Top 30 stories from Hacker News. Top 10 include comment highlights. Compiled at 20:10 UTC.
696 points by gslin · 391 comments
You gather your most expensive people into a room to make your most important decisions. Then, somewhere in the second hour, the room quietly gets worse at making them. Not the people. The room.
What HN said:
gpt5: I really wish a Apple or another major OEM would integrate CO2 monitor into watches or smartphones. Suddenly, everybody would be aware of the CO2 level in the room, get alerts, etc. and the problem will just solve itself.
sarchertech: If you want citations search my post history, but CO2 cognitive impact studies have a replication issue. We’ve been studying the impact of CO2 for decades, at much higher levels than you see in office buildings and have never recorded any cognitive impact (until many thousands of...
vertnerd: As a high school teacher, I first noticed this effect when I started using a CO2 monitor in my classroom as a proxy for air freshness during COVID. The CO2 levels in our supposedly "no problem with the air" classrooms shot up to 2000 ppm within minutes of the start of school and...
deanc: I’m not saying this isn’t a legitimate concern but this really seems to have exploded amongst the tech community as the next obsession. I see this pop up on X every few weeks.
378 points by tylerdane · 175 comments
You can learn new things. Pixel art, touch typing, 3d modelling, music, calligraphy, wood working, knitting, a language. Whatever is practical and calls to you, you can learn.
What HN said:
frankie_t: It's a very common thing to blame the lack of time and "finding" the extra time by suggesting to give up phone or some other form of procrastination. But in my experience, time is almost never a problem.
mordechai9000: The best thing for being sad," replied Merlin, beginning to puff and blow, "is to learn something. That's the only thing that never fails. You may grow old and trembling in your anatomies, you may lie awake at night listening to the disorder of your veins, you may miss your only...
HexPhantom: One thing I wish this emphasized more is that adults often confuse learning with consuming material about learning, which is why my useful rule has become: if I'm not producing errors, I'm probably not practicing yet
jitbit: I live in Europe, had to get my permanent residence permit and for that - you have to pass the language exam. Boy, that was some experience! I'm in my 40's so learning a new language has activated the brain muscles that were last active when I was...
337 points by programLyrique · 93 comments
The most powerful AI platform for enterprises. Customize, fine-tune, and deploy AI assistants, autonomous agents, and multimodal AI with open models.
What HN said:
InsideOutSanta: There's a lot of criticism of Mistral being unable to compete with large model, and that's fair. But I think it dismisses what Mistral is actually doing, which is making specific capabilities available at high quality in tiny models.
boulos: This is nice work, but I found the bug finding example to be weird: > One such bug was in the sign function for zigzag decoding of the datrs/varinteger library. On input Std.U64.
andai: Halfway thru the article it shows a comparison with several frontier-ish LLMs. But they're all from half a year ago. "Our new model is better than all these Chinese models from 3 generations ago" is pretty funny to me.
Groxx: >One such bug was in the sign function for zigzag decoding of the datrs/varinteger library. On input Std.U64.MAX, the expression (value + 1) overflowed, causing crashes in debug mode and silent corruption in release mode—an edge case that testing and fuzzing would typically miss.
334 points by latchkey · 132 comments
The fastest open source LLMs for enterprise.
What HN said:
minraws: Can you folks add performance per watt as a metric to these comparisons, I honestly want to understand where AMD fits in the stack in terms of actual performance to dollars.
hassaanr: While cool, quantization to FP4 is practically never lossless in actual use. A lot of providers are advertising high TPS on Kimi and GLM, but the models are functionally lobotomized and no longer close to frontier quality. Would love to see this not be true.
nxtfari: I think we should make it illegal to not specify the quantization in the headline for these types of posts.
p1esk: There’s noticeable accuracy degradation when they switched from fp8 to mxfp4
314 points by javxfps · 147 comments
What HN said:
Mg6yDfjp5U: I recently left Google having worked on a number of projects with various YouTube teams. I think I can explain why it's being handled this way by YouTube. This is a fairly nuanced/involved issue, so the task of classifying the bug likely made it's way to one of the engineers resp...
CMay: In the example provided of leaking a private video, you already need access to the private video to even comment on it. That scenario is not much of an exploit. Unless there's a better example of what can be abused, the more realistic concern is authority laundering where a comm...
wxw: > Attacker leaves the comment on a creator's video. > Creator opens YouTube studio's comment tab. > Creator clicks a suggested AI prompt (Designed by YouTube) > Injection fires, attacker-controlled content appears in the response.
b-kf: bit meta but can I just applaud the article? Descriptive title, immediately comes to the point, no elaborate fluff, factual... what a nice change of pace. 95% of other users finding this would have done much worse.
313 points by theanonymousone · 41 comments
Explanation of everything you can see in htop/top on Linux
What HN said:
imrehg: I've relatively recently migrated over to using btop[0], and it's the kind of modern interface, useful and informative, that I needed. As others mention it - it seems to shows the Watts used as well :) (and network, and GPU, and disks,....) [0]: https://github.
cogman10: 2 Settings I change on every htop which makes a HUGE difference. 1. I disable user threads. Those mostly just clutter up the htop view while providing no useful information. 2. I enable the process tree view.
AnotherGoodName: I appreciate the note on virtual memory not being reliable. This is what Windows task manager reports by default and it's terrible. Resident size is the most reliable metric.
cb321: A different usage paradigm from *top that I have come to like better is to do differential ps-like reports and system-wide (like vmstat) reports which leaves everything in your terminal scrollback buffer as in: https://github.
235 points by chatmasta · 114 comments
What HN said:
throwaway260704: Using a throwaway account for obvious reasons, but I’m very involved in this space using LLMs from multiple providers. I’m aware of at least two instances in which the intermediate infrastructure “swapped” responses, once impacting Claude models and once impacting GPT models, fro...
dofm: Just add a line in AGENTS.md that says "never talk about Minecraft unless you're explicitly asked", I'm sure it'll be fine after that.
jonhohle: I’ve been seeing this in Gemini in the past few days. Often during a prompt with a reasonably large input set, I’ll get answers that appear to belong to someone else. It may be trigger hallucination, but it seems like it may be cache collisions or something else.
trq_: Hi, it's Thariq from the Claude Code Team here. Thanks for the detailed report. We’re confident this is a hallucination but of course take these reports seriously and the team is looking into it. We’ll report back if anything turns up.
178 points by Cider9986 · 77 comments
IE=edge" http-equiv="X-UA-Compatible"> Google Books (or similar) all book scans — $200,000 bounty (#234) · Issues · AnnaArchivist / annas-archive · GitLab <link as="font" crossorigin="" href="/ass...
What HN said:
ahmedfromtunis: I live in a country where the selection of available books, especially in English, is very limited. Buying online from foreign markets comes with a long list of administrative hurdles and limits.
stephenlf: The only legal hurdle keeping Anna’s Archive away from its noble goal (piracy laws) has been shown to mean zilch in the age of AI.
dr_dshiv: https://SourceLibrary.org has about 16,000 rare books translated — most for the first time. 50,000 books archived (will be translated when we have $$ for it). More tokens than English Wikipedia and about .75 petabytes.
trilogic: Who is behind Annas archive, there is a lot of english speakers involved in the team and forums! Anyway as long as buying isn´t owning no issues here.
163 points by jnord · 100 comments
Faced with observations of early black holes and galaxies that weren’t expected to exist, scientists have come up with a wealth of new theories to explain them. Now they just need to figure out which ones are true.
What HN said:
nzeid: > Recent ideas suggest that little red dots could be black holes cocooned in thick gas, possibly representing a completely new type of object called a black hole star, in which the tight shroud of gas emits light like a stellar atmosphere.
sulam: I thought I read somewhere that many of these little red dots are turning out to be nothing more than bog-standard brown dwarfs in our own galaxy that are confusing the signal.
hparadiz: Little red dots are my favorite new concept in astrophysics. This idea that there could be so much matter orbiting a black hole that the matter reaches sun levels of pressure which in turn starts steller fission without there being an star. Mind-blowing
hyperman1: There was a time where Hawking's A brief history of time gave a decent overview of the universe to beginners. Does anyone know how well it holds up today and if anything better exists?
142 points by andrenotgiant · 44 comments
ie=edge"/> var db_debug = function(){} if(document.cookie.
What HN said:
dsauerbrun: Maybe I'm too stupid to understand the article... How does this achieve performant querying for olap and oltp purposes? Based on my understanding, olap queries will go to the parquet files which are stored in a columnar fashion and oltp style queries will go to a caching layer th...
saisrirampur: But why? I’m skeptical of the idea of unifying storage just because it sounds “elegant” or “cool”. It’s not obvious to me how a single storage engine can compete with purpose-built OLTP and OLAP systems like Postgres and ClickHouse, without significant tradeoffs.
Avalaxy: Super cool stuff. Being able to combine your analytical platform and transactional database into one storage layer without having to set up ETL pipelines in between is really a game changer. Especially since it's just postgres, instead of some proprietary database.
andrenotgiant: Here's what I don't understand: Part of the value of doing an ETL pipeline via streaming replication is you get the full history of data in a table. An SCD type 2 table where each row also has a valid_from and valid_to timestamp column.
AI/ML
Business/Tech
Other
Programming
Science
Security
Stories and comments sourced from Hacker News public API. Not affiliated with Y Combinator or Hacker News.