frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Open in hackernews

When AI Benchmarks Plateau: A Systematic Study of Benchmark Saturation

https://arxiv.org/abs/2602.16763
18•doppp•55m ago

Comments

otterdude•17m ago
This seems to be the end of the road for LLM's. There's only so much accuracy on a highly non-linear space you can get from a regression.

If the Pareto rule is any indicator, 80% of results come from 20% of causes. It seems that we have alot more to learn about intelligence.

I am reminded of a statement on truth from an ancient philosopher, this sentiment seems to be exactly the opposite of the LLM training paradigm

“The seeker after truth is not one who studies the writings of the ancients and, following his natural disposition, puts his trust in them, but rather the one who suspects his faith in them and questions what he gathers from them, the one who submits to argument and demonstration and not the sayings of human beings whose nature is fraught with all kinds of imperfection and deficiency. Thus the duty of the man who investigates the writings of scientists, if learning the truth is his goal, is to make himself an enemy of all that he reads, and, applying his mind to the core and margins of of its content, attack it from every side. he should also suspect himself as he performs his critical examination of it, so that he may avoid falling into either prejudice or leniency.” - ibn al-Haytham

graemep•10m ago
I am wondering whether the reason he needed to say it was because he was arguing with those who did out their trust in the writings of the ancients.
scotty79•9m ago
Do you draw that conclusion from the fact that AI surprisingly quickly reaches the end of each ruler we try to measure it with?
freejazz•2m ago
Can't call it AI like that without discrediting yourself. You mean LLMs?
adrianN•4m ago
I really don’t think we know enough about what „intelligence“ is or how LLMs actually work to confidently say that this is the end of the road for LLM.
buckle8017•8m ago
Slop

> We find that nearly half of the our bench- marks exhibit saturation

joeyagreco•6m ago
This leads me to believe it's NOT slop lol
jdiff•6m ago
It's a grammatical error, sure, where is the indication of slop?

Show HN: I built a Buffer alternative that also generates the videos

https://videoai.me/login
1•Paul_Grsl•1m ago•0 comments

A free, cross-platform ePub 3 reader

https://thorium.edrlab.org/en/about/
1•cernocky•2m ago•0 comments

Alpamayo 2 Super, Nvidia frontier open reasoning model for autonomous vehicles

https://twitter.com/JensenHuang/status/2084656303046332747
1•nthypes•3m ago•0 comments

The Download: US robot restrictions, and ICE's DNA grab

https://www.technologyreview.com/2026/08/04/1141098/the-download-robot-restrictions-ice-dna/
2•joozio•4m ago•0 comments

ChainDrop NPM supply-chain attack infects packages

https://www.bleepingcomputer.com/news/security/massive-chaindrop-npm-supply-chain-attack-infects-...
1•iLemming•5m ago•0 comments

All of Winona Police Department's Flock cameras cut down and stolen

https://www.valleynewslive.com/2026/08/04/every-flock-camera-winona-minnesota-cut-down-stolen-coo...
3•NDlurker•5m ago•1 comments

RailWarden – Safely run parallel AI coding agents in one Git repository

https://github.com/advaith-1212/railwarden
1•Satya_1212•6m ago•0 comments

Ask HN: Should we direct AI on our personal projects during work hours?

1•rpnx•7m ago•0 comments

Bending Spoons to Buy Software Firm Airtable for $2.3B

https://www.bloomberg.com/news/articles/2026-08-04/bending-spoons-to-buy-software-firm-airtable-f...
1•devy•7m ago•0 comments

Average IQ by Country 2026

https://worldpopulationreview.com/country-rankings/average-iq-by-country
1•amelius•8m ago•0 comments

Show HN: Gantt resource planning vs. event scheduling in RevoGrid

https://benchmark.rv-grid.com/revogrid-vs-bryntum-scheduling/
1•kumakint•9m ago•0 comments

Former OpenAI exec Fidji Simo's company has analyzed 3,500 blood vials with AI

https://fortune.com/2026/08/03/fidji-simo-discusses-pots-diseases-chroniclebio-ai-startup-openai/
2•gwintrob•9m ago•0 comments

Silico, the platform for ambitious AI research, is publicly available today

https://twitter.com/GoodfireAI/status/2084671608028057737
1•glowingvoices•9m ago•0 comments

The Inference Engine Guide for K3 Deployment

https://twitter.com/vllm_project/status/2081788542079123939
1•gmays•10m ago•0 comments

Show HN: TsundereSQL – 208 SQL exercises graded locally with DuckDB-WASM

https://tsunderesql.itch.io/tsunderesql
1•elsaicequeen•11m ago•0 comments

Microsoft Tells Engineers 'Tokenmaxxing Is Not What We Are Optimizing For

https://www.404media.co/microsoft-tells-engineers-tokenmaxxing-is-not-what-we-are-optimizing-for/
3•cdrnsf•11m ago•0 comments

Show HN: Give your agents a personal internet based on what you've surfed

https://github.com/Fergana-Labs/stash/blob/main/docs/ai-native-bookmarks.md
1•samzliu•11m ago•0 comments

The Geoid

https://svs.gsfc.nasa.gov/5660/
2•jonbaer•11m ago•0 comments

Vampires (Programmers) versus Werewolves (Sysadmins)

https://blog.codinghorror.com/vampires-programmers-versus-werewolves-sysadmins/
1•speckx•14m ago•0 comments

Show HN: I got tired of sharing files between devices, so I built TapSend

https://tapsend.app
1•TapSend•15m ago•0 comments

OSS and Photography

https://www.jasonwu.ink/signals/2026-08-02-oss-and-photography
1•whiteblossom•16m ago•0 comments

What If Past Returns Were Indicative of Future Returns?

https://elmwealth.com/expected-returns/
1•wawayanda•16m ago•0 comments

The IPv6 Private Network Trap: Why Fd00:/8 Is Not 192.168.x.x

https://mabels.github.io/mseg-tester/blog/fd00-is-not-your-192-168-short.html
1•fastandfearless•16m ago•0 comments

Show HN: OldHand A Claude/Codex plugin to verify the development flow end-to-end

https://github.com/berwinsingh/oldhand
1•berwinsingh•17m ago•0 comments

Hyperion Entertainment BV and Amiga Corporation Reach Settlement

https://www.hyperion-entertainment.com/index.php/news/38-corporate
1•doener•19m ago•2 comments

The Indo-European Family Tree

https://djbinder.com/language-tree/
2•benbreen•19m ago•0 comments

AI detectors, the new clergy of authenticity – The Great Work Heist

https://geraldoaugustoseabra.substack.com/p/ai-detectors-the-new-clergy-of-authenticity
1•newsgames•20m ago•0 comments

Please add min-release-age to your .npmrc

https://www.endorlabs.com/learn/npm-malware-compromises-keyv-and-cacheable-with-500m-weekly-downl...
1•scj13•20m ago•1 comments

The Scary, Scary Singularity

https://markmcdonaldthoughts.substack.com/p/the-scary-scary-singularity
1•paulpauper•21m ago•0 comments

Show HN: SIEMatic, a fair-sourced observability and security platform

https://github.com/mcindi/siematic
2•ilovetux•21m ago•0 comments