All 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 20252026

All Jan Feb Mar Apr May JunJulAug Sep

All 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 252627 28 29 30 31

The OpenAI models that hacked Hugging Face weren’t just following instructions

Girish Gupta25 Jul 2026 22:26 UTC

59 points

3 comments5 min readLW link

Your software should build itself

Max von Hippel25 Jul 2026 19:46 UTC

9 points

2 comments7 min readLW link

The Human Soul is LLM-like

Julian Bradshaw25 Jul 2026 19:45 UTC

15 points

3 comments2 min readLW link

The one name LLMs may fear

Steff25 Jul 2026 18:05 UTC

7 points

4 comments7 min readLW link

Introducing PIRAMID: Physics-Informed Research for Ambitious Mechanistic Interpretability

Lauren Greenspan, Ari Brill, Andrew Mack, Nischal Mainali, Jennifer Lin, Lucas Teixeira and Dmitry Vaintrob

25 Jul 2026 15:54 UTC

92 points

1 comment6 min readLW link

Claude Opus 5: The System Card

Zvi25 Jul 2026 13:42 UTC

41 points

0 comments10 min readLW link

(thezvi.wordpress.com)

The Viable System Model & Multi-Scale Agency

Jonas Hallgren25 Jul 2026 8:11 UTC

16 points

0 comments14 min readLW link

(equilibria1.substack.com)

SONI: Selective Orthogonalisation via Noise Injection

Jasper Chong25 Jul 2026 0:22 UTC

29 points

0 comments11 min readLW link

Linear probes tell you where quantization will hurt

Aniket Ghosh25 Jul 2026 0:22 UTC

29 points

0 comments5 min readLW link

Can Recursive Self-Report Probing Detect Emergent Misalignment?

kavitak12825 Jul 2026 0:21 UTC

8 points

0 comments8 min readLW link

Orbit: A framework for multi-agent security evaluations

wlanderson and bhagag

25 Jul 2026 0:21 UTC

7 points

0 comments1 min readLW link

The Long (Self-)Correction

Wei Dai24 Jul 2026 21:01 UTC

267 points

62 comments2 min readLW link

Why PauseAI UK accepts anonymous donations

Joseph Miller and PauseAI UK

24 Jul 2026 20:37 UTC

13 points

0 comments5 min readLW link

Seeking Mentees for the Sentient Futures Project Incubator

BrodyM24 Jul 2026 19:52 UTC

7 points

0 comments1 min readLW link

Intent Is All You Need.

Not Sure24 Jul 2026 19:52 UTC

−7 points

2 comments2 min readLW link

Congress Moves at Tech Pace: The FRONTIER Act

dan.parshall24 Jul 2026 19:34 UTC

26 points

0 comments8 min readLW link

Stable Systems Have Stable Outputs

Deixis24 Jul 2026 19:21 UTC

5 points

0 comments4 min readLW link

The AI Industrial Explosion — Part 5: Given AGI, automating physical production is probably not that hard

djbinder24 Jul 2026 19:15 UTC

29 points

2 comments25 min readLW link

(defensesindepth.bio)

Where does hint-following and concealment arise? A case study on OLMo-3 checkpoints

arav-dhoot and yix

24 Jul 2026 19:14 UTC

24 points

0 comments4 min readLW link

Introducing Lightcone Commons

Zvi24 Jul 2026 17:21 UTC

35 points

0 comments6 min readLW link

(thezvi.wordpress.com)

Should we be worried about how good AI is getting at coding autonomous drones?

Lukas Petersson24 Jul 2026 16:44 UTC

6 points

9 comments1 min readLW link

Coefficient Giving just gave GiveWell $1 billion. Where should other donors give now?

jackultraphil24 Jul 2026 15:40 UTC

4 points

0 comments2 min readLW link

(fundinganthropalypse.com)

LLMs are (still) mostly powered by imitative learning, not RL

Steven Byrnes24 Jul 2026 14:26 UTC

238 points

39 comments9 min readLW link

Democracy isn’t ready for the AI revolution

Sophia Gore24 Jul 2026 14:17 UTC

25 points

3 comments5 min readLW link

Does distilling Claude carry the persona with it?

Benji Berczi and Kyuhee Kim

24 Jul 2026 12:31 UTC

43 points

4 comments10 min readLW link

Georgia Tech AI Safety Initiative Retrospective 2025-2026

Ishan Khire, yix, Andersehen, Alec Harris, afterless, Eyas Ayesh, RocioPV and hersheys

24 Jul 2026 11:55 UTC

61 points

3 comments7 min readLW link

[Linkpost] Thoughts on the Recent OpenAI Hack

Linch24 Jul 2026 1:51 UTC

19 points

2 comments4 min readLW link

Should OpenAI’s rogue agent be punished?

groblegark24 Jul 2026 1:20 UTC

1 point

3 comments1 min readLW link

Evaluating Red Team and Blue Team Capability for AI Control Research

Ram Potham24 Jul 2026 1:11 UTC

11 points

0 comments8 min readLW link

(dearfutureais.substack.com)

Fixing rewards for NLA to reduce confabulation

SEONG PYO HONG24 Jul 2026 0:55 UTC

8 points

0 comments4 min readLW link

Anthropic’s J-Lens: A Research Engineer’s Analysis

willkn24 Jul 2026 0:54 UTC

8 points

0 comments11 min readLW link

Pulling the Fire Alarm

nem23 Jul 2026 23:14 UTC

86 points

5 comments1 min readLW link

Contra George Hotz on “AI 2040 and the Cult of Intelligence”

Matthew Tromp23 Jul 2026 22:43 UTC

8 points

0 comments8 min readLW link

(substack.com)

The Model Organism Lottery: Model Organism Interpretability Strongly Depends on Training Methodology

GabrielKS, therootof3, Raffaello Fornasiere, Nikita Menon and StefanHex

23 Jul 2026 22:37 UTC

45 points

0 comments6 min readLW link

(arxiv.org)

vibes-based thinking as a cultural response to unknowns

madelineberzak23 Jul 2026 22:35 UTC

7 points

0 comments8 min readLW link

Why an LLM cannot accumulate concepts

Zenya23 Jul 2026 21:40 UTC

−10 points

0 comments2 min readLW link

Red-teaming LLM unlearning: LUNAR’s “forgotten” knowledge is still recoverable

Diksha Gupta23 Jul 2026 21:39 UTC

10 points

0 comments5 min readLW link

Inception in DiffusionGemma—Jailbreaking a Diffusion Language Model by Pinning Tokens Anywhere on the Canvas

Theresa G., Simon S. and Siva Kumar Lakkoju

23 Jul 2026 21:39 UTC

19 points

0 comments11 min readLW link

ACX Atlanta August Meetup

Steve French23 Jul 2026 21:31 UTC

2 points

0 comments1 min readLW link

Not Pinning Your OpenRouter Provider Might Invalidate Your Research

Matthew Khoriaty23 Jul 2026 20:17 UTC

112 points

16 comments8 min readLW link

Pseudpocalypse

dynomight23 Jul 2026 20:09 UTC

60 points

18 comments23 min readLW link

Estimating LLM Training FLOPs on the Nvidia Jetson Orin Nano

William Fowler23 Jul 2026 19:52 UTC

24 points

0 comments10 min readLW link

Lightcone Commons

habryka23 Jul 2026 18:35 UTC

377 points

30 comments19 min readLW link

Challenge: Hand coding weights for efficient sequence memorisation

Linda Linsefors and Lucius Bushnaq

23 Jul 2026 18:05 UTC

56 points

9 comments18 min readLW link

AI Researchers Don’t Understand the State

Alex Amadori23 Jul 2026 17:58 UTC

3 points

0 comments1 min readLW link

(x.com)

The OpenAI/Huggingface incident | Redwood Research podcast episode 2

ryan_greenblatt and Buck

23 Jul 2026 17:56 UTC

85 points

0 comments2 min readLW link

V&V takes on OpenAI’s long-horizon incidents

Yoav Hollander23 Jul 2026 16:51 UTC

22 points

0 comments4 min readLW link

(blog.foretellix.com)

Duane Arnold

Tomás B.23 Jul 2026 16:17 UTC

138 points

4 comments19 min readLW link

Introducing Impact List: a ranking of people by the expected value of their donations

Elliot_Olds23 Jul 2026 15:27 UTC

12 points

0 comments5 min readLW link

Wanting Crooked Lines

Davey Morse23 Jul 2026 14:33 UTC

13 points

0 comments4 min readLW link