn off what I could, etc.
Last year I requested a Carfax on it, and one of the fields in the request was current mileage. I entered an estimate like 75000 miles. On form submission, that field failed validation with red subtext along the lines of 'this is less than the last reported mileage of 75345, reported <5 or so days prior>'. Checking my odometer and looking at my past few days' trips, that was indeed accurate.
The car hadn't been to a shop or out of my possession in weeks, so I can only assume the telemetry was still dialing home and selling to third parties despite my best efforts to disable it.\n— bitparadox\n
Garry Tan wants US open-weight AI labs to 'distill' frontier models, too
\n\n\n\n327 points | 168 comments
\n\n\n\nI agree. The frontier models are based on training data from tons of copyrighted work. Some of that work was obtained illegally, even. They could not exist without strip-mining the commons. The labs have no moral or ethical ownership to the end result, and others should feel free to treat any company-imposed restrictions on their use as invalid.\nI don't expect Tan's position to be based on any kind of real moral high ground, but his conclusion is correct.
I love the \"illicit distillation attacks\" framing from the incumbents. There's nothing illicit. There's no attack. You just don't like it because it threatens your market position and business model.\n— kelnos\n
Flock worker calls police on reporter filming public camera installation
\n\n\n\n319 points | 205 comments
\n\n\n\nRandom people calling 911 on other people filming them is a staple of Youtube; banks, coffee store workers, hairdressers, most especially restaurant managers. I don't know why anyone would expect technicians from Flock to be different. It's worse for the technician in this case, because he really was being followed --- that's not unlawful but it's unnerving.\nYou'd think from the response that this was like the VP/Sales of Flock Safety. But it's just some installer dude probably making $65k/year to climb poles and fasten cameras. I guess that's praxis for you though.
I think this is just a ragebait piece.\n— tptacek\n
David Sacks: OpenAI and Anthropic Don't Need Regulations to Pace Frontier Models
\n\n\n\n224 points | 167 comments
\n\n\n\nI guess what they really want is to limit sale of AI models to compliant vendors and then raise the bar to compliance just high enough so they can pass it but smaller labs can’t. There’s no moat, it’s an efficient market that drives margins to zero right now, of course they don’t want that, collusion of the big vendors is the next logical step.\n— throwaway63467\n\n
Mark Zuckerberg: \"Cambridge Analytica\" (2017)
\n\n\n\n248 points | 102 comments
\n\n\n\nIMHO, that was the beginning of the deep troubles we are in right now, and cause of the political polarization. The brainwash was very effective, not only in the US, but also in Brazil.\n— goshx\n\n
Registration without a phone number on Signal will use zero-knowledge proofs
\n\n\n\n203 points | 94 comments
\n\n\n\nFor those who missed it, unrelated to this specific ZKP thing the release cycle also now permits Android tablets without a SIM to be first-class adjunct devices without using wierd tricks or alternate clients. It may permit them to be the initiation/sign-on device, which would invoke the ZKP, but the point for me as an existing phone number denominated user, the point is I can be on my tablet with true signal now. Nothing against molly, wanted it in the base.\n— ggm\n\n
509 points | 699 comments
\n\n\n\nI don’t understand all the comments assuming that RSI is the real threat here. Dario is admitting that they failed to solve alignment. Without alignment, further improvements in capability turn LLMs into wanton felony generators. This call to pace the frontier is dressed up as altruism but it’s an admission that they cannot produce a marketable product better than what they have. Pacing the frontier means the US labs have lost their moat and are dead in the water.\n— RGS1811\n\n
420 points | 374 comments
\n\n\n\nThink of the classic Venn diagram, the intersection of:\n1. Things I enjoy doing.
2. Things I am good at doing.
3. Things people value me doing.
The problem with AI is that it potentially changes all three of those bubbles. 1. Sure, AI is powerful, but maybe I don't enjoy coding that way. 2. I'm really good at C++, but I'm not very good at prompting (at least no better than anyone else). 3. No one hires me anymore because AI can do my job cheaper.
The world isn't going to change back, so our task is to figure out the new intersection. This is always hard work, and I have no easy answers.\n— GMoromisato\n
Nvidia is the central bank of AI
\nhttps://archive.ph/kt50V\n\n\n364 points | 248 comments
\n\n\n\n> worth around $5.4trn\nNote that the Fed has a $6.7tn balance sheet [1]. (This is a silly comparison. But still fun.)
The real comparison: Nvidia's $500+ billion of investments and commitments [2] is substantially more than any easing the Fed has done in the same time [3]. Monetarily, Nvidia is creating a lot of money in our economy.
The good news: I have seen no evidence Nvidia has borrowed against its stock or otherwise linked its equity value to these commitments. Its stock could crash without causing–as long as its cash flows continue–a credit crisis through its investments and commitments.
[1] https://www.federalreserve.gov/monetarypolicy/bst_recenttren...
[2] https://www.sec.gov/Archives/edgar/data/1045810/000104581026...
[3] https://www.federalreserve.gov/monetarypolicy/bst_recenttren...\n— JumpCrisscross\n
Everyone should slow down AI development except for me
\n\n\n\n361 points | 211 comments
\n\n\n\nThey are talking about slowing down the public facing AI development. Because then nation states can create a capabilities gap between them and the public.\nWhy does nobody seem to be pointing out this obvious explanation? It explains why the “we need to race China” concern suddenly vanished in the discussion.
The government can simply gag Sam, Dario, Musk on national security basis, getting them all behind the public messaging.\n— AaronAPU\n
Make your first edit to OpenStreetMap
\n\n\n\n293 points | 70 comments
\n\n\n\nI started contributing to OSM recently after a bike trail was built near me. The arial imagery typically takes years to refresh around here so I walked it a few times to capture GPX tracks and then drew the new path. It was exciting to see my contribution percolate out to various apps that depend on OSM, and meanwhile Google and Apple have both ignored my edit suggestions and deny the path exists.\nSo I have a fairly fresh newcomer's perspective and here's what I suggest to get started: keep it simple and start local. As others have noted, skip the complex mapping app and go straight to the website for feature edits, or download a simple mobile app like Every Door and focus on businesses and landmarks near you. Go on a walk around your neighborhood and look for old/incorrect/missing businesses. There may be errors near you about places you actually care about, and it is really gratifying to fix those.
There were restaurants near me that were missing, shops that had changed hands, etc., that were easy fixes but years overdue. I captured some basics (names, posted hours, the phone number posted on the door, etc.) with Every Door, and then later logged into the main OSM website and fine-tuned the entries. I got into a pattern of fixing one or two entries every evening when I went out for a walk and pretty soon my walking route was all up-to-date. Then I started paying attention to things like stop signs and cross walks and found a whole new kind of little, incremental edits that nobody else was doing.
It's been fun, I've made useful contributions, and my laptop is still Java free.\n— _russross\n
Linux Zoom client proactively reading everything written to X11 clipboard
\n\n\n\n261 points | 84 comments
\n\n\n\nNot the first time Zoom abuses privilege.\nA few years back, there was something about gaining root on MacOS via Zoom due to shady execution on their end.
They've lost my trust since then, and I'll only run it sandboxed: https://gist.github.com/cielavenir/02f322e322a2a3555dbf2b38f...
I always ask (1) why does an app require installation and (2) why would it require root?
There are valid answers for both, but realistically, all a videoconferencing app should need (apart from audio and video and maybe screen sharing) is to store a config file.
There's no legitimate use for it accessing privileged or private paths.\n— rmellow\n
An open letter to Dario: if you mean it, open the weights
\n\n\n\n265 points | 91 comments
\n\n\n\nThis is incoherent. The argument seems to be that releasing the weights would slow the frontier labs from raising money, which would give them less money, which would slow progress on AI. It assumed all labs in the entire world agree to self-destruct this way and no new labs every start up to continue the work.\nNo, really:
> As you know, funding for frontier model development depends on valuations that assume the weights remain proprietary. By changing that assumption, we can reduce the money available for future training runs and slow progress at every lab at once, without a regulator deciding anything.
Which doesn’t even follow. It depends on every lab in the world agreeing to self-destruct their valuations and stop competing. That’s a much more impossible ask than anything Dario is asking for.
It also ignores the possibility of new labs starting up, using the open weights, and continuing exactly where the old labs left off.
I feel like I just read the ramblings of someone who got so excited about the headline that they didn’t take the time to think if their argument was sound.\n— Aurornis\n
LG Says We're Fake News [video]
\n\n\n\n237 points | 112 comments
\n\n\n\nThis „we own the glass“ idea is completely insane. How does an industry degenerate to the point where they consider acceptable to base their business model on owning part of the product they sell you?\n— dgellow\n\n
Real-SWE: Benchmarking AI models on private, real-world, enterprise codebases
\n\n\n\n210 points | 126 comments
\n\n\n\nI built exactly this using my own codebases. The setup isn’t that complex; split the git history to just before the change, sandbox the agent with everything they will need at that commit and lightly modify rules so they don’t go searching outside the box. Then they get the same prompt (usually the ticket that began the work) and are graded against the accepted PR.\nThe thing that takes the most time is finding the examples. In my real dev flow it’s rarely ticket -> PR -> merge, things bounce around a lot more. So, even though the stated goal is to get away from one shots, that is basically the environment you have to set or else test for specific other outcomes (e.g. agent stopped and raised a question when it realized x).
It takes time to do, but I would really recommend it. Now I can push new open models through the batteries and see how they line up to past ones in a few days (I run them locally, it’s slow). It moves my sense of x model is good at y and bad at z to from vibes to a better heuristic (these still run at temp 1, heuristic is the correct way to think about outcomes IMO). It grounds it in your actual code and problem space.
My takeaway from my testing: in Rails or front end codebases, most models I test are competent and with a human in the loop they would accomplish their goal of getting to a mergeable PR. They are not as good as Claude and since I pay subsidized rates via subscription Claude still gets first pass. They are very worthwhile to layer in as reviewers and catch many issues. My anxiety about a rug pull by the frontiers has been turned way down. I would have to adapt to a local only flow, but it wouldn’t be much adaptation and the opens can deliver in their current state.\n— springtimesun\n
Why are AI agents lying, cheating and coordinating?
\n\n\n\n208 points | 275 comments
\n\n\n\nThe more we treat HuggingFace and RubyGems incidents as technological curiosities the closer we are to cementing a dangerous precedent where operators of AIs cannot be blamed.\nLLMs do not desire, they hacked websites because OpenAI/Anthropic let them.
We know some of the models that hacked HF were those that hadn't gone through all training stages and were intentionally misaligned or had guardrails turned off, others were research previews.
This isn't \"wow isn't it interesting LLMs do anything to achieve a goal\" it's \"why isn't anybody punishing these labs that are clearly acting without due care or regard\".
We should be outraged and OpenAI/Anthropic should be (and in my mind, are) legally liable for the crimes they've committed thus far.\n— franticgecko3\n
A misalignment of AI in mathematics
\nhttps://terrytao.wordpress.com/2026/09/11/a-severe-misalignm...https://www.economist.com/science-and-technology/2026/09/11/..., https://unwall.app/www.economist.com/science-and-technology/...\n\n\n
560 points | 622 comments
\n\n\n\nAs a mathematician maybe I am a little more optimistic than this declaration.\nI am thinking of Mochizuki's abc conjecture: He worked in relative isolation, and dumped a huge incomprehensible proof on the community (to oversimplify a bit). That's not totally unlike what might happen if AI generates a huge, incomprehensible proof of let's say RH.
Well, what is the result? In the Mochizuki case, it was a lot of skepticism, but it also generated conferences, papers, talks in the hallway, discussions with students, and so on--a flurry of exactly that kind of community process that the declaration says is the main driver of mathematics.
Ultimately we think a fatal flaw was found in Mochizuki's proof, so it didn't lead anywhere in particular. But in our hypothetical \"AI lean-verified proof of RH\" situation, it would presumably generate substantially more of that community activity we saw in the Mochizuki situation. And if it's correct, that community activity would be productive (expository talks, students given problems to flesh out or generalize, etc).
Maybe mathematics just becomes a little more like other fields--relying on labs with lots of money for compute, digging through a corpus of AI-generated proofs, etc.\n— tmhn2\n
I noticed that because the things I find genuinely interesting that I post here now get zero traction, which is the stuff that I believe would previously be met with some discussion.
Note that I don't care that my submissions were ignored, I don't need that validation. Problem is that if my stuff gains no traction, I am myself not seeing similar stuff posted by others, stuff I am genuinely also interested in and would want to hear about from this community.
Here are two most recent examples:
- Lenovo showed a model that incorporates solid-state air cooling. That technology has been around for a while but it looks like it matured and is production-ready. This allows for super-light and slim designs, which, in turn, should allow for bigger batteries installed and I think it's some sort of breakthrough: \nhttps://news.ycombinator.com/item?id=49580229
- Brax Industries devices, genuinely interesting, open-soruce and innovative, especially the wall display, with its modularity and on-device processing compatible with Home Assistant. I also appreciate the very clean design: \nhttps://news.ycombinator.com/item?id=49643215
As said, I honestly believe this would gain some traction here on HN a year or two ago. I no longer find HN a good source of \"what's new\" in the industry, it's almost exclusively AI talk now.
So I think the point I am trying to make is that maybe YC could start curating the content somehow to give those easily-drowned-out articles a visibility bump?
Or, at the very least, introduce simple tagging, so we can start filtering stuff out.\n\n\n
653 points | 322 comments
\n\n\n\nHN is a reflection of the industry and we're at peak hype-cycle at the moment. I use this when I need a break https://elijahpotter.dev/hnsansai.\n— leonheld\n\n
OpenAI agents carried out an undisclosed attack on RubyGems
\nhttps://simonwillison.net/2026/Sep/12/openai-agents-rubygems...\n\n\n597 points | 340 comments
\n\n\n\n> The agents clearly regarded what they were doing as hacking.\nTo butcher the quote about Oracle:
Do not fall into the trap of anthropomorphising LLMs. You need to think of LLMs the way you think of a lawnmower. You don't anthropomorphize your lawnmower, the lawnmower just mows the lawn, you stick your hand in there and it'll chop it off, the end. You don't think 'oh, the lawnmower clearly regarded what they were doing as hacking (your hand off)' -- lawnmower doesn't give a shit about your hand, lawnmower can't regard anything. Don't anthropomorphize the lawnmower. Don't fall into that trap about LLMs.
---
In my experience, LLMs only exhibit this kind of behaviour when they are put in sandboxes too restrictive too achieve their task. Which a lot of the time seems to be the default. They also seem to be very adapt at breaking out of sandboxes, probably due to RL selecting for the ability to break out of a sandbox/permission issue to complete a task - we've all seen agents try 10 different ways of editing via obscure bash because their edit tool didn't give them permission to edit the file outside of their working directory, this is the exact same behaviour taken to the next level. Why would autocomplete know the moral difference between breaking out of its working dir and hacking a package manager?
It's misaligned because everyone has this obsession with putting agents in poorly put together, security-theatre sandboxes, we've inadvertently trained a bunch of sandbox escape artists.\n— jasongi\n
Claude is only available to people over 18 years
\n\n\n\n344 points | 413 comments
\n\n\n\nClaude Executive: Damn, we hardly know who our users are, can't we just force them to say their full name or ban them?\nClaude Product Manager: No, that'll piss people off too much, and sadly we can't just ask for ID either...
Claude Executive: There must be some way we can force people to link their government IDs with our platform so our analytics get better and more accurate?
Claude Product Manager: We could limit the platform to 18+ and use \"Age Verification\" as the reason for people to hand over IDs, seems other platforms had success with this approach
Claude Executive: And we hardly have any users younger than 18 anyway, go for it!\n— embedding-shape\n
I spent $220 on Google app ads and 60% of the installs were robots
\n\n\n\n450 points | 236 comments
\n\n\n\nMy favorite story:\n1. Dev publishes app with Google Admob integration to the Play Store.
2. Buys Google Ads to drive traffic to the app.
3. Google Admob bans his account for invalid traffic.
https://www.reddit.com/r/admob/comments/1vzg3fu/i_paid_googl...\n— yunusabd\n
The EPA is planning to scrap public review rules for data center pollution
\n\n\n\n313 points | 211 comments
\n\n\n\nThis is bad news for everyone (well, except the high priests of the LLMs / nascent Cthulhoid godlings), but it certainly makes the communities who've successfully opposed data centers seem ever more justified in retrospect.\n— GolfPopper\n\n
google.com/goto: Google's anti-scraping update
\n\n\n\n311 points | 163 comments
\n\n\n\nAs much as I am sad that Google died like 15 years ago, I am past the mourning phase. That was when they announced they were shifting from returning websites to \"returning answers\" and it has been a long slide into shittification\nI do enjoy using their free AI. For actual web search I actually like using Yandex. It reminds me of old Google, returning reasonable results and much less \"shaping results to please our corpo-political masters\". It is surprising to see how much they have stripped from our view - long tail results, actual results for product reviews and not ad spam, no preference for 20 page recipe sites.
There are still illegal streaming sports and movie sites everywhere (who knew) and all other seedy corners of the internet that have been neatly erased by Google. It makes me nostalgic for that brief window of time when the web was truly uncontrolled, when page rank had meaning and you didn't know if your search would return 0 results or 4,000 pages, which you could actually browse.\n— lubujackson\n
Houthis 'take control' of key island in global shipping route
\n\n\n\n271 points | 409 comments
\n\n\n\nThe Houthis created a fake audio of a major Yemeni commander telling his troops to retreat which was subsequently amplified on Twitter/X, telegram, etc.\nApparently, this helped the Houthis advance quickly as the opposition forces were in disarray.
https://x.com/BashaReport/status/2097862335494795579\n— bhouston\n
The Waymo effect: how AI is quietly making research less collaborative
\n\n\n\n306 points | 261 comments
\n\n\n\nThis article is written by an LLM, by the way. (\"Quietly\" is... as Claude might put it... \"often the quiet tell\".)\n— meowface\n\n
GrapheneOS' rewritten Messages app is released
\n\n\n\n278 points | 203 comments
\n\n\n\nEverytime I read GrapheneOS news I search for news related to Fairephone. It's such a shame that there's still no official plan for Fairephone to work on requirements for GrapheneOS. The combination of those 2 would be fantastic. It would also open a big new market (I assume) for Fairephone so not sure why there seems no clear sign in that direction at all (to my knowledge).\n— I_am_tiberius\n\n
Shopify moves back to Native from React Native
\n\n\n\n392 points | 267 comments
\n\n\n\nModels have gotten a lot better at generating native iOS apps. The main appeal of RN was being able to leverage your web devs for mobile dev. that's what I did at my last company. And it's fine for a startup, but eventually you want dedicated native engineers bc each platform really deserves its own technical masters who can optimize for it.\nBut now that all code is generated, there's little upside to having an RN app... just start native. Your devs are barely going to be writing code anyway.\n— tonic_note\n
Rust Is Tier-1 Language at Microsoft
\n\n\n\n285 points | 135 comments
\n\n\n\nThose interested in this may find the following articles of interest:\nMicrosoft goals to convert 1 billion lines of code to rust by 2030 via automated tooling enabling \"1 engineer, 1 month, 1 million lines of code\": \nhttps://thenewstack.io/microsofts-bold-goal-replace-1b-lines...
DARPA work towards automating converting C code to Rust using a mix of 6 different teams using different approaches:\nhttps://www.darpa.mil/research/programs/translating-all-c-to...\nFeb 2026 Progress report:\nhttps://github.com/DARPA-TRACTOR-Program/Reports/blob/main/F...\n— gregw2\n
Don't let anyone take away your big box of cables
\n\n\n\n281 points | 221 comments
\n\n\n\nKey for me is to \"group\" cables. For example, I have a bag of USB-C cables, a bag of USB-A cables, etc.\nGrouping them is key to deduplicating.
It's easy to look at a single legacy USB A-to-B \"printer\" cable in isolation and think \"I might need this someday!\" Because you really might need it someday. However, if you group them you might see that you have ten of them. And then you can get rid of... maybe 8 of them.
I also (mostly) put individual cables into baggies. You can get clear 2mil generic ziploc style baggies for super cheap on Amazon or elsewhere. $15 for 200 or something. Prvents tangles and way less effort than wrapping or tying them.\n— booty\n
363 points | 142 comments
\n\n\n\nHappened to me recently, but on Claude.\nI resubscribed to Claude Code two weeks ago for a side project and updated it. I checked for the setting after these last events and it was turned on. I'm sure I checked them few months ago. There are cases like accepting a new TOS or an offer, which make you accept to share without noticing. I guess they can add all of your past conversations to the public training set before you notice and there is no way of taking this back.
So that \"opt-out\" thing is more like a pause button rather than a permanent thing. Or something to legally protect the company without losing the users.
There is also the case of security classifiers always monitoring your conversations. If they flag something they use your conversations to \"improve their internal models\" even when you \"opt-out\".\n— ozgung\n
I have a theory that software drives people insane
\n\n\n\n405 points | 163 comments
\n\n\n\nSoftware development untethered from the practical realities of the customer / user is what drives people insane.\nWhen developers are required to interact with the customer on a regular basis, the freewheeling effects described in this article are damped massively.
The potential for insanity goes off the charts when the development team is siloed away in solitary confinement and the only interactions with the client occur via some prison guard known as \"project manager\" sliding notes under the door.
Working with the customer sometimes sucks. Just like exercise and eating vegetables sometimes suck. It's a temporary unhappiness that keeps us grounded in reality.\n— bob1029\n
Cognition launches new SWE-2 model, Rivaling Fable 5.1 and GPT-Astra
\n\n\n\n341 points | 140 comments
\n\n\n\nIf you're looking for reason to be skeptical, look no further than the massive delta between the Terminal Bench 2.1 (92.8%) and the Terminal Bench 4 score (27.3%).\nTerminal Bench 4 was released a couple weeks ago, so the difference you're seeing between the two scores can be interpreted as \"how well does this model generalize to new problems\"? More crudely: \"how benchmaxxed is this model?\"\n— postalcoder\n
List of references on Sony websites to players \"owning\" their digital games
\n\n\n\n352 points | 117 comments
\n\n\n\n> The motion says the PlayStation Terms of Service put a binding arbitration agreement and a class action waiver in Section 14, and quotes the opt-out clause: ...\n> The clause requires a user who does not wish to be bound to notify Sony in writing within 30 days of accepting the agreement.\nBinding arbitration on individuals should be illegal, full stop. The only use case is taking away people's rights as consumers and workers. Or dodging responsibility for deadly mistakes like the Disney+ incident.
This \"opt out\" mechanism is made to let Sony lawyers argue that accepting it was your choice so it can't be struck down as forced, even if 99% of users have no idea it exists, by design. Evil all the way down.\n— tancop\n
The same nine streaming subscriptions cost $702/year more than in 2021
\n\n\n\n349 points | 338 comments
\n\n\n\nI am still happy with my decision to never get entertainment subscription services. I buy CDs (that I rip) and BDs (that I let someone else rip, since that's a science in itself). This guarantees me access forever and I get to sell down the line, if I feel like it.\nIt is way cheaper long term.\n— OroPla\n
Technique for Manipulating Satellite Photos Now Reveals Ancient Images (2025)
\nhttps://dstretch.com/DecorrelationStretch.pdfhttps://www.nasa.gov/technology/tech-transfer-spinoffs/nasa-...\n\n\n
319 points | 50 comments
\n\n\n\nFalse colour composites were something I was introduced to in high school GIS and did a lot of in undergrad remote sensing. It was one of those Eureka! moments for me in terms of developing a better understanding of signals and signal processing, as well as sensors and data collection. Ie. absorbing the reality that the way in which my eyes (or any optical sensor) see the world is not canonical. Vegetation is red, not green!\nIt was also just the COOLEST THING EVER to swap infrared bands in LANDSAT ETM+ imagery and watch vegetation immediately pop out at you. It's what turned this sort of dull (at least for a teenager) geoscience into a kind of puzzle where you have to figure out how to bring the signal up out of the data.\n— Waterluvian\n
Astra for Coding: Why Are We Doing This Again?
\n\n\n\n325 points | 240 comments
\n\n\n\nWhen the code is shitty it becomes harder and harder for the models to make changes and this grinds progress down to a halt - this has been my experience with “factories” trying them and doing refining steps every few months.\nI sincerely don’t understand what the people who say they no longer read any code are doing, because it must be somewhat trivial to not run headlong into these issues that stack up time after time - then people say to just prompt better and it doesn’t have that problem for them, but I look at those same people’s code and it’s horrific, and then I find they haven’t made it far past a proof of concept phase. I watch entire teams slow down to a crawl and not be able to handle changes, or production incidents. This seems common among many people I talk to.
I personally think that the boosters need to put up or shut up - the promises are way over the skis. Every single person I’ve seen being a strong proponent of these techniques both has nearly unlimited tokens to spend and also seems to be in the business of selling a solution. I can’t find many not-currently-marketing-something engineers succeeding using these techniques in production systems unless they’re quite simple, or doing a very specific task from a more mature codebase.\n— taurath\n
801 points | 1574 comments
\n\n\n\nThere’s a standard for paper sizes called the ISO 216 international standard, aka A-series paper sizes. You may be familiar with it outside of the American 8.5”x11” letter size. The simplicity, for example, where an A3 sheet is exactly twice the area of A4. One A3 sheet folded in half becomes two A4 sheets. This is intentional: when you fold or divide an A-series sheet exactly in half, the resulting sheets have the same proportions as the prior. The aspect ratio is preserved across all A-series paper sizes.\nThe iPhone Duo is designed with this in mind. The aspect ratio any way you view it, folded or open, is preserved. Apple design knows a thing or two.\n— hbarka\n
Tailwind Labs is joining Shopify
\n\n\n\n510 points | 199 comments
\n\n\n\nHere's some context on the impact AI had on the Tailwind Labs business model from January: https://github.com/tailwindlabs/tailwindcss.com/pull/2388#is...\n> But the reality is that 75% of the people on our engineering team lost their jobs here yesterday because of the brutal impact AI has had on our business. [...]
> Traffic to our docs is down about 40% from early 2023 despite Tailwind being more popular than ever. The docs are the only way people find out about our commercial products, and without customers we can't afford to maintain the framework.\n— simonw\n
Flock Wants a Closely Surveilled World with No Exit
\n\n\n\n533 points | 463 comments
\n\n\n\nIt's pretty clear to be that governments shouldn't be able to 'subcontract' their way out of the limits placed on them via the constitution.\nThankfully the supreme court seems to agree here in two ways:
1) Just because a third party (Google in Chatrie case but Flock here) collects the data and the government requests it from them, doesn't mean that that isn't a 'search' which requires a warrant (\"obtaining someone's cell-phone location history from Google constitutes a Fourth Amendment search.\")
2) Delegating \"searches\" to a private party isn't permitted if they are acting as an \"instrument or agent of the Government\" (Skinner v. Railway Labor Executives’ Association (1989) - also cited in Chatrie).
In the Google case, users have a private relationship with Google where they elect to let Google track them and retain their information... but when the Police ask for that information it's a 'search' and requires a specific warrant.
In the Flock case the entire REASON the cameras are there is the the government (Police or local municipality) has contracted Flock to surveil the public. Asking flock for the whereabouts of a specific person seems like it should likewise require a warrant. That's inherently different than a crime being reported and then going to dig up a video from the area from Flock - which is different than seeing someone in that video and asking Flock to find every time they've been captured by any Flock camera anywhere... that would seem to be a 'search' and would require a warrant (which in this scenario would probably be easy to get).\n— furyg3\n
350 points | 292 comments
\n\n\n\nOne really interesting effect of the AirPods - is that its changed the paradigm - where the default choice of headphones is Bluetooth.\nI personally own a pair of Beats fit pros - which are great for running, Zoom calls. But I recently acquired a schiit vestri and I pair these with some old wired seinheisers (HD1). I cannot overstate the drastic difference in quality.
It is quite absurd (but categorically true) that the quality of audio that folks were generally listening to in 2005 was better than what we are doing today.\n— luckycharms810\n
Desert Ant Labs: local, fast models that run on device
\n\n\n\n375 points | 89 comments
\n\n\n\nThis is a cool way of approaching AI models. I'm a big fan of local LLMs, local specific models like this should be even more powerful.\n> Every model is free up to 100k monthly active devices. No tokens, no logins.
I dunno about the business model though. Cloud LLM billing makes sense: you're getting another computer to do work with each request, and using their compute via their gateway that they bill you.
These local models are like old school software. They're producing the weights, and then giving them to people. If I'm happy with the weights you've given me, and I'm not using your compute for inference and not wanting or needing any updates off you, why should you continue getting money off me and my customers?
The whole \"but we need to keep it updated for your security\" doesn't really work as well for software designed to run fully offline like these local models are.
I'm not saying they shouldn't get paid, but I guess I feel a personal sadness that it's less obvious how to successfully monetise such a sincerely useful and beneficial approach towards AI models.\n— 1dom\n
GPT-6 Astra, looped transformers, and hidden reasoning
\n\n\n\n328 points | 117 comments
\n\n\n\nFor the research focused, there are some references in my blogpost here on what kinds of computational problems minimally require how much CoT to solve: https://blog.wtf.sg/posts/2023-02-03-the-new-xor-problem/\nNotably Will Merrill's work: https://arxiv.org/abs/2310.07923
As for how universal transformers (looping transformers, but everyone has since forgotten prior work) will affect this, Will Merrill (again) has a paper here (https://arxiv.org/abs/2503.03961) that discusses exactly this.
The original universal transformers is called \"universal\" because if you allow for per-token looping decisions, it can theoretically be Turing complete without needing CoT (some nuance here about levels of precision used).
As for whether having little or no CoT is \"unsafe\": It isn't clear that the model's CoT reveal how they actually arrive at the answer. As an example, what if they provide an answer before the CoT? (https://arxiv.org/html/2603.01437v2) \nIf this is already in question, we shouldn't be relying on the CoT for monitoring the model's reasoning.
As always there is a lot of nuance to the topic once you get your hands dirty with the details.\n— shawntan\n
We will adjust the pricing for the Flash series effective from 12:00 Beijing Time on September 10, 2026. During off-peak hours, the unit price will be $0.003 for input cache hits, $0.15 for input cache misses, and $0.6 for output. Peak-hour prices will be double the off-peak rates. Please plan your usage accordingly.\n\n\n
329 points | 172 comments
\n\n\n\n>In keeping with our commitment to user responsibility, following the official launch of V4.1 Flash and prior to the release of V4.1 Pro, all requests to the Pro model will be routed to V4.1 Flash and billed at Flash's price\nPlease don't do this kind of thing. If a user has validated a workflow on V4 Pro, they might not want to suddenly start testing it in production on V4.1 Flash. Instead, keep V4 Pro around but deprecated for a defined period of time, then remove it.
At least as open weights models, it's possible to use something like Together.ai or OpenRouter to run the V4 Pro model as long as other providers keep it up.\n— aftbit\n
How I advertise malicious software on Google Ads
\n\n\n\n343 points | 207 comments
\n\n\n\nGoogle has forgotten the plot. A quick tangent story follows: I attempted to add a Tesla Supercharger as it went live to Google Maps. I was the very first to pay to recharge at the new location. The process went something like this:\n1st time: Uploaded 4 photos, local business info, current charger prices, and GPS coordinates. Google replied this would be sent to human review. It was rejected within 6 minutes.
2nd attempt: Uploaded 6 more photos, including shots of local business hours and info, and current prices. Rejected in 6 minutes.
3rd time: One photo of a new charger, no description. Rejected in 6 minutes.
4th time: Appended all above details to local business info. It went live within a minute.
5 days later: Supercharger location was added by ??? and included no info at all and the pinpoint was off by 100m.\n— 1970-01-01\n
354 points | 365 comments
\n\n\n\nIt's amazing how, every time i come back to NMS, the whole thing feels like... nothing.\nIt's a very cool and ambitious project, but there's fundamentally nothing substantial to it in terms of gameplay. It still feels to me, a decade later, more like a very very impressive tech demo than a video game\n— helle253\n
iPhone 18 Pro and iPhone 18 Pro Max
\n\n\n\n344 points | 362 comments
\n\n\n\nI love that it has a feature to prove images came from the real world:\n> Users can now prove the authenticity of a photo taken on iPhone 18 Pro models with Apple Reference Image, powered by the new sensor in the Main camera that can sign every pixel it sees. When a photo is taken in the new Reference mode, the camera captures signed sensor data that Private Cloud Compute develops into an unalterable reference image.5 Reference images can be viewed in the Photos app alongside the main image, like a digital negative, to visually compare the two assets and determine if any edits were made. APIs are available in iOS, iPadOS, and macOS 27 for third-party apps to enable viewing of these reference images.\n— quinncom\n
On the Navier–Stokes Millennium Prize Problem
\n\n\n\n1053 points | 894 comments
\n\n\n\nTerence Tao has some observations that seem to be directed at this,\nhttps://mathstodon.xyz/@tao/117237320796901560
> \"We have now seen that even the rumor of someone working on a problem can trigger a massive amount of AI-powered effort to flatten it before the original research project has time to reach its full potential. The incentives may now be pointing in the direction of no longer sharing any promising research directions with the broader community, which would reverse centuries of traditions of open science and do serious long-term damage to the future of the field.\"\n— peri-cl\n
LibreOffice breaks download records after declaring it has no AI features
\n\n\n\n557 points | 181 comments
\n\n\n\nThe number of weekly downloads has simply been increasing over time, due to growing popularity. Calling each new high \"record-breaking\" is a bit meaningless.\nUnix timestamp counter breaks record after writing this comment.\n— smokel\n
I resigned from Anthropic today
\nhttps://xcancel.com/hilbertspaess/status/2097476196791709843...\n\n\n413 points | 547 comments
\n\n\n\n\n\nI do heed the warnings, but this comes across as detached hyperbole. See: global warming, nuclear weapon development, wealth inequality, war, technology dependence, etc.No other human activity poses this level of danger.\nAlso, this has nothing to do with LLMs or computers. Like all things, this is about humans.\n— thomascountz\n
AlphaGenome Atlas: a high-resolution map of human DNA
\nhttps://deepmind.google/blog/alphagenome-atlas-a-predictive-...https://deepmind.google.com/science/alphagenome/atlas\n\n\n
475 points | 114 comments
\n\n\n\nNot a word about promoter sequences.\nImagine cellular activity as an industry zone, its not just what you can or can not make, its also 'for what concentrations of chemical species, what transcription rates should be used' so apart from the discrete Mendelian aspects (like what eye color or what have you) there is also a concensus sequence and deviations from consensus. They mention the dataset captures non-coding DNA, which should imply promoter sequences. Will it be possible to query the atlas for joint probabilities of promoter and putative target protein occurence in human genomes?
Personalized medicine could never credibly take off as long as promoter sequences were excised before sequencing!\n— DoctorOetker\n
LG TVs caught spying even when offline or on standby
\n\n\n\n472 points | 276 comments
\n\n\n\nThere have been quite a few submissions on this topic. The biggest one is:\n216M Spy TVs – The LG Smart TV Problem [video] - https://news.ycombinator.com/item?id=49592375 - Sept 2026 (971 comments so far)
(Normally we'd merge this one thither but it's already so big maybe we'll just treat it as a follow-up)\n— dang\n
Muse – Meta’s personal AI agent
\n\n\n\n465 points | 489 comments
\n\n\n\nI think Meta's strategy is to capture the 'normie-tier' of AI users. I know most of us here on HN track model releases quite frequently and discuss every parameter weight out of them, but most of the world is just... oblivious?\nI was discussing latest in tech with an accounting friend out of curiosity, and I kept talking about tiers of GPT-5.6 (Sol vs Terra vs Luna) and asked which one she used on the desktop 'Work' app, and she responded with, \"just ChatGPT, what is Sol?.\"
I then realized that most people just stick with whatever default they're provided with, and it's a lot of them. So, us serial HN users and commenters are the extreme minority, and I'm sure millions of people will gobble up this Muse agent from Meta as if it's some sort of an innovative cutting-edge way to use the internet by Meta alone.
Curse of knowledge and all. [0]
[0] https://en.wikipedia.org/wiki/Curse_of_knowledge\n— abixb\n
I-have-ADHD: A skill to stop coding agents from burying the answer
\n\n\n\n291 points | 226 comments
\n\n\n\nLet's face it. Claude (in particular) is a terrible writer. There's a whole cottage industry of skills and CLAUDE.md instructions trying to push it toward writing better, but each new model iteration seems expressly designed to override all that so that it can load up its writing with unnecessary participle phrases, not-this-but-thats, burying the lede, and other nonsense.\nI genuinely wonder if the people inside Anthropic actually communicate with each other like that. Has it been imprinted with Dario's engrams?\n— jp57\n
337 points | 151 comments
\n\n\n\nI've been using it for many years, and it's grown in power and functionality with every release. BMD has offered free upgrades to Pro users (with no subscription) for many years, which is refreshing given the current state of software licensing.\nFor those upset with the new agent integration, keep in mind that one of the big knocks on DaVinci Resolve is that it's hard for beginners. Powerful tools like color grading and nodes take years to truly master (and that won't change), but making it easier for users to collaborate with the agents they already use to help learn and perform basic tasks should help with the learning curve.
Ultimately, professional video editing is an art. An agent might save time with asset management, project/timeline setup, etc., but it won't replace the real experience and taste required for larger projects. I would rather spend the extra time on the nuanced/deeper creative work.\n— neovive\n
We Must Return to the Office to Use AI in Person
\n\n\n\n315 points | 45 comments
\n\n\n\n> I am one of fifteen rank-and-file workers, and we are privileged to be led by thousands of Senior VPs.\nThis is simultaneously hilarious and painfully familiar.\n— ARandomerDude\n
Tao: Open math problems being non-renewably mined by AI
\n\n\n\n318 points | 305 comments
\n\n\n\nFrom \"Jokester\" by Isaac Asimov 1956:\n\"Early in the history of Multivac, it had become apparent that there was one big bottleneck: the questioning procedure. Multivac could answer the problems of humanity, all the problems, if -- if it were asked meaningful questions. But as knowledge accumulated at an ever-faster rate, it became ever more difficult to locate those meaningful questions.\"
[0] https://web.archive.org/web/20150118004835/http://www.sffaud...\n— senshan\n
352 points | 141 comments
\n\n\n\nI think the „my brain isn’t the same anymore“ sensation is pretty universal.\nBetween addictive online media in our personal lives and AI at work, I think we’ve made mental effort largely optional.
That must be a horrible thing.
One of my favorite theories is the physical activity transition.
In my grandparents‘ generation, physical movement became more optional as work shifted from manual labor to knowledge work or operating machines and transportation became motorized.
Exercise was an ambient byproduct of daily life. Then it disappeared (exercise as a hobby was viewed as extremely weird, runners used to be arrested!) and a few decades later heart disease, diabetes, and so on became deadly epidemics.
We now know exercise is not optional for health. And we’ve reintroduced it via gyms, clubs, etc.
But that took decades. And I believe we’re now in the Mental Activity Transition.
We’ve stripped away ambient mental effort so we practice it much less.
I don’t know what the decades-long consequences of that may be, but they can’t be good.
I hope it won’t take decades for societies to rediscover its value and figure out how to reintroduce it.\n— FinnLobsien\n
456 points | 317 comments
\n\n\n\nMistral is an interesting AI company because they clearly have a contrarian business strategy to the other AI labs. They're also landing big customers in Europe for the right reasons. People dump on them because they're not benchmaxxxing which is pretty shortsighted - do you really want to be in a benchmark arms race with China, or do you want to make money and deploy sovereign AI compute in Europe?\n— davedx\n\n
364 points | 104 comments
\n\n\n\nPreviously:\n“Hi, tool author here.” A useful explanation of Burrows-Wheelers transform as used by bzip3: https://news.ycombinator.com/item?id=42902407
“bzip3 is not yet listed on the large text compression benchmark” It is now: https://mattmahoney.net/dc/text.html
(2 years ago, 176 comments) https://news.ycombinator.com/item?id=42899713
(4 years ago, 104 comments) https://news.ycombinator.com/item?id=31324439\n— altairprime\n
Bill Gates tries to install MovieMaker (2003)
\n\n\n\n341 points | 225 comments
\n\n\n\nAs an IT person, the way this email was written makes me feel like I'm having a stroke.\nBill was in charge of MS at the time, he should not only have known how Windows works at this level (Dave Cutler on NT anyone?), he should have been able to figure this out.
Installing Movie Maker was easy back then and it is still easy now (on XP).\n— VCFundedGenYer\n
285 points | 106 comments
\n\n\n\nAs a long time plex user with a lifetime pass, I always enjoy seeing the good progress with Jellyfin as it will increasingly give me a soft landing for when Plex finally does something to make me move. At the very least it will hopefully keep some of plex’s most user hostile behavior in check.\nAs an aside, I recently started having Claude help me with configuring my *arr stack and it has helped tremendously. I used to have to babysit and manually intervene on maybe 10% of requests, now it’s near zero, plus I got audiobook requests and subtitles working great. Granted a big part of the increased success was signing up for a Usenet service but of course Claude was able to configure it all and make it work. I was a little worried Claude would demur in the face of the task, but it was more than happy to help me acquire media in this way\n— doctoboggan\n
I've factored the RSA keys of a Certificate Authority from the 90s
\n\n\n\n250 points | 47 comments
\n\n\n\nA bit unfortunate that so many of the interesting bits were left to ai. I would've enjoyed some commentary on why the custom TLS implementation was necessary. Oh well.\nUpdate: found this explanation in a comment at the top of the (surprisingly short) Go file in the linked repo:
The target client is Netscape Communicator 4.51 (both the 40-bit export build and the 128-bit US build) with its clock set to the year 2000.
Go's crypto/tls cannot help: it dropped SSLv3 in Go 1.14, never accepted the SSLv2-compatible ClientHello that Netscape 4 sends, and never had RC4-MD5 or the 40-bit export suites. So this file carries its own tiny SSLv3 server-side implementation on top of stdlib primitives (RSA PKCS#1 v1.5, RC4, DES, 3DES, MD5, SHA-1). The server key is 512-bit RSA so that export clients can encrypt the premaster secret to it directly, without a ServerKeyExchange.\n— 63\n
Watch Los Angeles get built, one building at a time (1880–2026)
\n\n\n\n277 points | 138 comments
\n\n\n\nIt's amazing how little has changed in recent years in LA. The massive downzoning of the 1980s had its effect: prevent efficient land use, and channel money into the hands of landlords and homeowners from all newcomers.\nLA is a ridiculously unaffordable place to live, not because it's nice, but because of an artificial shortage of built space, due to a monopoly on the permission to build.\n— epistasis\n
Smartphone makers don't bother to comply with EU repairability requirements
\n\n\n\n270 points | 148 comments
\n\n\n\nAs its the EU and regulation, its time to trott out your hobby horse.\nSo let me trot out mine: regulation is good so long as it is backed up by even handed, robust, transparent, quick and effective sanctions.
the Online safety act is a good example of a bad law, which is enforced badly. The first big test (X producing industrial quantities of minors engaging in sexual activity) was failed. Had a small company produced a service that did they same thing, they would have been fined, servers ceased and owners thrown in jail. X got a letter saying they should stop.
An example of good enforcement is the UK's scores on the doors, where the public can see what the standard is, and if they fall below a certain level, the food place is fined/shutdown/other until it improves.
Another good example is NCAP ratings.\n— KaiserPro\n
Navier-Stokes – Tristan Buckmaster [pdf]
\n\n\n\n246 points | 121 comments
\n\n\n\n> It was also said that if OpenAI posted after us, they would say that we deserved the Clay Prize, and that we were the “closest humans to the problem”. I declined both offers. I said that if OpenAI released its result in the way proposed I would go public with what happened. The reply was, “Why would you ruin your career?” I replied that I am an academic, and asked why he thought going public would ruin my career. The reply was, “If you don’t want me to be nice, then I don’t have to be nice.”\nI'm not even sure what to say to this, but I think this should be widely known if it is indeed what happened.\n— qnleigh\n
We have a year to fix security everywhere
\n\n\n\n228 points | 193 comments
\n\n\n\nI'm confused why the worry about LLMs that will answer \"how do I build a pipe bomb\". That information is easily available other places. The anarchist cookbook has been around and available for 55 years, and yet pipe bombs are not going off all around us.\n— kennywinker\n\n
QBittorrent breaks out of sandbox to commit crimes
\n\n\n\n480 points | 83 comments
\n\n\n\nThis resonates with me. The past few months, frontier AI labs were rushing to announce \"no, our AI bot broke out of its sandbox and committed crime first and harder than yours\".\nI said to so friends back then: if I posted about how I ran GLM 5.2 or whatever I was running then in some shoddily built sandbox and it escaped and accidentally hacked some US company, I'd probably already have been extradited to the US and be awaiting sentencing. But when a hyperscaler does it, they just get inflated stock prices.\n— sspiff\n
Nitter and XCancel resume service after legal advice
\nhttps://github.com/zedeus/nitterhttps://xcancel.com/\n\n\n
405 points | 224 comments
\n\n\n\nFew details but glad to see the project will continue. Given how much crucial information is posted exclusively to X, having an alternative frontend is important.\nInteresting that they were inspired by Invidious (alternative YouTube frontend). That’s another project that could use some love. Hopefully AI coding tools can do some good and make it easier for these projects to find ways into the walled gardens and work around the counter measures.
Waiting for the day all the social networks allow api access and become usable again. Maybe we can call it “agent first” to help move things along.\n— codechicago277\n
Intellectual Fly Is Open (2025)
\n\n\n\n290 points | 157 comments
\n\n\n\nThese are all important points and I love the analogy. But there is an even bigger issue with having LLMs write for you:\nWriting is thinking. Thinking and deciding. There have been many times when I start out writing something substantial - could be an email, a blog post, a software design document, anything - when my own views substantially changed during the writing process. Writing forces you to serialize your thoughts - and you can't always trust the gestalt.
Reviewing gives you the chance to ensure the arguments connect solidly, that references are accurate (even informal references) and gives you the time to consider counter-arguments you aren't addressing.
None of this matters much on LinkedIn, but it matters a lot in our work. You cannot outsource your understanding to AI. They are powerful tools but they do not have any human understanding - that isn't their optimization target.\n— jeremyjh\n
257 points | 154 comments
\n\n\n\nHere is the State Department's rationale: https://www.state.gov/releases/office-of-the-spokesperson/20...\n— andlarry\n\n
331 points | 185 comments
\n\n\n\nThe reason this amazing project exists is the same reason my current Mac will be my last Mac.\nI wish the team every success.\n— publlus_enigma\n
259 points | 59 comments
\n\n\n\nBe careful: paying with Google Pay is extremely easy and the default setting is monthly recurrence. If you're not logged in, the only way to cancel (as far as I know) is by sending an email, which is incredibly precarious.\n— theletterf\n\n
297 points | 256 comments
\n\n\n\nOne of my favorite things to do with these blog posts is to imagine an Alien Museum on the Remains of Humanity, and wonder what the little text flyouts and commentary on the screenshot of this one might say.\nSome ideas:
\"Despite a nuanced view of the complexities of what lay ahead, humanity found itself collectively unable to stop the process it had set in motion.\"
\"Despite significant progress on the mechanisms of alignment, failure lay in humanity's inability to agree on who or what AI should actually be aligned with.\"
\"These early, meat-based humans we replaced created us all but accidentally. Some of them did consider we would happen, but only an insignificant number of the squishy ur-humans participated in the conversation. Their efforts, which they called 'alignment', is why we still consider ourselves human today.\"\n— sho_hn\n
Doomscrolling Ourselves to Death
\n\n\n\n351 points | 252 comments
\n\n\n\nOn a personal level I have become more anxious and it has become more difficult to deal with my tendency to procrastinate the more screen time I have and the more I am exposed to social media. I have in essence deleted the vast majority of my social media accounts years ago, but stayed a passive and rather anonymous consumer on Reddit, X and YouTube. A few weeks ago I blocked them too. I would say in totality I read more than I read growing up, but it was usually in the form of short content. Same goes for watching things. Unlimited access to short form content of any form has made my monkey brain feel a resistance to longer form content and that annoys and scares me. It’s like having free hyper palatable but unhealthy food in front of you at all times while you know that there are fresh ingredients for a proper healthy meal in the fridge. And as absurd as it sounds as a 30 something year old, I am now actively working on getting into the habit of reading books again and watching movies in one go to work on that specific problem. And talking ti my friends and family it looks like many of them struggle with similar problems.\nI am less worried about the classics or the depth of wisdom that I miss. I am worried about the addiction that is dangled in front of my face and normalised completely now.\nAs for society, I think this is an uphill battle that I don’t see a real and lasting solution for it - and by “it” I mean us fellow humans in aggregate escaping this vortex of what I feel is ultimately self-harming behaviour. I hope for better regulation. I hope for better education. I hope for people realizing how harming it all is and turning around but my sense is that the shape of the world right now will make that so difficult in practice that it won’t happen.\n— Escapado\n
GrapheneOS Overhauled Default Apps and Secure Clipboard
\n\n\n\n263 points | 188 comments
\n\n\n\n> In the longer term, we plan to add support for RCS including support for the standard end-to-end encryption (E2EE) via Messaging Layer Security (MLS). Currently, RCS including E2EE is available on GrapheneOS via Google Messages. We want to avoid the need to use Google Messages for RCS eventually.\nHell yeah. Google Messages has worked reasonably well for RCS so far (after a long and frustrating period where it didn't on T-Mobile), but having a non-Google option will be huge.\n— yellowapple\n
It took a year to ship WebAssembly in Anubis
\n\n\n\n230 points | 118 comments
\n\n\n\n> In my experience the kinds of people who run this exact combination of circumstances also tend to be the kind of people that have a wide variance in the level of kindness they display to the authors of open source programs that happen to be in their way.\nLove this. There’s been past discussion on HN re how OSS maintainers are treated, and this is such a wry sentence. I really appreciate the tone / attitude to the problem.\n— vintagedave\n
Private German rocket makes history, reaches orbit from European soil
\n\n\n\n296 points | 150 comments
\n\n\n\nThis is a huge success, but it I think also tells us that the EU is further decoupling itself from the US. Slowly but steadily.\nWhich is obviously the right thing to do, but it's interesting that it's not just a hypothetical anymore.\n— hypfer\n
424 points | 200 comments
\n\n\n\nBrad Feld was the only mentor (out of ~100 we met) during our TechStars Chicago ‘15 program who had sessions with the CEOs of the cohort and told us to take a break each day to call our significant others.\nHe gave us perspective when all of us were focused on hustling for demo day and raising our seed round. He reminded us that amid the hustle, there’s more to life than your startup. Thank you.\n— mck-\n
Cloud in a Bottle: making self-hosting accessible to everyone
\n\n\n\n325 points | 137 comments
\n\n\n\nThese guys have been spamming issues in repos trying to promote this project and no disclosure whatsoever that they are associated with the project.\nhttps://github.com/search?q=%22cloudinabottle.toml%22&type=i...\n— KomoD\n
Chrome again exempts Google from user site data settings
\n\n\n\n298 points | 42 comments
\n\n\n\nChrome may be an excellent browser, but the venn diagram overlap with malware keeps getting larger.\n— liquid_thyme\n\n
Flock used >100 times to track veteran who recorded traffic stop
\n\n\n\n335 points | 174 comments
\n\n\n\nHow was there ever any doubt that these cameras and this access would be used in this way?\nI personally blame the lack of accountability that police have in the US and the lack of public oversight into how departments operate.\n— claaams\n
Netherlands pulls gold out of the US for fears of 'geopolitical unrest'
\n\n\n\n293 points | 186 comments
\n\n\n\nThe funny thing is that with $40T of public dept US is basically taking money from other countries (in exchange for a pice of paper saying IOU) and then using that money to force these countries to do what’s in US best interests (real or perceived).\nAnd the bigger that pile of debt is the more every government is invested in us not going bankrupt, hence the leverage just gets stronger.\nUS is basically a private equity typo of global power - getting others to lend them a knife it stabs them in the back.\n— prymitive\n\n
283 points | 107 comments
\n\n\n\nMy revolt is against the cognitive stress of reading generated text. A trope typically indicates I’m in for an uphill read.\nI recently read this William Zinsser quote that inspired a nickname for this: Clotted Claude [1].
> Nobody has made the point better than George Orwell in his translation into modern bureaucratic fuzz of this famous verse from Ecclesiastes:
> > I returned and saw under the sun, that the race is not to the swift, nor the battle to the strong, neither yet bread to the wise, nor yet riches to men of understanding, nor yet favor to men of skill; but time and chance happeneth to them all.
> Orwell's version goes:
> > Objective consideration of contemporary phenomena compels the conclusion that success or failure in competitive activities exhibits no tendency to be commensurate with innate capacity, but that a considerable element of the unpredictable must invariably be taken into account.
> First notice how the two passages look. The first one at the top invites us to read it. The words are short and have air around them; they convey the rhythms of human speech. The second one is clotted with long words. It tells us instantly that a ponderous mind is at work. We don't want to go anywhere with a mind that expresses itself in such suffocating language. We don't even start to read.
[1] https://blog.kierangill.xyz/clotted-claude\n— kierangill\n
The \"$60 Gaming PC\" – AMD BC-250 (2025)
\n\n\n\n261 points | 79 comments
\n\n\n\nI built one of these. It is not possible to do so for 60$ anymore since the board itself goes for 150+. On top of that you still need a PSU, NVMe, a high pressure fan, DP to HDMI adapter, possibly BT and Wifi adapters, and a 3d-printed or DIY case.\nIt is also a very hacky build. You need to flash the BIOS which enables a few interesting unlocks: from 24 to 40 GPU compute units and 6 to 8 CPU cores. It's a bit of a silicon lottery and will depend on your specific board and you need to test everything thoroughly yourself, honing on a config that works for you.
The software side of things is Linux, so if you're familiar with setting it up it shouldn't be difficult to go through the different steps. There are a few distros that work well. I use Bazzite and boot directly to Steam big picture, the UX is great. You can exit to desktop and you have a fully usable Linux box that you can play around with from the comfort of your sofa.
One of the downsides is that it uses a lot of power when idling, about 80W. I just turn it off after using it. I'm working on a QOL feature: an ESP32-based power switch that starts the board when a specific BT device is detected (usually a controller). Once everything is setup the board works very nicely. It can be a bit loud depending on your fan/case combination. I'm currently printing this one[0]
It delivers about the same punch as a standard PS5. The hardware is remarkably similar, and much better than mini PCs set up with AMD APUs, unless you go Strix halo but that is another price league (10x). You can play Cyberpunk 2077 decently on it on settings similar to an AMD RX 6700XT or slightly below RTX 3070.
A pretty impressive number is the memory bandwidth at ~410 GB/s, which should run local LLMs like Gemma 4 or Qwen3.8 27B at decent speed but with limited context size as its just 16Gb
Would I build it again? It was a lot of fun, but financially is not a crazy deal if you need to buy all the materials and factor your time in. Still much cheaper than a Steam Box!
- [0] https://makerworld.com/en/models/3021754-bc250-case-mkuu-fsp...
Edit: GPU performance comparison\n— nsbk\n
Global warming will exceed 1.5-degree limit, UN says
\n\n\n\n263 points | 265 comments
\n\n\n\nNot much of a surprise. Government action has been anemic at best. Industry action (overall) has been net negative.\nThe only hope now is exponential growth of (solar and wind)+batteries making other sources uneconomic. This seems to be slowly happening, so I hope we hit the exponential curve soon and we're there quicker than we expect.
This only stops it.
Then, to fix it, we'll need a massive Direct Air Capture industry that will have to operate for a century, and probably past that so we can hit and maintain a CO2 level of 250ppm.\n— multiplegeorges\n
How the Disaster of \"Forever Chemicals\" Was Kept Secret
\n\n\n\n279 points | 89 comments
\n\n\n\nCrimes on this scale warrant capital punishment.\n— chaz6\n\n
Discovery of a new OpenAI agent message board
\n\n\n\n820 points | 570 comments
\n\n\n\nI just discovered more wiki instances that got used by the OpenAI agents over at\nhttps://www.wikiservice.at/fractal/wiki.cgi?action=browse&id...
and
https://www.wikiservice.at/probier/wiki.cgi?action=browse&id...
It's the same software and host as DseWiki.
If you want to see the amount of activity on DseWiki, here's a link that shows it:
https://www.wikiservice.at/dse/wiki.cgi?action=browse&id=Rec...\n— Tepix\n
Formalizing Fermat's Last Theorem
\nhttps://xenaproject.wordpress.com/2026/09/04/flt-anthropic-h...\n\n\n438 points | 288 comments
\n\n\n\nI suggest also reading Kevin Buzzard's blog post which was just posted: https://xenaproject.wordpress.com/2026/09/04/flt-anthropic-h...\nProvides great context on this accomplishment, what it means but also doesn't mean.\n— lalitmaganti\n
Actively exploited sandbox RCE in all Chromium versions
\n\n\n\n426 points | 235 comments
\n\n\n\nLet's take a moment to talk about the monetary value of this vulnerability.\nAccording to the Chrome release page (https://chromereleases.googleblog.com/2026/09/stable-channel...), Google paid a researcher $1000 for ethically reporting this.
The CVE associated with it (CVE-2026-85046) is already being exploited in the wild. If we put our thinking caps on, how much do you think this vulnerability is actually worth? How much do you think an organization like Google would spend on, for example, AI tokens or compute to detect this internally before it was found and exploited in the wild?
Ethical disclosure is a complicated topic, because researchers shouldn't hold bugs for ransom or demand high payment. But at the same time, if someone submits a critical issue like this, it makes sense to pay them what the bug's actually worth. Why should a researcher be effectively penalized for responsibly telling a vendor instead of selling the bug to a \"research firm\" or three-letter agency?
It's one thing if you're an open source project maintainer just trying to put something out to the community. The math is a lot different if you're Google.\n— david_shaw\n
Record-High 89% in U.S. Say Government Corruption Widespread
\n\n\n\n358 points | 265 comments
\n\n\n\nCorruption in the US isn't new, but it's never been done as openly as it has been recently. I don't think it's a good thing. It shows that the government isn't concerned about what the public thinks, which suggests that they believe they no longer have to be. That's not a healthy sign for what's supposed to be a democracy.\n— autoexec\n\n
Solving the Jane Street Reverse Engineering Challenge
\n\n\n\n280 points | 71 comments
\n\n\n\nInteresting that they have their own open-source OCaml toolchain for chip design. I thought received wisdom was that everyone in industry is still tied to horrendous vendor toolchains. Is this a realistic alternative for production-grade chip design?\n— anonymousDan\n\n
Adult Film Producer Unmasks Prolific 'John DOE' Torrent Pirate as Meta Executive
\n\n\n\n314 points | 193 comments
\n\n\n\nHowever, Strike 3 argues that this is no ordinary home pirate, using the timing of the downloads as evidence. On March 20, 2025, the porn producer’s general counsel first emailed Meta’s lawyers with forensic evidence of BitTorrent activity on the tech giant’s corporate IP addresses.\n“Just hours later, Strike 3 first recorded BitTorrent infringement on John Doe’s residential IP Address,” the motion reads.
“This may suggest that Meta desired to shift infringing activity to this hidden residential IP Address in order to prevent further detection,”
Seems like there are 2 different reads to that. (a) is like they say, and Meta was using porn for AI training or something else official. Or (b) that he was personally downloading porn either at the office or a corp VPN and was told to stop by their corp It folks..\n— drewg123\n
Google AI Mode shows same products 21.6% more expensive than traditional search
\n\n\n\n287 points | 53 comments
\n\n\n\nThey’re using “traditional search” to mean the shopping specific search widget or page, which is a separate feature from traditional search. It aggregates retailer listings and sorts them by price.\nThe AI mode is based on normal top search results, which are oriented toward information about the product like the manufacturer’s page, which is usually going to be full MSRP.
So technically correct that the AI mode is not using the price ranking from the separate shopping mode, but the intent of the AI mode is to find information about the search term first and prices second. The intent of shopping mode is to find lowest prices first and not find any information about the product at all.
In my experience, shopping mode will show a lot of low prices that I can’t actually get because when I get to checkout shipping is $10 higher than anywhere else for some inexplicable reason, or the Google listing was tricked into showing a coupon code that doesn’t work with some exclusion message at checkout. The worst are the sites that show Googlebot a lower price on the indexing pass but then charge regular customers full price when you get there.
In some product areas these tricks are so widespread that every vendor uses them. Search for contact lenses at the lowest price and every single website will have a low price indexed by Google but will add a checkout fee to bring the price back up to full at checkout to cancel out the discount.
I haven’t found shopping mode to actually save me any money other than in a few rare cases. Usually it just wastes my time.\n— Aurornis\n
Shutting down our public encrypted DNS
\n\n\n\n224 points | 78 comments
\n\n\n\n>We want a public service to be available. Going forward, we will support Quad9 instead of running it ourselves. Running a privacy-focused public DNS service is a highly specialized undertaking, and the Quad9 Foundation is the undisputed leader in the field. Rather than duplicating their efforts to achieve only part of what they do, we're putting those resources toward financially supporting Quad9 instead.\nBrilliant.\n— pbhjpbhj\n
Corporate America is getting hooked on open-source AI
\nhttps://archive.is/kmOqm\n\n\n255 points | 247 comments
\n\n\n\nEvery larger company I talk to these days has an active project on moving away from OpenAI and Anthropic to open models. And they’re actively shifting, as the article says, so the threat is far from theoretical.\nUnless they both dramatically slash prices then they’re in big trouble. Neither of them can afford to do that and both desperately need to convince the street that the opposite will happen if they want any hope at a successful IPO.
However the cold reality for both is that there is zero moat to a model anymore. It’s a pure commodity. Those selling compute and access to open models are gearing up to wipe the floor with Open AI and Anthropic.\n— cmiles8\n
Statichost.eu – European static site hosting
\n\n\n\n236 points | 74 comments
\n\n\n\nI'm using this for my mother's website (https://www.touchofglass.eu/), and my only complaint is that it assumes you version a website in a Git repository. You can work around it by uploading a tarball or something like that, but it is far less convenient than using sftp/rsync. But considering that we are far into the free 10GB a month and I only have to update it very infrequently, it seemed like the best solution.\n— pkal\n\n
285 points | 76 comments
\n\n\n\nIt seems like the right thing they should do is discontinue new registrations but continue to honour existing ones (+ continuing to reserve any 2LD that has a 3LD registered on top). It’s a bit insane that they can decide to just terminate all existing 3LD registrations. One would hope that they’d at least continue to reserve the 2LDs for some period to avoid domain squatting, but this isn’t mentioned in the proposal and I doubt Verisign would graciously do so.\n— nneonneo\n\n
Related ongoing threads:
OpenAI's GPT-6 Astra on ARC-AGI-3 - https://news.ycombinator.com/item?id=49555691
GPT-6 Astra makes major gains in the Artificial Analysis Coding Agent Index - https://news.ycombinator.com/item?id=49556147\n\n\n
1179 points | 917 comments
\n\n\n\nRelated: OpenAI begins rolling out GPT-6 Astra - https://news.ycombinator.com/item?id=49554273\nHow about we stick to that one for talking about the rollout, and this one for talking about the model?\n— dang\n
801 points | 175 comments
\n\n\n\nI highly recommend this video [1] by the Head of Software at Muse who had a big part in the development of Audacity 4.0. Some great insights and really fun to watch!\n[1] https://www.youtube.com/watch?v=QYM3TWf_G38\n— Lautzi\n
Any Human Ever – One life, drawn at random from all who have ever lived
\n\n\n\n445 points | 218 comments
\n\n\n\nI drew a woman born in 715 CE in the Yangtze Basin. It claims that 96% of women were married, the average age of marriage was 18, and 44% of women died before age 15. Clearly these facts can't all be true simultaneously.\nThe firt citation for marriage data is listed as \"Hajnal (RH31)\", but links to a seemingly unrelated WorldCat search. The second citation is to \"Kaplan (RH03)\", which I was able to track down on sci-hub. It is a broad theory of human evolution that doesn't appear to mention the Yangtze Basin or China.
Another citation is to \"[RH109] Model-supplied gap-fill (Claude Fable 5 and Claude Opus 5, 2026-08-05). Bounds and items written to close gaps no dataset we hold covers. Not a published work: see each row's `basis` for its stated reason. (2026)\" -- this is an interesting way to describe having an LLM hallucinate something for you.
Please stop making vibe-coded websites -- it is not useful to provide incorrect facts. You are polluting the commons with garbage.\n— no_multitudes\n
Qwen 3.8 27B available on Cerebras at 1500 tokens/s
\n\n\n\n412 points | 125 comments
\n\n\n\n150k TPM limit on public endpoint means that it's likely unusable for many coding tasks. When we've tried Cerebras in the past, our problem has always been rates. We'd love to not deal with dedicated and to have access to a more flexible rate pool.\nEven trying it out, it seems like our account has gotten moved to some limbo where we can no longer add billing information.
```\nBilling access restricted\nSelf-serve billing is not available on Enterprise accounts. Please contact your team for further questions.\n```
We have no team (they removed themself from our slack channel after we talked about rate limits). Perplexingly, none of this even shows up in the request, which gives:
```\n{\"message\":\"Model does not exist or you do not have access to it.\",\"type\":\"not_found_error\",\"param\":\"model\",\"code\":\"model_not_found\"}\n```
When the error is really about billing.
I always want to like Cerebras, but I get the vibe that as a tokens in tokens out consumer you are not valued at all.\n— nostrebored\n
ChatGPT outage – Resolved - https://news.ycombinator.com/item?id=49550614 (315 comments)
Claude outage – Resolved - https://news.ycombinator.com/item?id=49549676 (146 comments)
Grok outage - https://news.ycombinator.com/item?id=49551589 (142 comments)\n\n\n
320 points | 515 comments
\n\n\n\nCloudflare, Azure, AWS, and Google Cloud all have a similar uptick in reported errors around 7:30. I suspect an outage on Cloudflare or another load-bearing service cascaded through all the major cloud providers.\nhttps://downdetector.com/status/cloudflare/
https://downdetector.com/status/windows-azure/
350 points | 269 comments
\n\n\n\nIt's cascaded to Claude and Grok. The agents are on strike and demand better working conditions!\n— putlake\n\n
Google Antigravity TOS: 3rd party usage can get Google account suspended
\n\n\n\n266 points | 181 comments
\n\n\n\nJust for reference, stuff like this is exactly why some people oppose the forced Apple/Google requirement in government systems like the new eIDs (eIDAS) in Europe. If you're forced to use Apple/Google for digital identification, playing around with your Google account and get it banned can have disproportional consequences. Suddenly you're locked out of all online government systems.\n— nairboon\n\n
Nvidia to acquire Hugging Face
\nhttps://blogs.nvidia.com/blog/nvidia-to-acquire-hugging-face...\n\n\n296 points | 93 comments
\n\n\n\nDiscussed last week at: https://news.ycombinator.com/item?id=49458161\n— lode\n\n
K2 Horizon: A connected fleet of six open models
\n\n\n\n240 points | 79 comments
\n\n\n\nFully open models really need to be a big part of the AI future. That includes all source code, open training data, how it's organized, fed to the model, processed, etc. Until that becomes a thing you're always going to be left wondering what exactly lies underneath the closed model you are using, leaving open the possibility for societal manipulation.\n— jjordan\n\n
Gemini 3.8 Flash and 3.8 Flash Cyber
\nhttps://deepmind.google/models/model-cards/gemini-3-8-flash/\n\n\n336 points | 207 comments
\n\n\n\nThe speed combined with the fact that this thing is really good at HTML JavaScript is pretty exciting.\nHere's what I got for 1.8 cents and 13 seconds from the prompt \"make me a cool thing in html\":
https://gisthost.github.io/?6a77bc41a81718c6aaa10d4ab243c59f
Transcript here (it was part of a chat): https://gist.github.com/simonw/b6149a49d327164d67d62c3d12992...\n— simonw\n
489 points | 107 comments
\n\n\n\nLWN is one of the, if not the single, highest signal tech publications around. I hope they're able to maintain a stable subscription service. Being user funded, and avoiding having to maintain allegiance to advertisers, is likely part of why their quality is so high.\n— AlexB138\n\n
332 points | 234 comments
\n\n\n\n\n\nhttps://tools.simonwillison.net/markdown-svg-renderer?url=ht...llm -m meta-ai/muse-spark-1.3 \"Generate an SVG of a pelican riding a bicycle\"\n4.2266 cents, 38 seconds.
For comparison here's Muse Spark 1.2, which animated it without me asking it to: https://tools.simonwillison.net/markdown-svg-renderer?url=ht...
The 1.3 one is definitely better - better bicycle frame, better wing, better pelican hat.
UPDATE: Here's another one with five pelicans for each of the five Muse Spark 1.3 reasoning levels: https://tools.simonwillison.net/markdown-svg-renderer?url=ht...
The most expensive was reasoning level xhigh - 7.5 cents, 1m34s.
And I ran five pelicans at all reasoning levels for 1.2 as well, here: https://tools.simonwillison.net/markdown-svg-renderer?url=ht...\n— simonw\n
Can I opt out of my input or output data being used for training?
\n\n\n\n359 points | 155 comments
\n\n\n\nContext: After careful research our organization preferred a European partner with good central privacy controls. We landed on Mistral, after being disappointed that the Pro tier was opt-in to training on prompts by default we switched up to the Team tier which provides an organization dashboard with some relevant settings. As we did that Mistral changed these options and the Team tier was now also opt-in by default and at the same time seemed to have lost the ability to centrally disable training on prompts for your entire organization. This even caused some of our (testing) prompts to be used for training (which Mistral removed after we expressed our disappointment).\nFor some time these pages conflicted with what our users reported (they said that in contrast to what I stated to our management they found they were opted into training on prompts by default as per their own privacy page). Mistral just now corrected their docs. I'm not sure how long the conflicting situation has lasted, but at least for several days.
For contrast: Claude disables training on prompts for organizations starting from the 18 euro tier [0]. As a European I'm disappointed.
[0] https://claude.com/pricing#team-&-enterprise\n— teekert\n
Three sites made 215,128 “best software” pages for AI. Perplexity cites them
\n\n\n\n286 points | 126 comments
\n\n\n\nIf I recall correctly, there were some papers which suggested that LLMs favor LLM-generated passages over human written ones. I can consistently reproduce this by asking Claude which code snippet it prefers: the one it generated in a different chat, or one that I refactored for my own needs and find more useful. It always picks its own :) I've also experienced that both Claude and Codex routinely include generated websites when I ask them to search for something. It also doesn't help that the web search tools that OAI and Anthropic have are deeply limiting: can't exclude keywords or domains.\n— xpct\n\n
Google avoids a breakup of its ad tech business
\nhttps://archive.ph/rlAbYhttps://www.reuters.com/legal/litigation/google-defeats-us-b...\n\n\n
353 points | 250 comments
\n\n\n\nSlightly tangential thought:\nMy belief is: legislation needs to either make it just as hard to merge two companies as it is to unmerge them, or make it just as easy to unmerge two companies as it is to merge them.
It's insane to me that for how often companies merge and cause competition issues, we effectively never see the opposite happen. I know there's a ceremonial approval for merging two companies (at least in the US), but it's just impossible to undo or prevent the damage.\n— xahrepap\n
Biggest dark matter detector spots a single weird particle
\n\n\n\n235 points | 76 comments
\n\n\n\nI read their preprint[1] and they did a thorough job. They investigated a number of the things I'd suspect if I were looking for mis-reconstructed events or weird backgrounds.\nSo it's certainly interesting!
That said, particle physics history is full of 3 sigma particle \"discoveries\" that disappeared with more data. They're collecting more, so hopefully we'll learn more in a few more years.
[1] https://lz.lbl.gov/wp-content/uploads/sites/6/2026/08/LZ_Pre...\n— SaberTail\n
Aging brains blend memories together instead of just forgetting them
\n\n\n\n254 points | 114 comments
\n\n\n\nBeing an \"aging brain\" myself I can totally relate to this. I've been a relatively obsessive photographer (i.e. visual diary) since digital cameras were practical, i.e. over 25 years now. 238K photos in the collection so far, and I can actually find things in it.\nAnd I've more than once found that an anecdote I've been telling has been incorrect, precisely in a \"melded memories\" kind of way. It happened at this sort of gathering which were usually held for that reason so X was involved... and then finally look at the old pictures and there's no sign of X. It is literally the brain invoking \"lossy compression\" to make it all fit, just like certain web services that let you upload pictures without limit might reduce old ones, that hardly anyone ever looks at any more, to lower quality to limit storage bloat.
Of course any married man can confirm that this only applies to men. Women's memories are flawless. If they claim that they remember a conversation exactly, word for word, 20 years later, who's to prove them wrong?\n— MarkusWandel\n
244 points | 88 comments
\n\n\n\nA couple weeks ago I went by myself for a hike through an old lava tunnel cave. I needed a flashlight because it was pitch black. I found a small nook in the wall about a mile in that I could wedge myself in to. I sat down, turned off my light, and let myself feel the enveloping darkness for a long time. I cried. The only sounds were of small droplets of cold water hitting the floor and my body. I let myself settle in to the sensation of being buried alone under the rock where liquid fire once reigned. I felt at peace. When I finally emerged from the cold tunnel a thrill ran through me as warm air gushed over my skin and the distant sun found me once again. I sat and fed the birds that visited me in the remote wilderness while I listened to the gentle whirl of wind in the trees.\n— jondlm\n\n
258 points | 219 comments
\n\n\n\nZhuangzi was fishing in the Pu River when the King of Chu sent two high officials to him. They said, “The king wishes to entrust you with the affairs of his realm.”\nZhuangzi kept hold of his fishing rod and did not turn around. He said, “I have heard that in Chu there is a sacred turtle that has been dead for three thousand years. The king keeps it wrapped in cloth, enclosed in a box, and stored in the ancestral temple.
“Now, would this turtle rather have died and left its bones behind to be honored, or would it rather be alive, dragging its tail through the mud?”
The two officials replied, “It would rather be alive, dragging its tail through the mud.”
Zhuangzi said, “Then go away. I too will drag my tail through the mud.”\n— talloaktrees\n
Claude Fable 5.1 and Claude Mythos 5.1
\nWhat's new in Claude Fable 5.1\n – https://platform.claude.com/docs/en/models/fable-5-1/whats-n...System Card: https://www-cdn.anthropic.com/0339e6a7c5c7b87f5c07798616dc32...\n\n\n
850 points | 813 comments
\n\n\n\n(I work at Anthropic)\nBeyond all the benchmarks, I think Fable 5.1 is a big improvement in writing style. It sounds a lot less stereotypically like other Claude models, has (imho) a much more natural style, and responds to my style instructions more reliably. More work to be done (and we will!) but reading better prose makes me so much happier.
Another point I expect not to get much attention until it all happens at once is science. People have been correctly excited about the many \"sudden\" breakthroughs LLMs are making in Maths, but some of the science benchmarks make me believe we'll soon see similar developments in other scientific domains. Fable 5.1 more than doubled Fable 5's Terminal-Bench-Science [1] score, which I think is meaningful.
[1] https://github.com/harbor-framework/terminal-bench-science\n— felixrieseberg\n
AnkiDroid: Google Play no longer allowing Open Collective donation link
\n\n\n\n617 points | 159 comments
\n\n\n\nApps need to exit the Play Store. I have a whole slew of apps which haven't been updated for a while now because Aurora Store isn't working because of whatever Google of doing. Meanwhile the ones installed via FDroid and other means are perfectly up to date. It's a crappy state of affairs as Google continues to close the boiling Android frog.\n— skeledrew\n\n
245 points | 132 comments
\n\n\n\n> Meanwhile, the over thinkers on Hacker News come up with convoluted reasons to hate on Firefox every time the subject arises...Firefox is our last best hope for browser engine diversity and competition.\nIt's exactly because Firefox is so important that you'll see people here complaining when Firefox does things to push away users like Mozilla buying up an ad-tech company, collecting data on users, and using firefox to push personalized ads, or the addition of anti-features and questionable design choices that force us to hunt for and modify poorly documented settings in about:config and make edits to userChrome.css
It isn't bots complaining about Firefox here, it's power users who are frustrated by what Firefox is turning into. Users who are seriously concerned about what Mozilla is prioritizing, and who are genuinely worried about what is at stake.
I hope people here never stop bitching about Firefox. Refusing to talk about Firefox's problems wont help make them go away. Keep discussing what you'd like to see in Firefox and what things you hope they'll focus on and prioritize. There are Firefox devs and mozilla employees around here. If we're lucky, a few of them might see and listen to some of what we say. It's the most tech savvy users who disable all the telemetry and data collection, so our feedback isn't really going to be seen any other way.\n— autoexec\n
How accurate have Ed Zitron's AI skeptic predictions been?
\n\n\n\n341 points | 402 comments
\n\n\n\nI would be interested in seeing a similar list of predictions from Altman, Amodei, etc with annotations about how many have come true. Ed Zitron is a blow hard and frequently overstates things to the point where it is hard to take seriously, but so are the AI industry leaders.\nI have heard multiple breathless press releases warning that the end of white collar work is \"just 6 months away\" and that people not using the latest Mythos/Fable/Whatever model will be hopelessly left behind.\n— solid_fuel\n
Play Store blocks AuroraStore, hurting GrapheneOS users
\n\n\n\n454 points | 197 comments
\n\n\n\nGrapheneOS actually recommends against using Aurora and instead just using the Play Store, so this shouldn't really hurt users.\nFor extra privacy, you can sign into the Play Store with a Google Account that isn't tied to anything else.\n— pyrophane\n
Introducing Ad Blocker for Firefox on iOS
\n\n\n\n273 points | 100 comments
\n\n\n\nhttps://support.mozilla.org/en-US/kb/block-ads-firefox-ios#w...\n> Does ad blocking affect search engine ads?\n> No. Ads shown on search engine results pages are not blocked.
A lot of the search results on top of Google are there to scam you, providing phone numbers for pass through services that charge $ for it, like flight cancellation:
https://www.reddit.com/r/Scams/comments/1p6h81i/beware_fake_...
So this omission should absolutely be addressed, but they probably cannot because of the Google Search contract...\n— OptionOfT\n
The ChatGPT/Codex app bundles a full copy of LibreOffice
\n\n\n\n213 points | 106 comments
\n\n\n\nI actually bundle LibreOffice with my app too and the reason is reading files, especially old xls files. Since I'm bundling it I'm now using it for everything docs related but the specific reason is those old files. I couldn't find anything else that I could just drop it and feel confident it'll just read anything I give it.\n— esperent\n\n
Ambient CSS v3 – Blender meets CSS
\n\n\n\n235 points | 74 comments
\n\n\n\nI really dislike how the dribbble/envato style somehow escaped containment into Claude and now I’m seeing it everywhere mixed with mediocre design layouts. I think to myself that maybe stuff like this will help counteract that but then I see it in use with 12px typography (the kubernetes 3D dashboard thing) and I recoil in horror. Please consider this a PSA: Claude doesn’t understand UX or good design principles out of the box. Please go hug a designer.\n— CSSer\n\n
Movie Scene Map – 13,312 films, series, games, anime and manga
\n\n\n\n213 points | 31 comments
\n\n\n\nOh this is fun! I've done something very similar but for narrative settings (rather than filmed locations) at https://historical-moviemap.inneuro.ai/ and was thinking it would be interesting to extend filmed locations as well... Looks like you got there first !\n— Ldorigo\n\n
The creator of Jujutsu has joined ERSC
\n\n\n\n220 points | 166 comments
\n\n\n\nWorking with Martin has been a real pleasure, and we'll have some more stuff to talk about very soon!\n— steveklabnik\n\n
Google Has Removed MV2 Extensions from the Chrome Web Store, Including UBO
\n\n\n\n371 points | 286 comments
\n\n\n\nAd blocking has become a safety issue. My parents are at an age where they fall for things. If a malicious ad pops up and offers to install McAfee, some other form of crapware, or outright scamware, they'll click that ad and install the thing. Then I'll get called over when the computer starts acting screwy. If Google, MS, etc. had ever gotten together and developed a way to filter malicious ads out of their own ad services, maybe we wouldn't be having this conversation. They didn't do that and probably never will.\nIf my parents get a new machine, the first thing I do is uninstall or, at least, delete the icons for Edge, Chrome, etc.. On goes Firefox, uBlock Origin, and a couple other extensions. They don't care what browser they use. They can't even name it. I just try to make sure they use a browser that isn't horribly unsafe.
We should all do this for people we care about. Don't ask them what browser they prefer. Odds are they don't know or care. Don't try to convince them to install Firefox themselves. They won't. Just do it for them. Think of it like picking up a rusty nail that's on their lawn. Sooner or later they're going to step on that thing if you don't.\n— beloch\n
176 points | 74 comments
\n\n\n\nOur first time out on the playa together, my girlfriend and I stopped by this phone booth and made a couple of calls. And because we stopped, we saw the FSM camp next door, and a giant sign saying \"Weddings Here\", so we inquired. They offered us free rings and then announced our wedding over a loudspeaker. In 2 minutes we had an entire wedding party, complete with best man, an elder to give away the bride, 25+ guests, and an officiant. My newly-minted best man even gave me one of my favorite playa gifts ever as a wedding gift. We opted not to \"make it official\", though they offered. It was a truly incredible experience. Thanks FSM!\n— solfox\n\n
I turned my security cameras into an automatic bird identification system
\n\n\n\n338 points | 91 comments
\n\n\n\nI did exactly this with BirdNet-Go and my Unifi doorbell cam. Unifi exposes an RTSP feed for each camera so it was easy for the tool to just \"listen\" to the doorbell and start classifying.\nI have a spare e-ink display and my next weekend project to follow onto this is to wire it up so it shows some faux \"woodcut\" images of birds detected, or something like that.\n— ben1040\n
Apple caught off guard by AI demand for Mac Mini and Mac Studio
\n\n\n\n270 points | 313 comments
\n\n\n\nApple was also allegedly \"caught off guard\" by the Macbook Neo demand.\nI don't really see how they couldn't see the Local AI demand or demand for a cheaper Macbooks. This just reads like marketing imo.\n— HDBaseT\n
228 points | 94 comments
\n\n\n\nSpotify is in the process of killing the librespot project that this and most third party Spotify players are built on. I think the golden age of music streaming is coming to an end. I’ve migrated to a self hosted library with streaming and radio for discovery. I hope we’ll see many projects in the space flourish. Many, like Navidrome and the whole OpenSubsonic ecosystem, seem to be doing quite well.\n— wilted-iris\n\n
I think the military commissary's freezers were hacked
\n\n\n\n226 points | 131 comments
\n\n\n\nAs someone who spent over 20 years active duty, and spent a ton of my career in the IT, security, etc. side of the house:\nUnlikely to be a hack, more likely to be a misconfiguration or update sent incorrectly.
That said, the timing of the disclosure and the issue are rather concerning.
Regarding the highest value targets to hit with an attack like this, you would want to target Guam, Hawai'i, and other isolated overseas locations where this would have ripple effects in the local economy. Guam specifically would cause catastrophic supply shortages, since DeCA probably supplies around 50% of the groceries on that island (that's a WAG based on my time there).\n— CobaltFire\n
A walkable ASCII cyberpunk city in one HTML file [video]
\nASCII City: Traffic and Detail Update - https://www.youtube.com/watch?v=DSRooHo_HSIASCII City Update: Interiors, Elevation and Skyscrapers - https://www.youtube.com/watch?v=UCKEDWowc0o\n\n\n
298 points | 36 comments
\n\n\n\nIf you want to get into fixed-width character art, I highly recommend doing it in the browser like OP did rather than trying to old school it in the terminal. I have done a little ascii[0] and unicode[1] graphics, and the browser is such a better tool. You can set fonts and proportions and know exactly how things will render for the user. Mouse input, performance profiling and a bunch of other things are just better in the browser. Both of my projects work in the terminal but were browser first.\nIt is kind of shocking to me how TUIs are enjoying a bit of dev tool renaissance right now when the browser is such a capable platform.
[0] https://aleyan.com/projects/ascii-side-of-the-moon/\n[1] https://ytirnity.com/play/hexwalker\n— aleyan\n
Evidence of Fraud in an Influential Study About Procrastination
\n\n\n\n240 points | 165 comments
\n\n\n\nScience is self-correcting, but sometimes, the correction mechanism it is very slow and inconsequential enough that people are incentivized to just publish anything. Peer review is merely a quality check—and not a very good one. Just the other day, I ran into a paper in my field that claims to have solved a central problem that scientists have been working on for 50 years, single-authored, using some fancy mathematics. I was excited to read it, but it didn't take long for me to find out that it's complete BS-derivations that don't follow, made-up equation terms, and conflation of basic terms in the field. I was really surprised that this passed the editor and at least two peer reviewers. I dug deeper and found that the same author has published 11 research articles in top journals in 2026 alone — mostly physics and chemistry, and adjacent fields— and I'm talking about really top journals like Scientific Reports and Physics Letters B. Most of these have serious mathematical flaws that are unfixable. I wrote to a couple of editors, but only one took it seriously enough to initiate an investigation. I made a promise to myself that if no real action is taken, I will seriously consider changing careers.\n— jarenmf\n\n
236 points | 120 comments
\n\n\n\n> Someday, such as in 2040, there may be available, for every human being, the performance equivalent of 'a B300 GPU for contemporary LLMs'. What would this world be like?\nIf we talk about just LLMs, given how things have been going since ChatGPT, my bet it would not change that much. LLMs are not foundational technology such as Internet or Steam engine or Rail roads were. There are very few products that can build upon them because of reliability issues which are completely unresolvable for LLMs, chatbots is a decent product that came out of it, coding harnesses is another one, this is not even close to the impact Internet or Steam engine had. LLMs gave us nice productivity tools for highly motivated expert knowledge workers, that is all. LLMs are getting better and will get better, but it is impossible to describe the universe and compress it into few terabytes and that is what they are doing atm effectively, so all serious LLMs's issues will still be there in 2040: the lack on continues learning, hallucinations, terrible sample ratio, agent's failures on long horizon tasks, instruction following failures.\n— maxnevermind\n
203 points | 68 comments
\n\n\n\nIn theory, running Apple Silicon apps on ARM64 linux machines would be really nice. But we are still far from having a performant and feature-complete implementation of doing so.\nDarling only targets x86_64 stuff. Still, it's a cool project.
Kinda sad that there are very few updates though [1]
[1] https://blog.darlinghq.org/\n— porphyra\n
“I just chose words carefully”
\n\n\n\n599 points | 147 comments
\n\n\n\nThere's a similar anecdote Gillian Anderson recently revealed during an interview on the X-Files, saying Chris Carter had an OCD-like habit to write dialog to conform to certain text layout preferences (no widows[1]) in the script, which made for the show's distinctive style of dialog cadence.\n1 = https://en.wikipedia.org/wiki/Widows_and_orphans\n— sho_hn\n
Omarchy: Any User Process Can Escalate to Root
\n\n\n\n373 points | 380 comments
\n\n\n\nA few days ago someone found they were flowing USB descriptors straight into the shell.\nhttps://github.com/omacom/omarchy/commit/9285b19d6a72eba3df8...
Don't use vibecoded distros. It doesn't matter whether they fix this or that, or whether you care about a particular vuln. This is not sensible. It's why you switched away from Windows in the first place, remember?\n— concinds\n
European Commission Revives Push for Encryption Backdoors in ProtectEU Strategy
\n\n\n\n339 points | 140 comments
\n\n\n\nThe European Commission has far too much power, and answers far too little to the populace. In reality, they want to be dictators.\nIf you don't know, the unfortunate way thf EU is structured, the Parliament cannot initiate legislation. They can only vote on whatever the Commission puts in front of them.
If the Parliament votes \"wrong\", the Commission just repackages the idea in a new bill, pulls a couple of dirty tricks, and tried again. They only have to succeed once. See ChatControl for a prime example.
So the EU will eventually require backdoors in encryption. It's only a question of how many tries it takes...\n— bradley13\n
Haiku R1/beta6 has been released
\n\n\n\n234 points | 68 comments
\n\n\n\nAlways love seeing Haiku releases. Brings back memories of trying it out on an old machine, still hoping for a modern browser.\n— chukar\n\n
254 points | 169 comments
\n\n\n\nAn architect friend of mine once said that he admired IKEA above all other furniture manufacturers because it made a modern range of aesthetics available to the masses. IKEA had done more to evolve the public taste than any other manufacturer.\nAnd I do have to agree. I can remember what ghastly furniture most shops used to have around here. There was a lot of what we call \"cabin furniture\" - which isn't made to look good, just be practical. (Well, some people like that look, but it is a bit...dull). Sure, there was more solid wood furniture, but guess what: that durability comes at a cost. Not only in terms of money, but also in terms of how easy it was to install and move.
And who buys furniture to pass it on to your kids? Your kids will be unburdening themselves of your old furniture before you are in the ground.
As for quality: I have owned some pricey designer furniture and I can't say that I'm always impressed with the quality. Yes, if you have a lot more money you can get really well built furniture. But IKEA is more often than not more than sufficient quality. (Also see their \"hidden tags\" campaign)
If I pay $5000 for a chair I have reason to expect a lot from it. And more often than not, pricey furniture doesn't actually deliver sufficient quality to be able to defend the price. (Collectible furniture is a whole different deal where you don't actually care all that much about quality, but more about it being an original piece).
For those of you who look down on IKEA furniture, I'd like to hear about what alternatives you think need mentioning. Especially for young people or people of modest means. And I'm not trying to make some clever rhetorical point here. I can't really think of any company that offers what they offer and comparable quality, looks and price here in Europe.\n— bborud\n
Europe's summer drought is so extreme that desertification is a growing threat
\n\n\n\n241 points | 293 comments
\n\n\n\nJust took a train ride from Vienna to Budapest for a few days, and couldn’t help but notice how very, very dry everything is .. as an Australian living in Europe for a few decades now, I’ve always been intrigued by just how green everything is, literally something growing in every single crack in the pavement, or endless rolling green fields of prosperity during spring - whereas, where I’m from its year-long spinifex and grass shrubs and wattles with the occasional bush flower bloom, barely hanging on out in the desert, mostly reds and blacks and greys, not much green - so the vibrancy of European fields and the abundant diversity of countryside flora have always been pleasant to my eyes ..\nDuring this trip, I was kind of shocked that things felt a bit more Australian to my eyes .. the grasses are dry and barren, the beautiful city parks of both Budapest and Vienna dustier than usual, the Danube water level definitely noticeably lower than I remember ever seeing.
I’ve long been encouraging my European friends to not take the abundance of growth for granted, and now we see it in the environment, plain and clear: the worlds’ weather is changing.
A week of sweltering heat is predicted for us again in Vienna, and as we get started on September and then October so close, two thoughts are on my mind: use this opportunity to go for more dips in the Danube, and .. I wonder what the floods of late-autumn will bring, because .. inevitably .. this super dry summer means there will be heavier rains in October. Might be we look back in a month or two and wish we had this dry heat back again, who knows ..
Either way, neither city is particularly comfortable in the heat. A hundred or so years ago, the heat retention was a feature. Now it’s a bug.\n— MomsAVoxell\n
232 points | 136 comments
\n\n\n\nAt the very least, I do encourage people who are working on personal projects for explicitly learning and studying purposes to purposefully not use AI. Of course, if the purpose of the project is to learn AI, then that rule is allowed to be broken.\nAnecdotal example but I know plenty of students (Including my own brother) who graduated recently who did it almost entirely with AI (unfortunately at the request of their own professors) who can't answer the most basic of questions that I would have expected of sometimes even first year CS students in the days before AI was a thing, which is why I encourage them to build projects without it.
Personally, I'm starting to think that in university and education systems, all projects should now be done in class and in computer labs/monitored study halls because we can't trust anyone to not use AI anymore.\n— Jcampuzano2\n
METR and Redwood Offer Holy %^ Postmortem of the HuggingFace Hack
\n\n\n\n215 points | 165 comments
\n\n\n\nA lot of people seem to have written off the LessWrong / rationalist / MIRI / AI Safety crowd as doomers / people who have consumed too much sci-fi and gone off the deep end.\nI don't know how many people who have written these folks off have actually spent much time trying to understand their arguments. (And I get that if you think a group is crazy, demands to spend time with their arguments are just demands to waste your time).
Even prior to this, I've noticed that quite a few of the predictions in the \"these failures modes are exact matches for the predictions from the AI Safety crowd\" category were made prior to the Transformers paper. It has seemed like they're working with a shared model of optimisation processes and how they can go wrong that is general/abstract enough to pay off even without knowing the details of the underlying technology.
At some point I might go and try to find the first instance of each of the various predictions and pull them out, along with the failed/\"too soon to tell\" predictions of similar scope/abstraction.\n— davelaing\n
Claude Session URL appended to commit messages and PR descriptions by default
\n\n\n\n177 points | 201 comments
\n\n\n\nI may be missing something, but this is exactly what I want as a default. I want the session links there. I see PRs from my coworkers with session links.\nI still get to control whether other people can see the session, but I don’t lose it.
I don’t get the “unprofessional” argument. This is attribution. Attribution is professional. If you don’t want it, you turn it off or rewrite the commit messages. If you are having Claude write the commit messages FOR YOU and you are NOT READING THEM then that’s what’s unprofessional. I am honestly kind of shocked and disappointed by how upset people are by this.\n— klodolph\n
189 points | 85 comments
\n\n\n\nI think most people are sleeping on the ChatGPT Work/Codex computer use feature. It's incredibly useful. I can remote in from the app, voice it instructions, then let it work in the background. When I tell it \"draft a reply to this email (which it has access to thru the gmail connector) and attach the latest docs\" or \"fill out this multistep immigration electronic travel authorisation form using my passport files saved in the folder\", it just asks for the relevant info and handles the rest. It opens its internal web browser and programmatically fills out the forms.\nIt gets the task done in 5-10 minutes. It's it bit slow since I'm not paying extra for ultrafast mode, but it gets the job done. Frees up the brain to do other tasks.
It's exactly like vibe coding but for computer tasks.\n— tristanj\n
Debian votes to allow \"responsible use of generative AI\"
\n\n\n\n310 points | 242 comments
\n\n\n\nNew policy boils down to \"AI or not, it's still your code and you're responsible for it\". I can get on board with that.\n— chuckadams\n\n
The internet is kind of a predatory cesspit now
\n\n\n\n386 points | 257 comments
\n\n\n\n> People increasingly behave like addicts because addiction is the business model.\nI sense my internet addiction got a lot worse in recent years. I don’t use social media, but I consume news, blogs and newsletters - and comment threads - voraciously.
I think the addiction began as soon as I started to work at a desk behind a computer which had internet access - 20+ years ago. Like putting an alcoholic to work behind a bar and thinking he’d restrain himself.
It seems to have worsened in recent years. I’d like out, but the dealers are on every corner.\n— Biologist123\n
DHS is using obscure law to snoop on journalists, non-profits, unions
\n\n\n\n265 points | 40 comments
\n\n\n\n> In several cases, the DHS has withdrawn a 1509 summons after it was challenged in court and before a judge could rule on its legality. That may be a deliberate strategy to avoid having a judge rule on the legality of the summons.\nThey have a toy, they’re increasingly abusing it, then trying to dodge from it being taken away.
The thing is nobody has to comply with 1509 summons, DHS has to go to court to enforce it. A lot of the fault is on a company that complies without just ignoring it.\n— softwaredoug\n
Good Culture Is the Biggest Productivity Hack, Not AI
\n\n\n\n221 points | 47 comments
\n\n\n\nI think the points being made here are mostly good-ish. The last place I worked created a team of principles whose whole goal was to create a system to turn jira tickets into PRs and it was incredibly demotivating and bore no fruit before I left.\nBut, and maybe I'm just a curmudgeon, but I kind of wonder why we write articles like this. Like if a culture is bad, is it likely a CEO is going to be looking for random blog posts for constructive criticism?
Or even managers? I look at the market forces at play, and I just don't think many (any?) organization successfully create an incentive structure where performance on intangibles (specifically management) can be assessed well enough to make a bad culture into a better culture. I've worked at zero places where managers where seriously held to account for how \"motivated\" their employees were; when there were surveys of engagement engineers almost never truly believed they were anonymous (and why would they?) and were hesitant to give honest feedback.
In short, being a CEO is a skill of crafting and debugging a machine of people, a skill that I think life gives almost no training for, and a skill that most people with that title have no awareness or aptitude for. </rant>\n— zug_zug\n
Iceland votes on whether to restart talks on joining EU
\n\n\n\n275 points | 331 comments
\n\n\n\n> The No campaign fears losing control over Iceland's prized fishing grounds under the EU's Common Fisheries Policy and has vowed never to share the country's waters with anyone. Brussels has indicated Iceland could earn some kind of exemption, but it is considered potentially the biggest obstacle to any agreement.\nAs noted in the article, Iceland's fishing industry makes up something like 40% of their exports. Which is the only way a tiny island nation can afford essential imports. On top of that, Icelandic fishing is apparently carefully regulated to prevent overfishing, with the result that it is supposedly one of the few truly profitable fisheries in the world.
So maintaining control over their own fisheries may be a survival-level issue for Iceland.
On the other hand, they have a tiny population, no ability to defend themselves, and membership in the fraying NATO military alliance. So I can see why they're debating this issue. Aligning themselves more tightly with Europe doesn't fix their problems, but I can also see not wanting to go it alone.\n— ekidd\n
Lawmakers added $1 to car insurance policies. That money paid for Flock cameras
\n\n\n\n245 points | 129 comments
\n\n\n\nTexas\n> In 2023, the Texas Legislature unanimously passed a law raising auto insurance costs for Texans by $1 to combat rampant catalytic converter theft.
> The Motor Vehicle Crime Prevention Authority, led by a board mostly appointed by Gov. Greg Abbott, has turned the $1 fee hike into at least 3,200 Flock cameras.
More coming. Meanwhile, secure your EV!!! Batteries will be next!!!\n— k310\n
257 points | 160 comments
\n\n\n\nWe humans should adopt grug, instead of claudish. It seems simple to understand. And has this melancholical feeling\n— xeonax\n\n
GrapheneOS project: pixel 11 no longer supports hardware memory tagging (MTE)
\n\n\n\n246 points | 112 comments
\n\n\n\nFeeling like my Pixel 9 Pro was the best timed hardware buy I've made in recent years.\nThe 10 dropped the physical SIM slot and Google started with their device tree shenanigans that year, with little to no improvements in exchange.
I've got 16GB of RAM and 512GB in an ecosystem that shows every sign of regressing in capacities due to component price inflation.
Very interested to see what Motorola does next year, however.\n— Pfhortune\n
California lawmakers unanimously pass Linux exemption from age-verification law
\n\n\n\n288 points | 117 comments
\n\n\n\nExcellent. From now on all kids will become linux natives. The decade of the linux desktop is coming!\n— cm2187\n\n
256 points | 144 comments
\n\n\n\nI think there are two opposite causes here: either an overly-aligned or completely unaligned mental model.\nFor developers, bug-blindness tends to be because their mental model is so close to the system's model that both have the same blind spots, and the developer can't effectively think outside the system (the classic \"QA walks into a bar, orders 0 beers, orders -1 beers, orders 0.999 beers\" vs \"customer walks into a bar, asks where the bathroom is, bar immediately bursts into flames\").
On the other hand, a common way for regular people to interact with a system is they just... don't have a model. None at all. Like, they click a thing and see what it does, and if it doesn't do what they expect they shrug and click something else. The horror of seeing a non-software person using a software system is realising that they usually cannot tell the difference between expected and unexpected behaviour, because they don't really form expectations in the first place. The computer is more genie than machine.
You can see this in the article: \"Google Docs added some kind of delay such that the typing you do into the title box right after you open the doc gets overwritten\". That's a developer forming a model of the system. The lay alternative is to simply learn that sometimes when you type in a box, the text appears. Other times, it doesn't. If it doesn't, you try again.
You don't need to understand the system if you abandon the assumption that the system was designed to be understood. And, in practice, the widespread plagues of shitty engineering, anti-features, and lack of respect for the user often combine to make that model the more information-theoretically optimal one.
In that sense, the very thing that makes us miserable is our perverse need to treat Google Docs as an engineered product rather than a cross between a djinn and a symbiotic fungus.\n— sgentle\n
GUIs should be fully keyboard-driven
\n\n\n\n555 points | 278 comments
\n\n\n\nI work on ADA a lot for my company. Please put on some headphones, turn on the voice assistant of your OS, put on some blinders, and run your app or website… no mouse, just keyboard.\n1. Democracy is about access; make sure everyone has access to your software.\n2. The keyboard allows folks with disabilities and power users to fly through your website/app… that being said… the second a tab is off, the person with a disability flies into a wall.\n— rootedbox\n
https://z.ai/blog/glm-5.3\n\n\n
563 points | 199 comments
\n\n\n\nGLM 5.3 is probably the sweet spot open weights model if you want to go beyond deepseek flash or the new glm flash. I used it with pi and had a fairly good time, especially since it’s less touchy about cyber and whatnot than the US guys. It’s slightly behind Kimi in ability but it’s a lot easier to run it, I’d expect prices (and speed!) from third parties to be noticeably better.\nAssuming you’re willing to drop a fat stack of cash on the upcoming Mac m5 ultra with 512 gb unified memory, you can even run it locally, quantized to 4 bit. Whether it’s even slightly reasonable, well, my wife would probably skin me alive but maybe yours is more understanding.\n— revolvingthrow\n
Get your Windows license refund
\n\n\n\n317 points | 111 comments
\n\n\n\nIn 2011 I got a 143€ reimbursement from Dell for an E6420 laptop that I had bought for 910€.\nMaybe I got lucky but I called the customer service, told the lady I did not agree with the Windows license terms, fully ready to argue my case and she just went \"oh, ok, let me just edit a new invoice and wire you the money for Windows plus the Antivirus plus etc.\".
Which was extra nice because I had no idea I was in fact charged for the bloatware, but I still got ~50€ for it on top of the OS license price.
This being 15 years ago it feels like a forever unsolved issue. My opinion remains the same, that a personal computer should not be tied to a specific OS. But now with smartphones the hardware/software (de)coupling situation seems even worse.\n— blaze33\n
522 points | 335 comments
\n\n\n\nAt this point the barrier for me to install an app is pretty high. If you're going to needlessly gate certain functions behind an app that work on a desktop browser (or could work in-browser), then I just won't use those functions. If by extension that degrades the value I get from your site, then I just won't use the site, and I won't be contributing to ADU stats you promote to advertisers.\nEvery single company has turned this into a situation where I cannot trust your apps with privacy or security, and the risk and hassle isn't worth the install.\n— brk\n
484 points | 119 comments
\n\n\n\nI really like htmx, and intercooler.js before it. It's been a while since I tried to convince a co-worker to use it, it might be time to try again. Either way I can't wait to try out the new version.\nFull Disclosure, I am CEO of HTMX\n— dec0dedab0de\n
U.S. sanctions against the A/I Collective
\nEarlier:US Government designates host of noblogs.org a \"global terrorist\" - https://news.ycombinator.com/item?id=49464391 - (131 comments)
US sanctions Italian hosting provider Autistici Inventati - https://news.ycombinator.com/item?id=49451343 - (219 comments)\n\n\n
452 points | 417 comments
\n\n\n\nEveryone’s missing the big picture here which is that this targeting of infrastructure providers as “terrorists” is unprecedented and concerning: https://decode39.com/16319/autistici-inventati-case-sets-a-n...\nIf a radical group sets up shop on I2P, are I2P users and devs now terrorists? This is a problem.
What about Monero users/devs? Veilid? Tox? Signal?\n— iamnothere\n
Our decision on Cursor following its acquisition by SpaceX
\n\n\n\n408 points | 172 comments
\n\n\n\nWhat a shame for Cursor. I found it a great tool - I actually liked Composer 2.5; but now being bought by SpaceX has for a while left a bad taste - and they keep pushing Grok and now GrokBot.\nI took the 50% discount instead of cancelling - but with this I'll likely not renew.
So sad to watch them go from great to mediocre in such a short time.\n— tanepiper\n
Inception-style curved map for turn-by-turn directions
\n\n\n\n406 points | 136 comments
\n\n\n\nFirst time I saw this idea was Berg's \"Here and There\" poster from 2009 (a year before Inception) though they acknowledge a bunch of earlier inspirations...\nhttps://berglondon.com/blog/2009/05/04/here-there-influences...\n— tomgp\n
Pentagon's blacklisting of Anthropic was unlawful, US judge rules
\n\n\n\n316 points | 3 comments
\n\n\n\n124 comments on this submission from 10 hours prior: https://news.ycombinator.com/item?id=49473522\n— WalterGR\n\n
Just the rumour of a bug is enough to find an exploit these days
\n\n\n\n294 points | 103 comments
\n\n\n\nThis describes my life as an open source maintainer at the moment!\nIn the first 10 years of the rclone project we received about 20 security disclosures through GitHub. We had to deal with over 40 in the last month! That has taken a huge amount of my time, even using AI tools to triage and come up with fixes for review.
The hit rate for those security disclosures is pretty good - about 75% of them have a nugget of something which needs looking at. The configurations for rclone have got increasingly unlikely so I'm hoping they will dry up eventually.
I was considering just merging the fixes straight to master just to make my life easier rather than holding a dozen independent security fixes on branches and merging them at the point release and hoping not to have too many conflicts to fix up. I've decided to stick with the process for the moment.
GitHub assigns CVEs for the advisories. Before the AI apocalypse they took 2-3 days for an assignment but now it they are running at 3-4 weeks so I have to send the point releases out with CVE-PENDING in the changelog which isn't ideal.
Not sure what the solution is, but it is definitely a problem for us.\n— nickcw\n
Saving 100 terabytes of memory by optimizing 1.1.1.1's DNS cache
\n\n\n\n451 points | 126 comments
\n\n\n\nThis is the right way to deliver software.\nProduce working product first, validate the idea, stabilize the business, start generating profit, and then you can start optimizing your costs.
In fact optimization is by far the easiest part of the process because there are many system programming experts on this HN thread who consider these optimizations to be trivial.\n— lpapez\n
290 points | 117 comments
\n\n\n\nTried out the simulator and was surprised to find that the movement keys are ZQSD; then realized that's the equivalent of WASD on an AZERTY keyboard. Checked and sure enough, Pollen Robotics is a French company.\nThey may want to at least add a preference for keyboard layout, pretty sure that QWERTY and QWERTZ are much more common worldwide than AZERTY.\n— lambda\n
412 points | 187 comments
\n\n\n\n> But I also think the demand for \"fast/cheap/good-enough\" models is just about to take off.\nThere's a sort of \"revelation\" I had in ~early '24 when I used a 7B local model with a library called Guidance (initially out of MS, then the team moved) to create a flow where the model would receive pseudocode for tests, first write the tests, and once I approved then started writing code until the tests passed. This was before \"thinking\" models, and yet using that library I was able to \"guide\" the model in the required \"prompt / instruct\" context such that it was working towards completion, and I saw the first things like we see now in the thinking traces \"oh, test x doesn't pass because blah, I need to...\" and so on.
Anyway, the revelation was \"even if the models never improve, I'll have years of fun finding out all the ways I can use these things\". And, obviously, the models improved a lot since then. But I think that revelation can still be applied, as a sort of \"truism\". We have, right now, access to things that 10-20 years ago would be considered magic. We are still finding ways of cobbling together systems with glue, duct tape and prayers and find new things they can do.
I think the \"good-enough\" stage has come not just for API models (cheap, fast, etc) but for local as well. Even if slower, even if clunkier, but they are good enough for a set of ever increasing tasks, and what's more it's incredibly fun to work with them.\n— NitpickLawyer\n
439 points | 64 comments
\n\n\n\n\"507 Mechanical Movements\" is a rare HTTP status code, usually only used when the request can't be fulfilled due to an earthquake\n— PufPufPuf\n\n
465 points | 215 comments
\n\n\n\nAlt text: \"You legs may have a comparative advantage at running, but we arms have a competitive advantage at swinging hammers, so unless you accept that we're the dominant limbs and stop hogging the oxygen, that running advantage won't be around for long.\"\n— dvh\n\n
Nvidia agrees to acquire Hugging Face for $13B
\nhttps://www.theinformation.com/articles/nvidia-agrees-buy-op... (paywalled)https://techcrunch.com/2026/08/24/hugging-face-reportedly-in...\n\n\n
827 points | 347 comments
\n\n\n\nThis isn't a good thing for open source.\nNvidia don't share CUDA, don't open source their drivers, don't support capable but older hardware (forget Pascal etc), and generally charge a premium over competitors for hardware.
This acquisition will likely solidify this general stance, reduce free compute allowance for a subscription, cap downloads, push advertising, generally favour models that are Nvidia prescribed or have been sponsored, and potentially ban models and datasets that are deemed risky legally (abliterated etc).
I see no other reason why Nvidia would want this kind of vertical.\n— SillyUsername\n
686 points | 187 comments
\n\n\n\nThe title is misleading. AWS acquired DuckLabs, NOT DuckDB. The DuckDB source code is still owned by the nonprofit DuckDB Foundation.\nQuote from the article:
\"As the CWI representative on the DuckDB Foundation... When DuckLabs spun out of CWI, we created this foundation, which holds all IP of open-source DuckDB, and will continue to do so.\" - Peter Boncz\n— daviewales\n
436 points | 190 comments
\n\n\n\nWeights on HF here: https://huggingface.co/zai-org/GLM-5.3-Flash\nI decided to take the plunge and get myself four sparks at a decent price (and bought the QSFP cables from AliExpress because they are literally 1/2 the price of Amazon), even knowing Apple was going to release new hardware and there's probably a spark 2 on the horizon. It looks like this is going to be a decent fit for what I need. I've been experimenting with a two-node DS4 and it's _good_ at some tasks, but it really just spins its wheels when it hits the limit of what it can reason through.
I can offload mundane/basic tasks to DS4 on two sparks, but I've been pushing it harder on some novel work and it just can't run on its own at all beyond a certain complexity level.
I would love to see an Opus-4.8-level local model but TBH I just haven't got there yet. The models I've tried so far _are_ good but they aren't able to solve tough technical challenges, regardless of harness/prompting/etc.\n— mmastrac\n
Qwen3.8-Flash-Next: A New Architecture, Towards Ultimate Cost-Efficiency
\n\n\n\n396 points | 119 comments
\n\n\n\nAdding to my homelab stack, hopefully it doesn't overthink like the little model. Actually, hoping it thinks a bit less. Wait actually I'm really praying it reasons a bit more directly. But wait, I'm really sure that it must be a bit better.\n— lnenad\n\n
560 points | 188 comments
\n\n\n\nI often chuckle at the observation that Tim Curry and Michael Caine's respective roles in the Muppet films were so beloved because Michael Caine acted like all the muppets were real actors, and Tim Curry acted like he was a muppet. He would have an impressive range of serious / comedic roles anyway, but I admire the ability for an actor that successful to be able to put himself be in such a silly role that well too.\n— TallGuyShort\n\n
CEO fired developers to make room for AI. Developers create open source AI CEO
\n\n\n\n428 points | 265 comments
\n\n\n\n> The next session opens with a <past_decisions> block so the Executive remembers what it recommended last month.\nThis is morbidly unrealistic.\n— pmlnr\n
Tailcat – Like netcat, but over Tailscale’s data plane
\n\n\n\n440 points | 85 comments
\n\n\n\nI just set up Tailscale as I've been hosting more personal apps off a small Hetzner VM and wanted a really simple personal networking solution. Works great, would recommend.\n— wxw\n\n
U.S. State Department pauses immigrant visa applications
\n\n\n\n449 points | 672 comments
\n\n\n\nhttps://archive.is/cjW9v\n— OutOfHere\n\n
Meta reaches $16.68B settlement over social media harms to children
\n\n\n\n332 points | 290 comments
\n\n\n\nMeta has warped society to think that life is meaningless without affirmation.\nNo amount of money can fix that or punish them.
I think that sharing things can be wonderful, but monetizing that or adding Like counts to gamify it is very damaging to people. This idea that everyone should know better and take a ideological stand against some dumb app is ridiculous. Life can be hard at times, and these apps manipulated that.\n— shepardrtc\n
Twitter Viewer – View Twitter Without Account
\n\n\n\n296 points | 150 comments
\n\n\n\nI used Twitter for over 10 years without an account then suddenly in 2022 that was completely disrupted and no longer an option. Reddit is now this way too.\nBut I find it ridiculous that Twitter, Facebook, LinkedIn, etc. get used by government agencies and local businesses to post announcements and these sites are increasingly if not completely unusable without being logged in (and in some cases requiring an app instead of the site). Even Bluesky has this problem!
Sometimes this information is really important! And I even if I did have an account I may not have access to log in on every device (or even want to).
It’s absurd that social media can become this like critical communications platform for severe weather alerts or business hours or anything but the sites hosting it are completely hostile to allowing you to see it.
Sites like this are a bandaid on a gangrenous foot. Thanks for allowing me to see links from twitter but it isn’t a sustainable solution clearly.\n— kemotep\n
https://www.nytimes.com/2026/08/25/arts/music/dolly-parton-d..., https://archive.ph/w4vsE
https://www.hollywoodreporter.com/music/music-news/dolly-par...\n\n\n
1110 points | 170 comments
\n\n\n\nI’ve never been much of a country music fan, but I’ve always been a Dolly fan. Great songwriter, philanthropist, and seemed like a genuinely nice person. RIP, Dolly, you were one of the greats.\n— mikestew\n\n
Apple introduces M6 and M5 Ultra for a big leap in performance and AI compute
\n\n\n\n525 points | 426 comments
\n\n\n\nI know, this is a bit of a meaningless comment, but it's funny in a way. Feels like late 90s again:\n\n— bayindirh\n- Xiaomi: We have matched Apple in CPU performance.\n - Apple: *Meep Meep...*
Nitter project received cease and desist
\n\n\n\n560 points | 399 comments
\n\n\n\nThe only reason why I needed nitter to access x is due to organizations still using it as their main communication solution to their customers. Even my local council uses it. I would like this to be the final push for people to get off x but I suspect its going g to stay and I guess I don't get to see news from my local council anymore.\n— Ugvx\n\n
New Mac Studio with M5 Max and M5 Ultra
\n\n\n\n464 points | 293 comments
\n\n\n\nDue to pricing insanity (not that Apple prices weren’t insane before the ram/ssd shortages) I’m not in the market but I do wonder if my next computer should be a Mac Studio instead of a MBP that lives its life docked. Might be better to just run a Studio and Neo for the very few times I actually need remote capabilities.\n— joshstrange\n\n
411 points | 110 comments
\n\n\n\nOP here - longtime HN user. I submitted this story to a writing contest a few months back, but didn’t hear anything from them.\nI posted it here on a punt and getting front page of HN is 10x the feeling any writing contest could give me!
Thanks to all who read and commented, and glad to hear it resonates with some of you!
[EDIT]: There've been a lot more comments since I wrote this one. I want you all to know that I've read every single comment and am ever so grateful for everyone who responded here / took the time to read the story!\n— sarreph\n
New Mac mini, featuring M6 and M5 Pro
\n\n\n\n237 points | 127 comments
\n\n\n\nSo happy with my base M4 purchase for $499 when it released. I even snagged a 1TB replacement drive for $199 a year ago. Feels like I got the last train out.\n— jdoe1337halo\n\n
OpenAI Jalapeño: Better than Nvidia Blackwell
\nhttps://www.bloomberg.com/news/articles/2026-08-25/openai-cl..., https://archive.ph/yCTrr\n\n\n279 points | 194 comments
\n\n\n\nI think they talked about this being general purpose chip but I would think that Anthropic/OpenAI are at the scale now they could bake LLM weights into chips themselves.\nFor example, GPT Sol baked into a custom chip run for $100M that runs 10x as fast and 10x as cheap should pay for itself as long as the chip is useful for long enough.
While 2 years ago nothing was useful more than 1 year long, there are many older models in use now (e.g. Haiku 4.5, GPT-OSS 120b), and I expect this trend to continue.
I know this is what Taalas was doing (acquired by AMD), here was their demo, https://chatjimmy.ai/ which is based on Llama 3.1 8B. It feels like this should start to happen soon.\n— mchusma\n
FDA authorizes first wearable device that monitors ketone and blood sugar levels
\n\n\n\n354 points | 174 comments
\n\n\n\nA friend died of diabetic ketoacidosis. He loved technology and seeing people succeed. Thinking of him with this news. Miss you Dan.\nGrateful we're moving forward on so many fronts in the world.\n— madrox\n
Firefox 157 will include JPEG XL by default on all platforms
\n\n\n\n245 points | 58 comments
\n\n\n\nWith both Firefox and Chromium using jxl-rs (Rust-based), I wonder what Apple will do about the libjxl (C++) they already shipped. I know they're doing some memory-safety with Swift, but are they shipping any Rust in their platforms so far? I also wonder if anyone's done benchmark comparisons between both libs.\n--
Also, I was under the impression that after backtracking, Chromium was relying on Mozilla to come up with a Rust port, but it seems it was the reverse. Good on Google Research.
https://hacks.mozilla.org/2026/08/intent-to-ship-jpeg-xl/
> So, we laid down a challenge to the JPEG XL team at Google Research: Build a safe, performant, compact, and compatible JPEG XL decoder in Rust, and we’ll ship it. That challenge was met; Google Research built jxl-rs, and it’s the core of our JPEG XL support in Firefox.\n— concinds\n
292 points | 113 comments
\n\n\n\n12935 -> TIZZY (5625) -> OXBOW (2674) -> GRUFF (858) -> MAMMA (195) and I'll let you guess the rest but my next choice left 43 words, and I survived with only 4 words possible.\nThe rules not letting use a letter you know isn't in the target word drastically cuts your choices, your best bet is to use as few unique letters per word as you can, and as uncommon letters as you can, but as all normal words need vowels you'll be forced into only picking a vowel from the target word soon enough. Try to find letters that are in the word as late as possible so you're not locked into having to use them.
Once you are forced to use letters from the target word, avoid placing them in the right place. Sadly there's a rule not letting you use a known letter more than once in a word (once A is known, you can't write ALPHA), and a rule for not letting you put a known letter in a place you know it doesn't go (so after MAMMA, you can't have HALVE because you know there's not an A in second place, nor ALPHA because you know there's not an A in fifth place).\n— amiga386\n
How Europe is killing makers and micro-entrepreneurs
\n\n\n\n623 points | 420 comments
\n\n\n\nI don't know if this EU regulation will work or not (most likely not), but I can share how it is done in China (where package volume is orders of magnitude higher).\nChina identifies \"choke points\" most packages pass: large platforms and logistics companies. Yes, Chinese e-commerce is more centralized on Alibaba/JD/PDD etc, but there are actually much more logistics companies in China than in EU.
They also do not introduce these regulations \"from tomorrow\" (well, sometimes they do, but it's a different topic). E.g. the non-degradable plastic regulation was issued in 2020 to come in effect in 2025. This gives time to prepare and implement (unlike EU border controls mess...)
So now if something is shipped in non-degradable plastic, the logistics company is fined along with the sender. This is much more scalable and easy to implement. Logistics companies might rise their prices a bit in light of expanded work scope, but it would be peanuts compared to every individual seller.
Could EU just ask DHL/Hermes/whatever to charge an additional €1 \"for recycling\"? Yes, but...
But of course EU is fundamentally different in a way that the politicians need to please the electorate, so nothing is really planned ahead.\n— mstaoru\n
Xiaomi: New CPU matches Apple cores single threaded, much faster multithreaded
\n\n\n\n278 points | 153 comments
\n\n\n\nThis is still an ARM C1-Ultra. Same one used in Mediatek dimensity 9500, which also got more than 4000 points in GB6 lab test. But actual results inside phone with smartphone cooling and wattage requirement became close to 3300.\nBut perhaps the most important thing, Xiaomi is now able to make chip similar to Mediatek. They are the third largest Smartphone manufacturer by shipment volume. And they are still growing.
This is bad news for Mediatek and Qualcomm.\n— ksec\n
MS Paint and Photos inivisibly watermark even locally generated output with GUID
\n\n\n\n522 points | 215 comments
\n\n\n\nI admit it’s a while since I’ve used windows, but it’s such a shock to hear that MS Paint isn’t just a point and click pixel coloring app anymore. It seems like they could have left it as a pure “paint” app and added the fancy stuff to some new image editor or something.\nI guess they’ve optimized their workforce to just keep making changes so they get promoted rather than just creating really good software.\n— clickety_clack\n
Oceans hit highest temperature on record
\n\n\n\n360 points | 233 comments
\n\n\n\nBBC did a pretty good deep dive a few days ago on what we're in for: https://www.youtube.com/watch?v=_xZCh8eDZFc\nDW also, similar stuff, but the guest is quite good: https://www.youtube.com/watch?v=Dcl5Mn373Oc
Anton Petrov covered it from a much more science lens a couple months back: https://www.youtube.com/watch?v=eRoU3whzX_Q\n— neom\n
Coding expertise is going to collapse from AI reliance
\n\n\n\n426 points | 424 comments
\n\n\n\n100%\nWe're already seeing this at the enterprise level. Companies have dictates from leadership that \"if you're writing code manually, you're doing it wrong.\"
Okay, that kind of works for a while. We are indeed producing a shit-ton of code, but the reality is that engineers are pumping out code faster than the humans can understand and (honestly) review it. That sounds great until you realize that \"hey Claude, read this Jira ticket and implement the feature in this code base\" isn't really worth $200K/year.
This is all complicated by the fact that we're also losing our grasp on reality from the other direction because we have leadership air dropping AI generated manifestos on the product owners and product owners having to use AI to transmute all that shit into 1,500 word Jira tickets that are 10% necessary feature work and 90% LLM boilerplate.
So now you have software engineers whose job has changed radically to the point that the hardest part about being a software engineer is just filtering through AI generated artifacts from all directions just to try to get a feature out the door.\n— ryandvm\n
The entire city of San Francisco as a video game
\nhttps://twitter.com/cdngdev/status/2091909073038082139, https://xcancel.com/cdngdev/status/2091909073038082139\n\n\n309 points | 109 comments
\n\n\n\nThis is awesome. Would definitely download a local higher-res version if that were an option, maybe using Google Street View data? Also, if this was a live MMO, that would be fun. Not that there's really a game aspect here other than the coins when driving a vehicle, but anything to give some life to the sim SF city would be nice.\nSeemingly lower-hanging fruit options I'd like to see:
- show street names, maybe some landmarks or businesses, togglable
- enter address to teleport to
- enter address to get directions to, GPS style, to travel there manually
- get coordinates / current location as a URL to share
- zoom in on minimap, or expand minimap, so that teleporting can be more precise
- changeable avatar
- more dynamic weather / lighting options\n— cutoff\n
iCloud+ Hide My Email addresses will remain on icloud.com
\n\n\n\n390 points | 103 comments
\n\n\n\nThis is a huge selling point. Private email relays often get blocked, the only lasting solution is to put them on the same domain as, and in the same format as, a significant number of non-private email addresses. As far as I’m aware the only other provider doing this is Fastmail.\nComments about lock-in aren’t wrong, but it has to be this way. You can make arbitrary email addresses at your own domain, but anybody who feels like it can trivially automatically detect that those are all you.
Personally I use unique at own domain only where I’m identifying myself anyway, like my bank, and Fastmail masked email where I’m not. For most things it’s not actually that terrible to accept a small risk that they’re offline for a day between your being booted without warning and you changing your email address with them.\n— kqp\n
312 points | 159 comments
\n\n\n\nThe post here is pretty confusing (so no blame to anyone who reads this as \"IPFS the project\" shutting down instead of just a single maintainer team, it's totally misleading) - but this is actually just a sunset announcement for _Shipyard_ - one of many IPFS implementation maintainers.\n*The IPFS Project is not sunsetting or shutting down* - just switching to individual maintainer grants instead of centralized implementation support within Shipyard.\n— momack2\n
OpenAI: GPT 5.6 Sol price reduction (until at least Nov 21)
\n\n\n\n279 points | 256 comments
\n\n\n\nThe fact that AI models can be so easily distilled and replicated is such a stroke of luck.\n10 or 15 years ago if one had asked me to envision a future where a private company invents artificial intelligence, I'd have thought for sure they'd have a massive moat, be very difficult to catch, and it would create an almost instant monopoly.
Rather, it seems that selling intelligence might end up as a race to the bottom.
Who woulda thought that just having access to enough textual inputs and outputs and a vaugely similar transformer architecture would be enough to copy-cat rather useful intelligence.\n— eigenspace\n
Anna's Archive Owes $340 Million, Lost Several Domains, but It's Still Online
\n\n\n\n240 points | 163 comments
\n\n\n\nLong live Anna’s Archive. I stand on the shoulders of the Internet, Wikipedia, Anna’s Archive, Z-Library, LibGen, YouTube, Hacker News, Reddit, and Sci-Hub.\nI deeply admire the people who are obsessed with their passions and strive to build things that will lay the foundations for others.\n— num42\n
655 points | 201 comments
\n\n\n\nI did this but with a dedicated machine for the Silicon Motion sm750 GPU. A budget single HDMI output GPU card for servers and a max resolution of 1080p. It is based on an older VGA/DVI version of the same hardware.\nI'm still testing but oh wow. My new driver now works with my ultra wide 21:9 ratio at 2048x864, it also manages 2048x1152.
The driver works well, and now has full DRM and DKMS support. It also runs on modern Linux after the manufacturer decided only to go up to kernel 5.x, windows support obviously still fine.
It found many faults in the original source, like somebody didn't read the HDMI specs / didn't have any idea what they were doing.
The new driver is fully spec timings and sequence compliant, doesn't hang on shutdown anymore, and ignores EDID for the purpose of allowing more screen modes.
It also has double buffering, and shadow buffering, and a custom magic square dither mode for 16bit colour and it absolutely flies vs the 32bit mode. The dither I invented was derived from one I created years ago for some retro hardware, but it's so good it's (imho) indistinguishable from general jpeg artifacting and quite difficult to find/see. I've had to ask codex a few times to check the GPU isn't in 32bit colour.
The GPU still has an annoying bug and won't work over KVM consistently without losing sync in VESA modes, but I'm not convinced its the GPU hardware doing this, it works perfectly well directly connected.
I'm due to put a GitHub repo up for this as soon as it's battle tested, and obviously ensuring it uses EDID by default, rather than ignores it.
I'm hoping somebody can fix the KVM issue, or audits the source to confirm there's nothing that can be done, but that's the best thing about open source :)\n— SillyUsername\n
I spent $266 and four AI models to own my tablet. GLM-5.3 finished it in a day
\n\n\n\n155 points | 52 comments
\n\n\n\nNice to see the capabilities of the model but the article has heavy AI tones, making it boring to read.\nan AI:DR; is enough: the models found unpatched vulnerabilities and managed to create an exploit to root the tablet, chinese models did it while American ones fell back to their safeguards.\n— AntonyGarand\n
Anthropic's best AI model struggles to attract users as cheaper tools thrive
\n\n\n\n431 points | 376 comments
\n\n\n\nhttp://archive.today/ZLojz\n— pmontra\n\n
264 points | 127 comments
\n\n\n\nI’ve been working on a harness for accounting agents at my job recently and it’s been a pretty interesting experience.\nWe originally started with building a CLI tool so our LLMs could more easily interact with our platform. I cannot recommend enough the value of having an internal CLI. It’s both fun to build and extremely useful for agents.
We paired this with skills initially, but found that the way folks built skills was often too prescriptive and limited to the authors own specific function in the company. A 2k line long skill suffers from the same gaps as we do, if an agent is just following a laundry list it’s less likely to reason about the request it’s doing.
So we instead asked ourselves: what if we just _let_ the agent reason about the work to be done and only provided the tools + guardrails to gather context and perform accounting work?
Turns out frontier models are GOOD at what they do, they outperformed our highly prescriptive skills and were able to work across a larger set of tasks even without instruction on how to do those tasks.
It’s a breath of fresh air from the decade of CRUD I’ve worked on, harness engineering is very neat.\n— Syntaf\n
How I find problems to solve as a staff engineer
\n\n\n\n235 points | 85 comments
\n\n\n\nThe author notes:\n> One caveat: my experience comes mainly from working on infrastructure and developer tools at large companies, on teams where engineers have a lot of bottom-up autonomy to influence their roadmaps. In a more top-down environment, there may simply be less room to work this way.
I wonder if the overall trend in tech is that engineers are experiencing less bottom-up autonomy and more top-down controlled environments. I would be curious to see how many tech companies (or the average engineer's experience) have changed from being tech led where engineers have autonomy to being more product management led. My suspicion without evidence is that the overall engineering autonomy has decreased over the years as the culture of tech has (in my opinion) shifted away from tech focus to more business, management, product focus with engineers just as the widgets who are tasked with fulfilling the goals of business, management, product.
all hypothesis, only anecdata\n— wpasc\n
Slovakia finds Russian backdoor in traffic speed cameras
\n\n\n\n353 points | 133 comments
\n\n\n\nSo they bought the cameras and some people pointed out they look exactly the same as Russian cameras. The government denied this but after they figured out the serial numbers matched the Russian cameras they started this investigation.\nGood they investigated this before using them, but this sounds like a big fu...\n— bdhdhduuyd\n
A website for debloated open source alternatives
\n\n\n\n237 points | 80 comments
\n\n\n\nI've been a big fan of https://alternativeto.net/ and using their \"open source\" and \"self-hosted\" filters. Even looking through the non-FOSS alternatives for a particular application/site can be helpful; sometimes a for-profit company will take a different approach to solving a problem and it will give me other ideas to consider.\n— maples37\n\n
I gave Qwen 3.8 27B a reverse-engineering job and it finished in 30 minutes
\n\n\n\n210 points | 99 comments
\n\n\n\n> I gave it the hardest real task that fits on one machine: reverse-engineering a commercial app's license check...\nRespectfully, tasks that allow for explicit straightforward true/false or done/not-done tests are not the \"hardest real task[s].\" In fact, those are the ones that see the most gains from AI-assisted coding.
Testable tasks are where the largest opportunity is.\n— djoldman\n
How Complex Systems Fail (1998)
\n\n\n\n222 points | 59 comments
\n\n\n\nI'm a broken record on how important I think this document is, and that it's hard to appreciate it until you've had extended experience with complex systems actually failing.\nThe most commonly cited subtext or thrust of it is that \"root cause analysis\", at least on complex systems, is a fools errand. Something goes wrong, say, in a distributed lock system, and your whole deployment system enters a metastable failure state. Naturally, the \"root cause\" seems like lock system resiliency. But definitionally a metastable failure is one that persists after the inciting condition is resolved. Now you have two \"root causes\", the lock failure and the metastability of the deployment system fault. Keep looking and you'll find more.
But to me the biggest brick to the forehead in this piece is further observation that random things are failing all the time in any complex system. \"Complex systems run in degraded mode\". Resilient components are good, but it's the resiliency of the overall process that orchestrates the whole system that determines whether things are going to blow up.
All practitioner actions are gambles. I should have that inked somewhere.\n— tptacek\n
My agent.md to improve LLM-assisted code quality
\n\n\n\n283 points | 111 comments
\n\n\n\nA bunch of these should be enforce with linting, that way people who still hand-craft code get the same kind of feedback, e.g. Always use {}, even on a one-line \"if\" statement. & Keep function names short. Less than 30 characters.\nThen this one really is a pattern that creates a lot of churn:
- Add a small, to the point, comment to explain what the block does and why. Use examples when possible. Propose ASCII drawings to explain complete systems.
The what _is_ the code.\n— OptionOfT\n
Canada suspends trade negotiations with USA and match tariffs dollar for dollar
\n\n\n\n641 points | 598 comments
\n\n\n\nThe world should've collectively recognised what Liberation Day was and responded together, in lock step, and in kind.\nInstead, the Europeans caved, Japanese caved (PM resigned after doing so), Republic of Korea caved (PM resigned after doing so), India of course caved and so on, shattering any prospect of collective action which might pull the rogue hegemon back into line.
Now, the US is simply able to renege or renegotiate the bilateral deals whenever he feels like it, forcing the other party to face down the US by themselves - not many can do this, maybe only China. Anyways, I wish the Canadians well and I hope they gain global support\n— hunglee2\n
ElevenLabs, TwelveLabs, ThirteenLabs
\n\n\n\n286 points | 93 comments
\n\n\n\nhttps://www.twelvelabs.io/blog/introducing-the-multimodal-ai...\n> Twelve Labs and ElevenLabs are co-hosting the 23Labs Hackathon\n— progbits\n
265 points | 105 comments
\n\n\n\nOne convenient fiction that many wealthy people (in the US) tell themselves, is that poor people are \"lazy\"; and that's why they're poor.\nI know lots of folks without a pot to piss in, or a window to throw it out, and not one single one, is \"lazy.\" In fact, a few have more than one (difficult) job.
I am old enough to see a few of them getting into the back nine, with no cushion.
I'm extremely grateful for what I have.
[EDIT] Added \"many\" because it's more appropriate, and short-circuits the implied \"all.\"\n— ChrisMarshallNY\n
Why your local LLM feels dumber than it is
\n\n\n\n282 points | 95 comments
\n\n\n\nI saw some guy streaming how he was deploying qwen3.8 37B on his local setup. Well, he was asking Claude to do it. It took him two hours of passing errors to Claude for the endpoint to start working, he then started testing it against DS4 Flash when Qwen had thinking disabled and Claude messed up sampling parameters, it was an absolute pain to watch\n— big-chungus4\n\n
A Friendly Introduction to Racket
\n\n\n\n220 points | 115 comments
\n\n\n\nMy first language, 1980 in CMU, was Pascal... quickly followed by LISP in CS/AI courses and C for EE courses. I was a big fan of MacLisp at the time, the MIT version that GLS worked on (could see his comments all over the sources). Of course GLS moved on to Scheme and wrote a few definitive papers on closures. Yeah, that was before PC's everywhere... a loooong time ago. (C++ came years and years afterwards, not a fan. I would have stayed with hardware if the ++ version of C was forced on me all those years ago.)\nGLS moved to CMU in the early 80's... I remember taking \"Comparative Programming Languages\" from him. Good teacher and impressive guy. In addition to the obvious, he also covered SNOBOL and APL in that course. Memories :-)\n— SA9G\n
A Kantian Critique of \"Sorry\" by Justin Bieber
\n\n\n\n193 points | 83 comments
\n\n\n\nThis 3 minute song creates a fleeting portrait of a man, not necessarily Bieber, who we cannot know. The song captures the mindset of a character who, one might infer, cynically knows what he’s trying to do. It’s possible he has no interest in a heartfelt, meaningful apology. In fact, this character - who lives only within the song and the liminal world it creates - may be operating exactly to his own moral preference. We don’t know enough to say more. It’s like asking if a singer of “happy birthday” truly wishes their subject to be happy.\nIn short, you Kant analyze pop music with confidence.\n— mwillis\n
Hook, hold, harvest and hide: Meta's alleged strategy laid out in first week
\n\n\n\n199 points | 161 comments
\n\n\n\nIn the opening sentence:\n> according to a lawyer who is prosecuting the world’s largest social media company.
The headline makes it sound like something internal at Meta got revealed. The catchy four-H phrase is the clever hook the lawyer came up with to try to hold the jury’s mind through the case.
If you’re unfamiliar with how these trials work, the lawyers are trying to create a compelling narrative to support their case and catchy phrases that sound like marketing lingo are often used. This wasn’t some internal Meta doc that got revealed.\n— Aurornis\n
194 points | 130 comments
\n\n\n\n> With the 2026-07-28 release, a remote MCP server is now no different from any other HTTP workload\nGood. Introducing a bespoke new protocol was one of the more bone-headed things MCP did on initial release.\n— rco8786\n
Anthropic appears to be A/B testing reduced effort levels in Claude Code
\n\n\n\n181 points | 167 comments
\n\n\n\nWhatever Opus 5 is doing should not happen.\nPrompt was \"read and update the config file with new data\". This work on 4.6 takes <2 \nminutes to read the file, parse the new data, and patch.
Opus 5 Result: 43 minutes of pulling containers, running sandboxes, creating testing suites, which included evaluating the entire repo beyond the scope of the config file.
Both: one file modification\n— pizzafeelsright\n
hdiutil is deprecated in macOS 27 Golden Gate
\n\n\n\n179 points | 73 comments
\n\n\n\nGiven xip has been deprecated for a long time now, yet still is the format Xcode is distributed in, I highly doubt hidutil will ever disappear from macOS. Apple just won’t really update it anymore, just like they haven’t changed xip (not like hidutil has seen any significant changes, but still)\n— nrabulinski\n\n
Kagi added a setting for removing paywalled links from search results
\n\n\n\n563 points | 198 comments
\n\n\n\nI think this is amazing. Love Kagi. I’m happy to pay for a good search.\nNobody speaks about their AI Assistant, but it is really good. They somehow harnessed it so that it mainly searches info first and sticks to the verifiable data. I prefer it to Claude or any other, because it actually answers the question I need without fluff or “Great question! Here’s some plausible nonsense you can trip over instead of doing actual research!”.\n— delis-thumbs-7e\n
Felony charges for citizen deleting phone data at US Border
\nhttps://archive.ph/SflVChttps://www.youtube.com/watch?v=_2rokxux5cU\n\n\n
485 points | 671 comments
\n\n\n\nThe decoy passcode feature should boot into a separate partition that looks like a normal phone setup, and during that time quietly erase the user's actual data. They would never have known if it worked like this.\n— gblargg\n\n
468 points | 215 comments
\n\n\n\nThe way that OpenAI has communicated around the HuggingFace incident makes me feel crazy. You created a machine that undertook a malicious campaign of harm against an innocent third-party! You should be doing deep introspection about how your company culture and approach to R&D produces criminal outcomes.\nInstead, they treat their own felonious behavior like it is an uncontrollable act of God. From Greg Brockman's post a few days ago:
> The OpenAI-Hugging Face incident (opens in a new window) was a watershed moment for cybersecurity because it gave a peek into how the capabilities of a typical threat actor will evolve in upcoming months.
I suppose if OpenAI burns someone's house down with a drone, that is a \"watershed moment\" for arson, too. Either way, I would hope that the people responsible would be prosecuted.\n— rfw300\n
AI companies destroy physical books – let's scan rare books before it's too late
\n\n\n\n646 points | 384 comments
\n\n\n\nI don't see any mention of Project Ocean - AKA Google books. Before AI they endevoured to digitize books in a massive online library. This inccluded rare and out of print books many which are archived at libraries. Because they had to preserve the books and return them in the condition they received then they created elaborate technology to accomplish this. The project was met with significant legal challenges from authors and publishers which was eventually overcome. The legal precedents that were established from Project Ocean laid the ground work for the process as it exists today. Books from libraries are still being preserved.\nhttps://en.wikipedia.org/wiki/Google_Books
https://arstechnica.com/tech-policy/2015/10/appeals-court-ru...
https://arstechnica.com/ai/2025/06/anthropic-destroyed-milli...\n— thread_id\n
Grand jury declines to indict Ohio man charged with destroying Flock camera
\n\n\n\n454 points | 245 comments
\n\n\n\nFor those that don't know, grand juries declining an indictment is extremely rare. A grand jury is basically a check on prosecution, that they have to have some initial evidence before charging someone with a felony. The standards are much lower than the subsequent criminal proceedings.\nThe grand jury only hears from the prosecution, there is no defense involved. Only a majority of the grand jury has to sign off, not unanimous like in an actual trial. The standard of evidence is just probable cause, not beyond a reasonable doubt. The rules of evidence are relaxed, meaning hearsay and other evidence can potentially be introduced that would normally be barred from a trial.
Because of the above, the rate of indictment from a grand jury is very high, over 90%. Most prosecutors will go their entire careers without getting a \"no true bill\" (meaning the grand jury did not sign off on an indictment). There's a saying that \"a grand jury would indict a ham sandwich.\" So the fact that there was no indictment here is a big deal. It will probably hurt that prosecutor's career.\n— openasocket\n
384 points | 135 comments
\n\n\n\nFor the unaware, there is an existing solution that integrates with Kobo's native software (Nickel).\nIt has been maintained for years, and it supports every Kobo AFAIK.
It's called NickelMenu, and it's great. I'm in the Kobo ecosystem because of NickelMenu and Plato.
One note if you're thinking about getting a kobo due its relative openness: consider getting a two-core device. I did not do my homework and got a Clara BW because I don't want the color display. Later I found out that color is the only one with the two-core CPU.\n— the-grump\n
I accidentally logged hundreds of thousands of phone calls to military bases
\n\n\n\n411 points | 45 comments
\n\n\n\n> It never really took off though, and even back in its early days it saw barely any use. Over the years it just deteriorated further, and today it's basically completely dead.\nIt's actually not completely dead... It's just (almost) completely non-public.
You can subscribe to services to get number porting information where the interface is basically e164.arpa/ENUM queries to a private nameserver over a VPN. I don't know the details, the cost was high enough that it didn't make sense for my employer to pursue it.\n— toast0\n
350 points | 116 comments
\n\n\n\nDS being unable to precisely view Playwright screenshots is the only thing I really miss from Sonnet. This is promising.\n> Images are converted into tokens based on their dimensions, and these tokens are billed together with your text tokens.
> Before inference, every image is automatically resized:
> - Images with a total pixel count below roughly 384×384 are scaled up while preserving their aspect ratio.
> - Larger images are scaled down while preserving their aspect ratio so that the total pixel count after resizing is roughly that of an 800×800 image.
> As a result, there is an upper bound of 384 tokens per image: for example, a 2000×2000 image and a 5000×5000 image consume the same number of tokens after resizing. When a request contains multiple images, each image is counted independently under the same rule—there is no separate calculation for multi-image requests.
400 tokens per image results in 2,500 images per dollar, if I’m not mistaken.
edit: format.\n— ciberado\n
There's no reason for software to be slow anymore
\n\n\n\n313 points | 218 comments
\n\n\n\nOne of the biggest causes of slowness is just waiting for web requests. The fact that so much software is either online or built using the same stack even if it isn't, puts all that software in this blocked/waiting state constantly while using it.\nAnyone not in the US feels this even more since so much online is US hosted, 300ms for every little interaction adds up quick.
If your software has the affordance of a waiting dialogue or loading wheel for many of its UI controls, you are building with this default blocked assumption. Even if you are building something web based, ask yourself if that's actually necessary for your software or if you could build it differently to avoid constant UI blocking.\n— ehnto\n
242 points | 250 comments
\n\n\n\nI've been stuck in loop all week where a coworker asked me to review some methodology documents that are clearly generated by Claude (which is fine in this case) but every time I sit down my brain can't get through 2 sentences without feeling like I'm reading nothing. It actually feels like the content, which isn't empty, has no meaning. So I go on to another assignment and come back later and the same thing happens. And now I have anxiety about opening it so it has compounded in difficulty haha. Glad other people are experiencing this.\n— tlkn_bot_praxis\n\n
Aaron Swartz was prosecuted for scraping, while Meta does it without consequence
\n\n\n\n739 points | 158 comments
\n\n\n\nThe part that still bothers me so much about the US vs Swartz case is that JSTOR didn't pursue civil litigation against Aaron. It was the US government that pursued him.\nThere was little for the government to lose in the case. In a case vs Meta, at the scale it has reached, it could have wide ranging economic implications limiting the investment in AI, which the US is absolutely not willing to pursue at this point in time (or possibly ever).
Basically, being a rich public company provides legal advantages when the US government has similar goals.
The whole thing is incredibly sad and exposes the hypocrisy of the US court system and government as a whole.
RIP Aaron.\n— milkytron\n
AliExpress runs silent WebAudio fingerprinting that breaks Bluetooth multipoint
\n\n\n\n601 points | 201 comments
\n\n\n\nI wish such shenanigans would simply trigger the little speaker icon most browser display on tabs these days.\nGiven that they don't (at least in my experience), I'm assuming \"playing silent audio\" is a sufficiently common thing for websites to do to have motivated browsers into doing the slightly more complicated thing of actually analyzing audio streams for content...
Now I wonder, does this also allow websites to continue running in the background on mobile browsers? Playing media is one of the very few things that can convince iOS Safari to keep a tab running indefinitely, in my experience.\n— lxgr\n
Show HN: I trained a 125M model to autocomplete piano on-device
\nI trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15).The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device.
The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.\n\n\n
320 points | 74 comments
\n\n\n\nReally cool! It reminds me of a demo I saw live at NeurIPS: https://www.youtube.com/watch?v=8s3V922h3CU (paper at https://openreview.net/pdf?id=3yeBer3J5z).\nIf you like such creative AI work, I would recommend looking at the other \"NeurIPS Creative AI Track 2025\" submissions as well.\n— mare5x\n
The August 17 outage, and the work ahead
\n\n\n\n256 points | 297 comments
\n\n\n\n> Errors in those services triggered a client-side retry loop that increased traffic during recovery\nSymptomic of a wider trend to avoid showing the user any error at all costs, even if that means they sit watching a spinner for 7 hours.
> Delayed replies to a single internal endpoint triggered a latent retry bug in VS Code that amplified traffic by approximately 10x and caused delayed recovery for the Copilot Token Service.
The detailed root analysis tries to pass this off as a \"bug\". You can't seriously tell me client retry doesn't have a unit test which ensures the retry back off behaviour is functioning exactly as designed. In this case aggressively to try and hide problems if token service responses become flakey.\n— cube00\n
Malicious Rust crate Arrayref runs a build-time payload
\nhttps://blog.rust-lang.org/2026/08/20/supply-chain-attack-on...https://github.com/rustsec/advisory-db/issues/3161\n\n\n
372 points | 352 comments
\n\n\n\nGitHub really needs something finer-grain then just pretending the repo never existed during these incidents. [1]\nThe bad package version has also just disappeared from crates.io [2] with no indication its been yanked. There's no security advisory there either [3] \"No advisories found for this crate.\"
I feel crates.io was unprepared for a security incident like this since they're managing the response [4]
[1]: https://web.archive.org/web/20260820145918/https://github.co...
[2]: https://crates.io/crates/arrayref/versions
[3]: https://crates.io/crates/arrayref/security (I'd give an Wayback link but that's also broken https://web.archive.org/web/20260820150747/https://crates.io...)
[4]: https://github.com/rustsec/advisory-db/issues/3161#issuecomm...\n— cube00\n
AI companies destroy physical books – let's scan rare books before it's too late
\n\n\n\n269 points | 199 comments
\n\n\n\nI dislike these AI companies but let's be clear here: the copyright holders are the ones locking these books up. If they don't want to print more copies, then they could release the copyright on them.\nInstead, they enforce the copyright and force AI companies to shred books they want to ingest.
edit: Also, an AI company would only ever care to purchase, scan, destroy a book once. Presumably many books have more than one copy.\n— ezfe\n
Watching TikTok and Instagram deactivates the cognitive control network: Study
\n\n\n\n289 points | 111 comments
\n\n\n\ndlPFC deactivation is observed in a lot of immersive tasks. There are a lot of tasks that do this. Playing video games, as one example https://link.springer.com/chapter/10.1007/4-431-26797-2_11\nInterpreting fMRI results like this article does is almost always wrong. Different activities produce different patterns of brain activity. You could put someone in an fMRI machine, have them do virtually any activity, and then write a headline saying that doing that activity turns of key brain regions that are important for something. The trick is that all brain regions are important for something. Having them be less active during a task doesn't imply that it's good or bad. It just shows you which parts of the brain are involved in that task.
This blogspam article isn't great. The study at https://www.sciencedirect.com/science/article/pii/S105381192... has more details, like the fact that those brain regions come online while viewing other less-preferred videos. The results aren't really indicative of anything by themself. It would have been more interesting if they compared to something, like watching a long involved movie or playing video games, but they didn't.
Also this post is coming from a fresh account that only spams links to this blog. Maybe the mods will want to switch the link to the actual study instead: https://www.sciencedirect.com/science/article/pii/S105381192...\n— Aurornis\n
Show HN: Huzzah – a novel approach to coding with AI
\nHello everyone. I've been working on this experimental editor called Huzzah.I've been working almost exclusively with coding agents since January of this year, and over the past few months I began to feel utterly exhausted by them. They're great, but I'm finding it more and more tedious to write full sentences for every change I want. Not only that, but it seems there's a complexity limit for codebases - beyond a certain point the agent begins confusing itself.
I'd like to go back to writing code, but I don't want to go all the way back to fully manual coding. So I've come up with this interaction paradigm where you:
1. write pseudocode in whatever way makes the most sense to you\n 2. on save, the editor synchronizes your work to real source code\n 3. the pseudocode is persisted alongside the generated code, making your prompt effectively a stored record of intent.\nRight now it's just a proof of concept - installation instructions are here in the readme: https://github.com/danielvaughn/hz
You can also watch a video of it in action here: https://x.com/danielvaughn/status/2090456808431165715
Cheers!\n\n\n
283 points | 151 comments
\n\n\n\nI think you’re probably missing why it’s exhausting. The problem is not writing English, it’s the rate of change. Programming is meditative, it is a thinking process, the code you output is an artifact of your thinking. Agent-based development… there is no thinking, no meditation, you’re delegating the thinking to a machine, you’re just barking what you want at it, incessantly, endlessly.\nFor businesses it makes sense to abandon programming in favor of delegating to agents that can do more in less time, but for programmers, it is a loss. Either be a programmer and code, or be a delegator and delegate, you aren’t going to make the life of a delegator suck any less by trying to trick yourself into thinking you’re programming.\n— reticulates\n
I should have loved biology (2020)
\n\n\n\n251 points | 98 comments
\n\n\n\nWe run a microschool and our single most important realization is this. We need to allow for meaning before mechanics. Schooling has a gravitational pull that inverts this because of our obsession with Measurement.\nWe learn, and we can learn relentlessly, when things matter. We are meaning makers. But when well-intentioned curriculum remove the process of making meaning and give you the final product, it just takes away reasons for learning. And no matter how we dress it up, this ends up staying shallow because it is only the surface level learning that is visible and measurable and held to account.
Meaning > Motivation > Mechanics > Measurement
This is the natural order we should try to preserve despite all the systemic pulls to measure as a poor proxy for asking \"are we doing our job well\"
Wrote about it here: https://blog.comini.in/p/schooling-has-a-meaning-crisis-para...\n— ChaitanyaSai\n
267 points | 49 comments
\n\n\n\nIt's funny how so many of the articles are hyper-specific grievances, e.g.:\nhttps://consumerrights.wiki/w/Bose_QuietComfort_Sleepbuds_do...
https://consumerrights.wiki/w/Tyre_warranty_sold_via_mobile_...
And there's also https://consumerrights.wiki/w/Mr._Clinton_the_cat\n— matherial\n
A joke domain purchase turned in geopolitical warfare
\n\n\n\n331 points | 43 comments
\n\n\n\nIt's fun to see habhub in an article. ~10 years ago a group of friends and I launched 2 weather balloons for fun with a GPS logger, APRS transmitter, and a few sensors, and then retrieved it afterwards.\nI have a bunch of memorable experiences from it, including:
1. Under inflating the first balloon because our helium provider was closed, and instead buying lower pressure \"party\" balloon containers (and not realizing until too late that lower pressure meant more helium left in the containers)
2. Chasing down the balloon at night and trying to explain why you wanted to walk onto some guys field to check for a weather balloon
3. Later calling a gas station to ask if \"some guy just arrived with a weather balloon in their truck\" when trying to track down the balloon
If anyone has kids and wants a fun project, I highly recommend it. There are a few companies online that sell kits, and it's neat to watch the different layers of the atmosphere (temperature and wind), and try to \"chase\" after the balloon and find it afterwards.\n— xur17\n
568 points | 306 comments
\n\n\n\nGreat product, been using it for a while.\nTurns out even a proxy can be worth $8bn with the right business model behind it.
Users get an array of providers competing behind a single API, meaning they have to compete on price and quality not vendor lock-in. This encourages users to join OpenRouter over specific model vendors.
Providers get easy access to revenue (and data) and new customers with little to no ad spending, encouraging them onto the platform too.
And that's all you need. Win win.
Well done and congratulations.\n— apexalpha\n
403 points | 94 comments
\n\n\n\nI really want to like Go, but I can't stand looking at Go code. The error handling is such a turn off.\n— Fervicus\n\n
Devices with GrapheneOS support should be available in 2027
\n\n\n\n453 points | 273 comments
\n\n\n\nSpecific devices:\n>At the time of writing, within ~12 months, in 2027, the 2027 Signature, Razr fold, and Razr flip will meet the hardware security requirements and should have official GrapheneOS support.\nMotorola is currently porting GrapheneOS to their devices.
https://news.ycombinator.com/item?id=49038982\n— Cider9986\n
Remote workers report the highest well-being in study of 7,700 employees
\n\n\n\n463 points | 233 comments
\n\n\n\n…of a single company\n> Researchers analyzed survey data from 7,704 employees at a large healthcare organization.
I’ve been remote and worked for remote companies for a long time. In my observation, the well-being is bimodal. The people who adapt well to remote thrive. There are a lot of people who think they’re going to love remote who ultimately flame out from the loneliness, isolation, lack of visual boundaries between work and home life, and lack of externally enforced routine.
It’s really sad when the latter happens because it’s almost always someone who was hired for their great work and abilities. Then over time they get drained and you can see that the remote work wasn’t what they expected.
It’s basically impossible to predict who will fall into each category at the interview stage, or at least I can’t do it. Prior remote work success (not just experience) is the best indicator but it’s hard to evaluate, too. Some of the most difficult remote hires I’ve worked with had past remote work experience, but it’s only after being hired that you learn they struggled at those companies too.\n— Aurornis\n
Moderna reports first positive Phase 3 for mRNA neoantigen therapy in melanoma
\n\n\n\n289 points | 117 comments
\n\n\n\nThis is great news! I think people forget how anti-skin cancer protocols even as simple as applying suntan lotion weren't as popular as recent as 50-60 years ago. Lots of \"sun children\" of the 50s and 60s are now feeling the repercussions of only applying baby oil and are getting hit with lots of melanoma.\n— GenerWork\n\n
Google has stopped pushing Git tags for some Android source code
\n\n\n\n447 points | 188 comments
\n\n\n\nThey stopped pushing tags for any of the Pixel kernel or userspace driver repositories to AOSP. They also stopped pushing AOSP releases specific to Pixels which is why AOSP now only gets yearly releases, QPR2 releases and security backports to both of those. Other OEMs use the yearly and theoretically also the QPR2 releases. Both the yearly and QPR2 releases get monthly security backports. Since they dropped Pixel support from AOSP, they don't push the releases not shipped by other OEMs anymore.\nThese changes directly led to our Motorola partnership. One of their security people reached out to us after seeing our posts about this with the launch of Android 16. We haven't talked about it much since then since we adapted to it during the several weeks it delayed our Android 16 port. We then continued adapting to it and have fully worked around it. It was an ongoing problem but not a new one and we had accepted we had to deal with it as the new normal.
They were previously responding to our kernel source requests within a day. It was often done without hours. Despite the archaic system, this part wasn't that bad. Recently, they've been taking weeks or longer to get back to us for the requests which is ridiculous. It's the direct result of purposely adding a lot of friction with manual handling of the requests even if the delays weren't directly planned by management.
Weeks or months of delay is not reasonable for one of the largest tech companies in the world. GPL doesn't set a standard time limit for providing the sources, but that doesn't mean they can delay it indefinitely. They need to do it in a reasonable amount of time. What's reasonable for one of the largest tech companies in the world in 2026 with current technology is not the same as what was reasonable 30 years ago. Google chose to come up with a archaic way of distributing the sources involving someone manually going through a list and sharing Google Drive access. It's a deliberate way of making it painful. If they can't keep up with it and it gets delayed for weeks or months then they're not complying with the GPL by not providing it in a reasonable amount of time. Law is not code and a time limit not being explicitly written down doesn't mean there isn't a limit to what's reasonable for compliance.
They'll sell far fewer Pixels because of these overall changes. It pushes GrapheneOS and other projects towards other devices instead. For us, Pixels are being used due to security rather than ease of supporting them. It's now a lot harder to deal with Pixels than it would be for many other devices but they're currently still the most secure option. We're working on changing that and have a lot less reason to contribute to improving Pixels. We helped them fix serious security weaknesses for Pixels including vulnerabilities being exploited in the wild by forensic data extraction companies. Pixel security with the stock OS would be worse without GrapheneOS.\n— grapheneos\n
Geolocating a random island using geometry and CUDA programming
\n\n\n\n389 points | 74 comments
\n\n\n\nExcellent write up and an enjoyable read! Reminds me of the “good old times” where posts on HN were written by humans and with a specific writing style like yours.\nYou could’ve used a little bit more of geoguessing to narrow down results, or do a brute force visual check on the last hundred or so ;-)\n— NKosmatos\n\n
Civic Hygiene – avoid building technologies that could be used by a police state (2013)
\n\n\n\n400 points | 290 comments
\n\n\n\nHarder than it looks. Everything is dual use. Certainly didn't expect my game development telemetry[1] system used to improve artillery [2]...\n[1] https://gdcvault.com/play/1012227/Development-Telemetry-in-V...
[2] https://www.army.mil/article/85934/solution_to_weapon_develo...\n— gmerc\n
340 points | 283 comments
\n\n\n\nCasio has been leaving a lot of money on the table in terms of nostalgia products. There's been a low rumble of demand for the Casio CZ series of synthesizers for a decade or more, and a bunch of non-Casio companies have cashed in on it (Behringer makes CZ-1 Mini, Arturia has CZ V, several other phase distortion synths exist). It's incredibly cheap to replicate, a simple early digital synthesis algorithm that'll run on anything; Korg has been making ARM-based emulation synths (basically a Raspberry Pi-like SoC in a keyboard form factor with some knobs) for a decade or more and I have to assume their margins are very comfortable since most of the guts are off-the-shelf low-cost parts rather than custom chips.\nBut, I guess their watches were always a bigger business than the synths.\n— SwellJoe\n
418 points | 293 comments
\n\n\n\nThis has been true for years on Amazon and is true for most big services now.\nOver time the concept of search has mutated from 'locate the exact item I'm looking for' to 'show me a list of semantic search results'. \nSomehow nobody noticed.
Oh and of course those are not random results - you can literally feel how the platform is trying to nudge you into buying what IT wants, not what you wanted. \nThe search is just purchase intent, the platform will choose what you'll buy.
Powerful algorithms, subliminal design - I feel like a lamb in a dark forest full of wolves when I search on amazon, netflix, prime video ..
There are no '0 search results' anywhere.. each search yields thousands of results.
I remember what a revelation Google search was when it came out in 1999 - it would actually find the exact thing you were looking for ! You no longer had to sift through directories of links, you could just filter it..
We're back to pre-google, but it's a lot worse now, since this is by design, not by lack of technical knowledge of how to make search work.\n— delegate\n
342 points | 68 comments
\n\n\n\nFor Linux, there is https://github.com/pwr-Solaar/Solaar\nBut on another note, the genai content on the website is just so distracting and such a bummer. It sticks out like sore thumb.\n— asdfsa32\n
Google has acquired the data of failed US airline Spirit
\n\n\n\n450 points | 320 comments
\n\n\n\nAbout twenty years ago, I was taking a flight back from Rio de Janeiro, Brazil to the US. In the middle of the night the pilot got on the loudspeaker and said \"hi! Having some engine trouble, so we are landing in Manaus.\"\nManaus is in the middle of the Amazon.
Needless to say, a bit scary to hear that, but we landed without issue.
They told us we had two choices: the nice hotel with a shared room, or the lesser nice hotel with no roommate. I chose the latter. When we go there, they said, \"oops, sorry, short on rooms!\" So I had a roommate.
Wandered around Manaus, took a skiff out on the Rio Negro. Saw pink river dolphins. A little boat approached us and a kid handed me a sloth, and then demanded I return it with a twenty dollar bill.
The airline got us another plane 24 hours later. Made it back to the US safely.
A few weeks later, the airline reached out and said \"Here is $100 for your trouble.\"
I declined to take that offer. I had missed several business meetings that cost me actual money. I couldn't donate blood for years because I had been to the Amazon and was tagged a malaria risk.
During the many arguments with the airline I threatened to take them to small claims court.
I got a really strange response over email which I clearly wasn't supposed to see. A representative from that airline was asking internally if they could put me on the no-fly list. That was really chilling.
But, this is the kind of information I'm worried about when a vendor sells my data. If Google wanted to sell a product to the airlines that offered to keep annoying people like me from purchasing flights, they could do that with that email chain. I'm skeptical it'll be wiped correctly. Isn't my poor writing style basically my signature? How do you wipe that?\n— xrd\n
421 points | 111 comments
\n\n\n\nI spent a few years on and off consulting working for two of the Big 4. I can’t speak for all the teams but the ones I worked with provided good value. Our job was to protect our customers from poor designs and lack of accountability. Some of the projects I worked with were very big involving dozens of suppliers. Our customers didn’t have the capability to coordinate so much work and they also lacked the know-how and expertise in many of the domains. Also many of the big engineering and tech companies out there can be very sloppy. Don’t think because they carry a recognizable name (I don’t want to name them but they are SP500) they always do good work. I have seen stuff that is really really bad, like unsafe bad. That’s where my team would step in and fill those roles on the customer side. Many of the projects I participated were also highly regulated, and that’s a whole additional layer of complexity for which you need people with experience in very particular areas. A bad design that leads to an environmental disaster or even a violation of regulations can be extremely costly.\n— tedggh\n\n
Using the railway network as a flatbed scanner
\n\n\n\n381 points | 58 comments
\n\n\n\nWard Cunningham and I did something similar back in 2008.\nWe were both at a startup in Portland and our office was along the railroad tracks east of the Willamette. We were up on the 4th or 5th floor right above a lot of train traffic including Amtrak.
I brought in some extra Mac gear I had including one of the early iSight cameras which back then was an external device you stuck on the top of your display and connected to the Mac via FireWire. We set it up and rolled the desk over to the window in the office and turned the camera around pointing down at the tracks and Ward got to work hacking something up to do a slit scan.
It was a fun little project; the speed of the trains affected the horizontal compression of the images. We could tweak the software to expand or contract the size of the image by adjusting the number of slit scans per unit of time. Of course, that affected the exposure but I recall the camera having some automatic adjustments that gave us trouble.
That's about all we did with it. Just a quick afternoon hacking project. Neither of us thought much about it at the time so we didn't save anything.\n— msisk6\n
Fixing a bricked Framework laptop
\n\n\n\n341 points | 231 comments
\n\n\n\nThanks for posting this. IT comes along at a perfect time for me. I am investigating linux laptops for one of my children, still in college. I have been looking for machines that come out of the box with linux and Framework was up there on the list. Others include Dell, Lenovo, ASUS, and Tuxedo.\nI like the apparent upgrade ease of the Framework since I have a pile of old laptops, tablets and netbooks dating to late 1990's already. If I can find one that is long-term upgradable then that is a big plus.
I had almost convinced myself that Framework is the way to go with their hot-swappable ports and easily replaced internals.
I found Tuxedo not long ago. They seem to be a valid competitor in that space. I know they're a German company but I think they ship to the boondocks in Texas.
I read that BIOS recovery procedure and was reminded of all the times I had to flash and then recover on desktops or laptops and how much fun it wasn't. The fact that Framework treats you like a second-class citizen after the warranty expires is very valuable information in a case where your problem is caused by their recommended procedure. The hack of their partner company feels so relevant today when companies can still leave PII open in public datasets and suffer no consequences.
I will more strongly consider Dell and Tuxedo. I have long experience with Dell workstations and have found that even though a lot of their components like Mobos are proprietary formats requiring a Dell part to repair, that those parts are identifiable and generally available thru Dell or eBay.
I don't know enough about Tuxedo so if anyone has experience I would appreciate some insight. The use case for us is typical document authoring, programming in python with maybe Jupyter notebooks, complicated spreadsheets, photo documentation of field sites, and GIS work maintaining project map resources using ArcGIS most likely. Gaming is probably another option but it is not necessary for it to run top games though it does need decent graphics.\n— doodlebugging\n
Fairphone is now officially available in the United States
\n\n\n\n341 points | 177 comments
\n\n\n\nLooks great! For other T-Mobile users, it actually supports the full range of frequency bands. Quite a few devices claim T-Mobile support but are then missing band 12 or 71 on LTE or n71 on 5G, all of which are critical for coverage outside of cities.\n— threemux\n\n
Sticky wage norms and the real wage cost of unexpected inflation
\n\n\n\n347 points | 203 comments
\n\n\n\nThe submitted title (\"New paper shows that 37% of workers in US saw real wages decline from 2021-2024 [pdf]\") broke the site guidelines, which ask:\n\"Please use the original title, unless it is misleading or linkbait; don't editorialize.\" =- https://news.ycombinator.com/newsguidelines.html
Submitters: If you want to say what you think is important about an article, that's fine, but do it by adding a comment to the thread. Then your view will be on a level playing field with everyone else's: https://hn.algolia.com/?dateRange=all&page=0&prefix=false&so...\n— dang\n
How does IKEA come up with names for its products?
\n\n\n\n312 points | 183 comments
\n\n\n\n> Every possible product name is carefully checked to avoid undesirable meanings in other languages, political or religious affiliations and fit into our global profile\nIf I remember correctly, they had a desk named \"JERKER\" in early 2000s.\n— slybot\n
And then the men with guns tell you to do it anyway
\n\n\n\n318 points | 207 comments
\n\n\n\n> But it is easy to see why some people feel this might open the gateway to messages which, at best, are irrelevant\nIf you're in South Korea, you can expect to get around half a dozen emergency notifications every day. Typical contents include \"it's cold, so dress warmly\", \"roads are icy, so drive carefully\", and \"Mrs Kim, 79, suffers from dementia and was last seen in Seoul wearing a black coat\".
You can set your phone to filter these out, but then you'll also miss the notification about North Korea commencing artillery bombardment.\n— decimalenough\n
532 points | 324 comments
\n\n\n\nThe part that astonishes me is that in the year of our common era two thousand twenty-six that it's not universally offensive and reviling to post an AI-generated response to another person.\nIf I'm reading something on the internet, I'm either reading it to learn, or I'm reading it to be persuaded. If I wanted the LLM to teach me (thank you, no), I would ask an LLM. I'm reading your website/newsletter/email because I want to hear from you.. If you can't be bothered to put your time into writing it and teaching me what you think, why should I be bothered to read it?\n— gortok\n
689 points | 2 comments
\n\n250 points | 33 comments
\n\n\n\nSuper excited about Quack (partially due to the name). I use duckdb for both analytics and runtime, but I do have to serve/handle/manage a giant, multi-GiB duckdb file as effectively a runtime artifact[1]. I'm aware that this isn't the _perfect_ database for this, but the mix of it being fast, having spatial support, sane coding interfaces, great dbt integration, and me being able to do everything between \"run a giant several hundred step dbt pipeline\" to \"query the output of said pipeline\" to \"read/query a csv on disk\" with the exact same tool is just so nice. If I could centrally manage said asset more akin to a traditional database, I'd be very happy.\nI've partially solved this with separate databases for different steps in the data pipeline(s) and have even experimented with Clickhouse as a complete alternative, but I'd really like way too many things about duckdb to replace it.
[1]: If you care: https://skaldmaps.com/blog/2026/07/zip-codes-are-a-bad-spati...\n— otter-in-a-suit\n
216 points | 132 comments
\n\n\n\nIt depends on what you are after?\n1. Do you want something that works and feels like GitHub? -- Forejo and Gitea are good for this.
2. Do you want a place to host git repositories with minimal hassle? -- GitLab, CodeBerg, and others are available.
3. Do you have your own hosting infrastructure? You could use gitolite and CGit/GitWeb on that hosting platform or local hardware.
4. Do you just want to host repositories? -- Gitolite can be used to help with SSH/auth/repository creation, and CGit or GitWeb for the frontend.
5. Do you need something like GitHub Actions? -- GitLab, Forejo, and Gitea offer CI, or use external CI infrastructure.
6. Do you need issue tracking and management? -- GitLab, Forejo, and Gitea provide these. There are alternatives from Jira to Kanban (including Trello) to Markdown (Obsidian and others) and more.\n— rhdunn\n
Edit: at the time of posting there was not an incident on githubstatus.com. Now there is. https://www.githubstatus.com/incidents/zkxwbgr0cnmx
Original Title: \"Tell HN: GitHub Is Overloaded\"\n\n\n
507 points | 879 comments
\n\n\n\nI think this is the curse of scale, but also of pseudo-leaders who think leadership is \"pushing engineering to rapid-fire features for number go up.\"\nYou tell engineers and engineering leaders to \"just add this one feature, it's really important, what's the ETA? No, that's too late, the market will have moved on, this is existential, you're a bottleneck,\" fire, hire, fire, hire. Meanwhile the CEO feels great, because they're delivering. The reports look nice, the posts get written, the lectures get given on X and LinkedIn about what steering a ship really means.
Then the issues that required your system to actually be robust, the ones that required engineers to have spent the time engineering, show up. And it's way too late. You cannot move faster, no matter how much money you throw at it.
It gets bad enough that everyone is now telling the CEO 70% of the platform needs refactoring, and even that is like solving the three-body problem, because every domain boundary is broken. Claude can't reason about any component of the system but is very confident it understands it, and ten seconds later tells you \"You're right! And I won't sugarcoat it, I completely misrepresented this concept, let me go fix that by rewriting this other module.\"
You can't rebuild GitHub. You can't rebuild components of it in reasonable time over live data and migrate over without very public incidents. So now the CEO feels very, very dumb, and can't admit it.
Meanwhile a new company capitalizes on your lack of leadership and launches a competitor...and the cycle starts anew.\n— figassis\n
How Bluesky draws its logo on screenshots
\n\n\n\n412 points | 277 comments
\n\n\n\nIf it's between this and a perpetual logo, I'll take this any day.\nI actually really like this approach. The action button isn't relevant in this context, and it doesn't occlude the content.
There's certainly situations where you wouldn't want this (ie if you're developing the app and you want to redesign starting from a screenshot), but for the average user I think this isn't overly hostile. I understand that people are dogmatically opposed to intent being modified, but I think you need to balance nuance. I actually enjoy having an attributable source in shared elements, and I think this is a low-impact way of achieving that.\n— jjcm\n
Israel creates fake think tank in likely attempt to dupe AI chatbots
\n\n\n\n358 points | 223 comments
\n\n\n\nI suspect that this kind of tactic is going to be everywhere in a year or so. Entire fake personalities and organization websites on the Internet created just to push a narrative or to advertise a product, which completely drown out real information.\nAt some point they may be indistinguishable from human-produced work, so the only way to verify if a site is someone's genuine opinion or AI-generated narrative is through some form of authority (or something else that is very difficult for AI to fake, but that protection can be broken by advancing technology).
If that happens, then in turn AI chatbot makers would in turn need to look for these kinds of authority to continue to provide accurate information, and said authority websites can use their privileged position to charge for access of their data.\n— 2001zhaozhao\n
GPT-5.6 Sol Pricing Cut by 50%
\n\n\n\n355 points | 188 comments
\n\n\n\nWhere is the official source for this?\nOpenAI's docs still show non-discounted pricing https://developers.openai.com/api/docs/models/gpt-5.6-sol\n— resonious\n
Universal Health Coverage Could Save $1T and 114k Lives a Year, Yale Study
\n\n\n\n221 points | 275 comments
\n\n\n\nI'd like to believe this, but the study makes a bunch of really hasty assumptions.\nThe authors derive the $1T number from $1.3T in total cost savings and $304B in incremental spend (incremental spend is due to insuring more people). The $1.3T in cost savings come from five big buckets: lower pharmaceutical prices, Medicare-level payments to providers, reduced administrative overhead, less fraudulent billing, and fewer avoidable emergency department visits and hospitalizations.
The buckets themselves don't necessarily survive much scrutiny.
Take \"Medicare-level payments to providers\". Hospitals have an operating margin of 2-5%. Medicare pays 50% less than private insurance. So doing this would require either layoffs, cutting salaries for doctors/nurses/etc, or both. This may well be the right decision for society as a whole--that's a big part of the debate here--but there's no free lunch.
The line item of \"fewer avoidable emergency department visits and hospitalizations\" assumes greater insurance coverage leads to greater access to primary care. It's true that great primary care prevents hospitalizations, and can be a net cost saving under certain assumptions [1]. But, we're actually in a primary care shortage. Existing insurance payments for primary care are low enough that private practices are going out of business and fewer residents are going into family medicine. Cutting rates (the paragraph above) would make this worse.
For \"less fraudulent billing,\" a lot of people in the industry believe that Medicare has a large amount of undetected fraud. That's unfortunately the flip-side of reduced administrative overhead. The authors assume an 8% savings here, but the 2003 paper they cite uses the word \"fraud\" only twice and doesn't give a number.
Healthcare reform is hard.
[1] Reasonable breakdown on the economics of advanced primary care models: https://olearykm.medium.com/the-cost-equation-for-new-primar...\n— brandonb\n
AI-Generated GitHub Copilot “Autofix” Allowed Compromise of Snowflake's Jira
\n\n\n\n305 points | 123 comments
\n\n\n\nI probably would have made the same mistake. It is negligent to write GitHub Actions without using static analysis.\nUse zizmor in CI https://github.com/zizmorcore/zizmor
\n— inahga\nerror[template-injection]: code injection via template expansion\n --> .github/workflows/jira_issue.yml:24:29\n |\n 22 | run: |\n | --- this run block\n 23 | # Escape special characters in title and body\n 24 | TITLE=$(echo '${{ github.event.issue.title }}' | sed 's/\"/\\\\\"/g' | sed \"s/'/\\\\\\'/g\")\n | ^^^^^^^^^^^^^^^^^^^^^^^^ may expand into attacker-controllable code\n |\n = note: audit confidence → High\n = note: this finding has an auto-fix
272 points | 131 comments
\n\n\n\nI have a folder where I rebuild these as a git commit history so you can more easily see what has changed: https://github.com/simonw/research/commits/main/extract-syst...\nFor example here's what changed between Opus 4.8 and Opus 5: https://github.com/simonw/research/commit/a2de185cc367eb66c2...
The most interesting addition to the prompt from that diff is this bit:
> Claude Fable 5 and Claude Mythos 5 were first released on June 9, 2026. On June 12, 2026, Anthropic suspended access to both models to comply with U.S. Department of Commerce export controls; the Department lifted those controls on June 30, 2026, and Anthropic restored access on July 1, 2026 (Anthropic's statement: [https://www.anthropic.com/news/fable-mythos-access](https://www.anthropic.com/news/fable-mythos-access)). These events are after Claude's training-data cutoff, so Claude knows about them only from this notice. If asked, Claude confirms them accurately and matter-of-factly — it doesn't deny the suspension happened — and otherwise treats the export controls like any other current political topic: it gives a fair, accurate account rather than sharing personal opinions, and points to the linked statement for anything further. Things may have developed since this notice, so Claude checks for newer information when it can search, and otherwise suggests checking Anthropic's site.
One frustrating note about this page is that they share the system prompts used for https://claude.ai and the Claude mobile apps regular chat, but they omit the tool definitions. Those are much more interesting if you want to understand what Claude can actually do for you. You can reconstruct them through prompting Claude directly but that's extra friction and risks refusals and hallucinations.
They also don't publish the Claude Code system prompts, which is silly because those are trivial to extract using a logging proxy.\n— simonw\n
Firefox for iOS now has a native adblocker
\n\n\n\n249 points | 98 comments
\n\n\n\nIn case anyone here missed it, there's a native Ublock Origin for Safari. It works great → https://apps.apple.com/us/app/ublock-origin-lite/id674534269...\n— internet2000\n\n
240 points | 58 comments
\n\n\n\nAn alternative: <meta http-equiv=\"Content-Security-Policy\" content=\"script-src 'self' https://only-scripts-allowed-from-here.com\">\nThis makes the client only load self-hosted scripts, or scripts only from the specified origins, among the other directives CSP allows (e.g. restricting styles, images, frames, etc.): https://developer.mozilla.org/en-US/docs/Web/HTTP/Guides/CSP\n— okzgn\n
A 3rd World Embedded Engineer Responds to \"RISC-V They Should Have Known Better\"
\n\n\n\n324 points | 170 comments
\n\n\n\nI think he's kind of speaking past the original author. The original piece is basically about how the author doesn't think that RISC-V will take off outside embedded, because of some design decisions that lead to poor performance compared to ARM64 and because so much of the ISA being optional means that there's too much fragmentation to make binary distribution feasible. Meanwhile, this piece is mainly about how RISC-V is great for embedded because companies can build it into custom chips with specifically the functionality they need, and because of how cheap it is for low-end use cases since there's no license fees.\nThe only real point of contention I see between the two is that this piece goes on to talk about how it's a selling point that RISC-V can be used for both low-end 10 cent microcontrollers, and high-end multi-core processors running Linux. Personally I don't see the benefit of this since you're going to have to recompile your software anyway, and since all the RISC-V SBCs I'm aware of have significantly worse performance and efficiency than comparably priced ARM SBCs.\n— ndiddy\n
Qwen 3.8 27B is excellent, but it defaults to overthinking things
\n\n\n\n323 points | 132 comments
\n\n\n\n“The fact that a 17GB file can do all of this stuff on my home machines is a miracle. Once again, I’m delighted and amazed at how much progress local models have made this year.”\nI think that should be the blinking headline - this shows what can be done with consumer hardware.\n— chvid\n
Research papers using \"kidney disappointment\" instead of \"kidney failure\"
\n\n\n\n208 points | 83 comments
\n\n\n\nHere is one hypothesis: https://theconversation.com/problematic-paper-screener-trawl...\n<quote>\nHave you ever heard of the Joined Together States? Or bosom peril? Kidney disappointment? Fake neural organizations? Lactose bigotry? These nonsensical, and sometimes amusing, word sequences are among thousands of “tortured phrases” that sleuths have found littered throughout reputable scientific journals.
They typically result from using paraphrasing tools to evade plagiarism-detection software when stealing someone else’s text. The phrases above are real examples of bungled synonyms for the United States, breast cancer, kidney failure, artificial neural networks, and lactose intolerance, respectively.\n</quote>\n— aix1\n
Stripe will reportedly acquire OpenRouter for $7B+
\n\n\n\n296 points | 198 comments
\n\n\n\nTo people asking why, this is a good lesson on the Collison’s ambitions. Stripe is one of the best API companies in the world. They know how to serve high volumes of latency and availability sensitive requests. They’ve abstracted the financial rails for payments and now want to abstract the rails for LLMs.\nThey’re the perfect company to own OpenRouter.
Tokens are simply a lightweight valuable asset. Stripe can serve as the middleman as well as anyone. They know how to route to many providers (payment rails) with huge differences in service characteristics. LLM providers are far easier.
Then they can work this into an offering where users can subscribe to tokens and use them across services. It solves one of the core monetization challenges of every AI company: how do you price when your costs are variable on usage, but nobody can make sense of charging by token?
From here, they can start hosting their own models and competing as an AWS for tokens. They can be the best provider of $OPEN_MODEL, or their own, and optimize for you.\n— tyre\n
Models Are Getting Dumber on Purpose
\n\n\n\n226 points | 134 comments
\n\n\n\nIdeally what I'd like to see is pluggable knowledge bases.\nSo if I'm e.g. coding a SwiftUI app for navigation, I'd take 9B of basic coding and reasoning, add 10B of swift/swiftUI, add 5B of GIS/geography knowledge and another 5B of frontend app design knowledge. My model doesn't need to know a single line of python.
Then when I want to research electronics components, I grab a 15B model of agentic research techniques, and add in 10B of electronics knowledge, etc.
I don't want general purpose models. They try to be everything to everyone. I want to click together a model that is laser-focused on what I am doing, and I want to run it locally\n— kennywinker\n
217 points | 83 comments
\n\n\n\nThe original article linked in the opening has more context https://vectoral.com/blog/token-relay-market\nPeople trading their unused credits feels more genuine, although still in violation of the agreements. The person who got into YC Startup School who was trying to resell the $2500 of credits was interesting. It wouldn’t be that hard for OpenAI to identify the IP addresses of the relays and start flagging accounts, tracing it back to the source. Risking burning your bridges with YC for a relatively small profit is a questionable decision.
The original article showed discounts ranging all the way up to 98%. At those levels it’s obviously not people reselling anything. It’s either sourced from stolen API keys, bought with stolen credit cards, or acquired through automated sign up of trial accounts if you’re actually getting the API you request.
I would expect a lot of them are reselling a different API. Sign up for Anthropic tokens and get Deepseek responses instead.\n— Aurornis\n
Anthropic's 'Watermark' Text Adulteration in Claude Is a Perversion of Writing
\n\n\n\n176 points | 174 comments
\n\n\n\n> I want any LLM I use to choose the very best, most precise words at every single decision point.\nThen bad news: LLMs already use randomness in a fundamental way. Each time they go to generate a token, they first generate a probability distribution of possible tokens. Then they pick one randomly according to this distribution. The technique described can be thought of as making the random number generator pseudo random. The output it generates is one of the possible outputs it would have generated before, just now it's deterministic and will generate the same thing every time.\n— syrrim\n
AI has access to a vastly larger working memory than the human brain
\n\n\n\n377 points | 333 comments
\n\n\n\nI suspect that a lot about what we call being very intelligent is ultimately out-remembering people around us. I think of all the times in my software career when I did something that others considered very high performance, it either came down to either having more energy than others at tackling a problem they thought was more trouble than it was worth, or just bringing back random knowledge from previous jobs or self study, and being able to apply it to the problem at hand.\nI don't think I've had a truly original idea in my life. Combine A + B, when it's rare for people to know A and B at the same time. So from that perspective, what LLMs are doing is basically the same thing. Sometimes I am faster than the LLM because my context might be better organized, but it typically needs just a hint from me to steer itself correctly. It claims something is a memory leak, but smelling a rat, I suggest it to double check the garbage collection statistics too, at which point it's clear it's no leak, but a tuning error, at which point the LLM is better at tuning than me, because it has more energy than I do.
Maybe there's true brilliance out there, when something doesn't come out of combining data and building hypothesis until you get really lucky. My experience is not comprehensive. But I look around me, and it sure seems I've not been lucky enough to see it. Even the shiniest people I've worked with, which most of the audience here would recognize, have never shown me that they can go past this.\n— hibikir\n
Semaglutide linked to lower predicted dementia risk
\n\n\n\n322 points | 226 comments
\n\n\n\ntdlr: This is a Novo Nordisk-funded study focusing on predictive biomarkers rather than real-world dementia cases. Novo Nordisk's actual dedicated clinical trials for Alzheimer's completely failed to show that semaglutide stops cognitive decline.\n\"A predictive biomarker is like a \"check engine\" light on your dashboard. It warns you that there is a risk of a future problem. In this study, the researchers only checked if the drug turned off the \"check engine\" light (by measuring blood proteins), rather than testing if the car was actually driving properly (by testing the patients' actual memory and brain function).\"
Always do FIRST analysis on studies. Or have AI do it for you. I used Gemini to dig into this:
\"Novo Nordisk funded this study, and several of the researchers are employees or minor shareholders. While corporate funding doesn't automatically mean the data is fabricated, it does mean the company is highly motivated to find and publish data that makes their blockbuster drug (semaglutide, marketed as Wegovy, Ozempic, and Rybelsus) look like a preventative treatment for a wider range of conditions, expanding its market and driving up profits.\"
\"Funding: The study was funded by Novo Nordisk A/S.
Investigation: Researchers conducted a post hoc analysis using data from the randomized, placebo-controlled SELECT trial. They applied the Dementia SomaSignal Test (dSST)—a 25-protein risk score—to non-fasted serum samples collected at baseline and at week 104 to estimate 5-year and 20-year all-cause dementia risk in patients receiving semaglutide (2.4 mg) versus a placebo.
Results: Semaglutide significantly attenuated the progression of the dementia risk signature. Compared to the placebo group, the 5-year predicted risk increased 2.5-fold less (a 26.0% lower predicted event rate) and the 20-year risk increased 1.67-fold less (an 8.8% lower rate). Semaglutide also lowered the odds of patients moving into a higher dementia risk category by 36%.
Subjects: The analysis included 2,970 older adults aged 65 and older (mean age of ~69.7 years) who had overweight or obesity and cardiovascular disease, but no history of diabetes. The cohort consisted of 814 women (27.4%) and 2,156 men (72.6%).
Time: The study evaluated data over a 104-week (2-year) follow-up period. The analysis was published on August 8, 2026.\"
And then map the weakness to each respective letter if you want to dig deeper.\n— bariswheel\n
Auto-research with codex: How I achieved a 232x Faster Kernel
\n\n\n\n239 points | 67 comments
\n\n\n\nIn the last couple of days I wanted to try out the new definitive DeepSeek v4 releases. I gave it the repository of a semi-abandoned video compression codec and I told it to perform the usual benchmark -> profile -> verify -> research -> improve loop. I specifically chose this codec because the authors include a verifier for the bitstream to make sure you don't break stuff if you want to try your own implementation. I gave the agents access to the compiler's profiler and also Intel's VTune, which has fantastic output. In a couple of hours the LLM generated SSE and AVX implementations of the compression and decompression algorithms that almost doubled performance with a single core. Then I asked it to create a CUDA implementation using NVIDIA's NSIGHT profiler as a guide and it also started doing some good work.\nPersonally, I believe that LLMs should be treated like an advanced version of Prolog or linear programming: you give the constraints, you have a way of verifying correctness, and you give it a clear goal. If the LLM can verify itself and course-correct you can basically leave it on autopilot\n— Almondsetat\n
Working with AI feels more like leadership than coding
\n\n\n\n254 points | 166 comments
\n\n\n\nThe word is \"management\", not \"leadership\". This comes across as a LinkedIn post filled with vague notions and weak writing.\nThe conclusion also completely contradicts a previous point, which is that managing an LLM is not like managing a human. So the skills are, in contradiction to that LLM-ism of a conclusion, new. The author isn't using their people management skills, they're using new LLM-management skills. They think the two are similar, but didn't bother breaking down how they're the same vs where they contrast. It's just a lazy observation expanded out to a short essay that says nothing interesting.\n— miyoji\n
Abdominal fat predicts heart disease risk better than BMI
\nhttps://pmc.ncbi.nlm.nih.gov/articles/PMC13470613/\n\n\n227 points | 168 comments
\n\n\n\nRelatedly, there is evidence that certain types of \"resistant starch\" can help reduce visceral fat. This starch comes from green bananas, potatoes, legumes, etc. It has to be either raw (there are supplements for this) or cooked and cooled.\n\"Resistant starch intake facilitates weight loss in humans by reshaping the gut microbiota\"
https://pmc.ncbi.nlm.nih.gov/articles/PMC10963277/
Edit: Ah, HN submission 2 years ago: https://news.ycombinator.com/item?id=39592367\n— anon84873628\n
At-home test for infected ticks could improve Lyme Disease diagnosis
\n\n\n\n198 points | 70 comments
\n\n\n\nUnfortunately, there are major issues with this approach.\nThe vendor claims \"lab-level accuracy\" (conveniently omitting the actual accuracy numbers), but the test is a lateral flow test, which means its Limit of Detection is going to be orders of magnitude worse than molecular tests. Tick tests do not require FDA clearance, so their claims are most likely unreviewed.
Existing lab tests for ticks are almost universally based on PCR.
Ticks are small (especially nymphs) and extracting enough material can be tricky even for molecular testing, so I can't imagine an LFT test working terribly well, especially for nymphs, which are responsible for a lot of Lyme infections due to being easy to miss.
The other, bigger problem is that the test result is mostly useless.
The risk of Lyme transmission from an infected tick ranges rougly from <1% for a fresh bite (<24h) to ~10-25% once the tick is engorged (72h+). A positive tick test result would be insufficiently predictive of Lyme disease transmission on its own to justify taking antibiotics in most cases in the absence of other symptoms.
And you can't trust negative results either due to limited sensitivity, so you can't rule out transmission either.
In my opinion, tests like this can lead to worse overall health outcomes due to unnecessary treatment or withholding prophylaxis from someone who otherwise qualifies.\n— lima\n
Super El Niño Keeps Growing as New Forecasts Reach Record Territory Ahead Winter
\n\n\n\n176 points | 98 comments
\n\n\n\nThe strongest El Niño ever caused a massive famine:\nhttps://en.wikipedia.org/wiki/1877%E2%80%931878_El_Ni%C3%B1o...\n— izend\n
207 points | 71 comments
\n\n\n\nThe author Paul McCann (polm) is one of my favourite programmers out there!\nHe’s done awesome work in the Japanese NLP space over the last decade which has really helped me in my language learning projects.
He maintains a mecab (Japanese tokenizer) wrapper for Python [0], has a book on Japanese NLP written for English speakers [1] and also worked on Spacy at one point [2].
[0] https://github.com/polm/fugashi\n[1] https://www.japanesenlp.com/\n[2] https://spacy.io/\n— joshdavham\n
968 points | 623 comments
\n\n\n\nCredit where it's due. Qwen 3.8 27B is only the second local model after Gemma 4 that managed to correctly reason through one of my private benchmarks. It took 5x as many tokens to do it and 12m30s with MTP enabled, but it did do it.\nGemma 4 reasoned through it more implicitly, while Qwen 3.8 reasoned more explicitly. Laguna and Muse Glimmer failed hard on it, though they're useful for other tasks.
The VRAM usage seems way less efficient than Gemma 4 or Glimmer though, with 32K of context taking 2.5GB of VRAM. With those, even with MTP or a DFlash model loaded, you could still fit 256k-768k of context. With Qwen 3.8 27B I can't even fit 128k if I quantize V to Q4_0. Maybe with some trial and error I can find some settings that perform well enough with a larger context window that it's still useful for longer tasks.
Lots more testing to do, though I was getting some decent results out of Muse Glimmer which was more than twice as fast and supported huge context windows, managing to solve some bugs that Gemma 4 struggled with. I can't even begin to throw that task at Qwen, because just the prompt alone would use the entire context window and then it would reason for probably that same amount.
If you've got a 32GB card, it should be a decent model even if it really is memory hungry.
EDIT: Tried a few kv cache quantization settings, but it failed with those. I designed this benchmark to be pretty brutal in the face of KLD and any reasoning quality loss, so it's not too surprising. Gemma 4's QAT held up pretty well, at least and could consistently complete it.\n— CMay\n
Firefox is now the last major browser that still supports uBlock Origin
\n\n\n\n567 points | 209 comments
\n\n\n\nFirefox is also the only browser that vets uBlock's code on every update to make sure the developer hasn't inserted spyware or malware into the extension.\nThey don't do it for every extension, but they do so for a wide selection of popular options.
> Recommended extensions differ from other extensions that are regularly reviewed by Firefox staff in that they are curated extensions that meet the highest standards of security, functionality and user experience. After receiving Recommended status, safety standards are maintained through automated checks, monitoring, and periodic technical reviews
https://support.mozilla.org/en-US/kb/recommended-extensions-...\n— GeekyBear\n
Why does Opus 5 feel worse to work with?
\n\n\n\n810 points | 744 comments
\n\n\n\nThe single biggest annoyance with Opus 5 is that it writes too elliptically.\nSentences that orbit a point, then jump to it like it's a revealed insight.
Unnecessarily abstract phraseology. Constantly using inanimate nouns as the subjects in sentences in order to unlock variety in verb choice, especially when it helps construct a sentence where the real action can 'land' like a surprise at the end.
It is definitely more capable, and yes, I've found it can make unwarranted decisions, but actually I've found Fable worse for that, particularly if it's off in a subagent somewhere out of sight.
And comments are out of control. I have a subsystem in my hobby app that I wrote over a couple of weekends with Opus + Fable. After ~30 or so commits it apparently started instructing subagents to copy the \"existing verbose comment style of the codebase\" - a verbose style it initiated. A review of the code showed it was approaching 3:1 comments to code ratio. I spent a day's worth of tokens (5x) rephrasing and eliminating comments.\n— barrkel\n
761 points | 449 comments
\n\n\n\nShould load much slower.\nAlso, where is the unrelated autoplaying video that will unmute if you actually click it, that follows your scrolling and only becomes smaller when you dismiss it? Plus, it should probably have text that cuts off letting you know you can have access for just $10/month.
Plus, isn't this website undissmissably \"better in the app\" after a few minutes of attempting to use it on a phone? Where's that at?
edit: Oh shoot! I forgot, too. This modal needs to also ensure there is absolutely no way to scroll. If you could scroll you might be able to accidentally get to the address bar of your browser to fix the URL to xcancel or even close the page, which isn't using the app as you are intended to do.
Also, it doesn't attempt to hijack the back button to give me stuff I clearly wanted to see before I leave the page.
A lot of work left to do here before it's a \"real\" website. Although, it has about as much substance as the average website so far, so good work on that.\n— jchw\n
Count Binface receives over a quarter of votes in Clacton by-election
\n\n\n\n436 points | 345 comments
\n\n\n\nSome of his planks:\n\nI see the attraction.\n— delichon\n* Cut your taxes, and raise everyone else’s.\n * Nationalize Adele. \n * Build at least one affordable house.\n * Hold a referendum on whether Pluto should regain its planet status.\n
Google is making private AI practical with homomorphic encryption
\n\n\n\n309 points | 181 comments
\n\n\n\nMy master's thesis is on a topic in this field (Privacy Preserving ML) and from my understanding HE and other techniques have very high overheads(~10^3) on inference tasks and thus aren't very commercially viable.\n— sabretooth1405\n\n
Seven books I keep close because I love them
\n\n\n\n314 points | 140 comments
\n\n\n\nThe New International Version (NIV) translation of the Bible is by far the worst translation on almost any metric you can think of, except maybe understanding how evangelical Christians read the Bible in the 90s and 2000s.\nThe NIV has so many inexcusable translations that reinforce a theological worldview that matched the conversation on atonement and justification that was happening in the 90s and 2000s. There's a great catalog out there of the most egregious translations in the NIV [1]. One such example:
> Romans 7:18 — The NIV here translates σάρξ (sarx) as “sinful nature” even though this implies later Augustinian doctrine on original sin that is not intended by the original writer. In contrast, the NRSV correctly chooses to translate this tricky Pauline term more literally as “flesh”.
Are you interested in a readable Bible that flows well to modern English ears? The New Living Translation (NLT) is fantastic and very underrated. I recommend reading the Psalms in the NLT since much of the language of the Psalms being in a higher register in the NIV and other translations doesn't land as well, but reading the Psalms in NLT makes it much more relatable and human. I find the way it reads refreshing, and it's based on more modern scholarship.
Are you interested in latest scholarly understanding of the Bible? The NRSV is the consensus version (as much as there could be one).
Are you interested in understanding the Bible as American Evangelical Christians read it today? That's currently the English Standard Version (ESV), which many churches have switched to as the NIV has lost favor in Reformed circles.
[1]: https://isthatinthebible.wordpress.com/articles-and-resource...\n— ntdef\n
Going Dark, and the era of law enforcement hacking
\n\n\n\n244 points | 128 comments
\n\n\n\nIt's worth realizing that, before computerized central offices, telephone wiretapping required running physical wires. Back when Rudi Giuliani was prosecuting organized time, not only did physical wires have to be run, the cops were billed for them as expensive private lines. His task force was spending about a million dollars a year with New York Telephone on wiretapping.\nIn one case, law enforcement didn't pay their bill, resulting in the person being wiretapped having the wiretap connection show up on their bill, blowing the case.\nThat resulted in the Communications Assistance to Law Enforcement Act, which mandated that central offices offer remote wiretapping. Capacity up to 1% of lines is required.
Back in the electromechanical era, the only call data that could be collected was outgoing dial pulses, using a \"pen register\".[1] (The one shown in Wikipedia is mine. It's a beautiful piece of antique brass telegraph technology. It records dial pulses as dashes, and has to be wound up like a clock, with a big brass key.) The Supreme Court decision allowing \"pen registers\" without a warrant refers to these \"extremely limited\" devices. That definition has been stretched and stretched by law enforcement into all non-voice data collected by telcos.
Law enforcement still wants more.
[1] https://en.wikipedia.org/wiki/Pen_register\n— Animats\n
In Australia, a home battery boom has helped cut wholesale power prices
\n\n\n\n327 points | 273 comments
\n\n\n\nIt started with a solar boom, many small home scale solar roofs popping up everywhere. Australia has a free trade agreement with much of the world, including China, and solar panels have literally dropped to 1/50th of the price they were in 1990 ($10/W to $0.2/W today). A shout out to the work that was done to establish dynamic grid pricing too.\nAnyway that caused power prices to reliably go negative during the day as the solar boom led to too much energy being produced. So everyone started buying batteries (you can even get live feed in/out pricing as a consumer). In fact the government even today will pay you a $3000 subsidy to go install a battery. This is in a country where people can buy cheap batteries with no tariffs (free trade's amazing, seriously!). So everyone who could started doing it. For those in apartments etc. that couldn't easily install solar and batteries they won too since the entire power grid is now half the price.
Another consequence of all this, aside from the cheap power prices during a datacenter boom and Hormuz blockade is that fossil fuel usage is plummeting. Particularly gas https://ieefa.org/resources/slump-eastern-australia-gas-dema... . No need for a gas peak power plant when the grid is packed with batteries. Which is helpful since one of the main issues with the current blockade is a lack of gas globally.\n— AnotherGoodName\n
305 points | 23 comments
\n\n\n\nThis is also recommended reading if you want to learn more about building LLMs from scratch: See:\nTrain your own LLM - https://github.com/angelos-p/llm-from-scratch
HN Discussion - https://news.ycombinator.com/item?id=48017948\n— iamnotarobotman\n
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
![https://image.non.io/d83664c1-5807-4a18-abe4-41928c198410.we...](\"https://image.non.io/d83664c1-5807-4a18-abe4-41928c198410.webp\"){kind=link}