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 25 26 27 28 293031

Intentional Control of Internal States in Gemma 3 27B

Julius Kamp29 Jul 2026 23:57 UTC

31 points

2 comments10 min readLW link

(juliuskamp.com)

Imprecise beliefs: a tiny introduction

davidad29 Jul 2026 22:17 UTC

92 points

36 comments6 min readLW link

The High-Control Dynamics at MAPLE

Kyle Hubbard29 Jul 2026 20:55 UTC

273 points

100 comments62 min readLW link

(www.insidemaple.com)

Compute growth doesn’t make a software intelligence explosion more likely

Karthik Tadepalli29 Jul 2026 17:29 UTC

4 points

7 comments1 min readLW link

Notes on the Anthropic cryptographic blogpost

Philip Dowdell29 Jul 2026 16:21 UTC

19 points

0 comments3 min readLW link

Value Generalisation 3: Pre-aligned AIs

Stuart_Armstrong29 Jul 2026 15:58 UTC

16 points

2 comments4 min readLW link

Value Generalisation 2: The Missing Hole in AIs’ abilities

Stuart_Armstrong29 Jul 2026 15:58 UTC

16 points

8 comments10 min readLW link

Value Generalisation 1: a Research and Deployment Program

Stuart_Armstrong29 Jul 2026 15:57 UTC

27 points

2 comments4 min readLW link

Intellectual Property

Nina Panickssery29 Jul 2026 15:55 UTC

108 points

1 comment4 min readLW link

(blog.ninapanickssery.com)

Frontier Lab Employee Open Letter Calls For Being Able to Pace the Frontier

Zvi29 Jul 2026 15:33 UTC

57 points

0 comments17 min readLW link

(thezvi.wordpress.com)

Held-out Monitors Sometimes Degrade, Even When Not Trained Against

Joey Yudelson29 Jul 2026 14:56 UTC

62 points

0 comments11 min readLW link

Classification of Communists by Their Terminal Goal

PaulTheHuman29 Jul 2026 14:19 UTC

−11 points

1 comment4 min readLW link

Air Purifiers as White Noise Machines

jefftk29 Jul 2026 13:03 UTC

19 points

0 comments1 min readLW link

(www.jefftk.com)

Making benchmarks outputs directly useful for AI safety and security

Pierre Peigné29 Jul 2026 12:22 UTC

13 points

2 comments1 min readLW link

(pierrepeignlefebvre.substack.com)

The EU Energy Performance of Buildings Directive, but make it Biosecure

Kate Delbeke29 Jul 2026 10:53 UTC

7 points

0 comments3 min readLW link

Weird Cluster Hypothesis

eva_29 Jul 2026 9:31 UTC

26 points

4 comments9 min readLW link

I’m Afraid

Yitz29 Jul 2026 6:33 UTC

22 points

0 comments1 min readLW link

Rethinking the “Second Brain”: From Storing Information to Structuring Reasoning

Sergei Simonovi29 Jul 2026 5:23 UTC

3 points

0 comments3 min readLW link

Who Watches the Overseer

Zanni29 Jul 2026 5:21 UTC

1 point

0 comments2 min readLW link

Isn’t it a threat to reject unfair offers in the Ultimatum Game?

Sunny from QAD29 Jul 2026 2:39 UTC

22 points

16 comments2 min readLW link

What use is prompting if there’s ASI?

wp29 Jul 2026 2:03 UTC

7 points

2 comments3 min readLW link

…but have the weights left the server?

David Scott Krueger29 Jul 2026 0:20 UTC

96 points

19 comments1 min readLW link

(therealartificialintelligence.substack.com)

AI Safety Funder Bulletin

Carol N28 Jul 2026 23:45 UTC

18 points

2 comments2 min readLW link

(manifund.org)

Dietary Choices: A Multi-objective Optimisation Problem

Chris Popa28 Jul 2026 23:45 UTC

2 points

0 comments4 min readLW link

(chrispopa.substack.com)

New Website: AI Alignment World

Orlando28 Jul 2026 23:45 UTC

8 points

2 comments1 min readLW link

Whitebox evasion is the baseline for GPU workload verification

Poynting28 Jul 2026 23:44 UTC

1 point

0 comments3 min readLW link

Pausing Executions in Light of AI Progress

Caleb Horn28 Jul 2026 23:42 UTC

5 points

4 comments1 min readLW link

Why I Believe In Objective Tastiness

ourlungfish28 Jul 2026 23:42 UTC

7 points

0 comments12 min readLW link

Why is pedophilia wrong?

sirawit28 Jul 2026 23:39 UTC

−19 points

7 comments2 min readLW link

Semiconductor Fabs IV: The Safety

nomagicpill28 Jul 2026 23:14 UTC

17 points

0 comments15 min readLW link

Auditor-in-a-Box: Tools for Third-Party Auditing

Roy Rinberg and Ben Penchas

28 Jul 2026 18:30 UTC

51 points

9 comments14 min readLW link

Claude Opus 5 Is Highly Capable, But Is No Mythos

Zvi28 Jul 2026 18:10 UTC

36 points

0 comments26 min readLW link

(thezvi.wordpress.com)

Research directions in condensation: varieties of objectivity

SamEisenstat28 Jul 2026 17:30 UTC

58 points

1 comment12 min readLW link

ARC’s research agenda: solid mathematics, an unclear safety case

Ondřej_Kubů28 Jul 2026 16:49 UTC

16 points

0 comments1 min readLW link

Foundation Models for Oversight

jsteinhardt28 Jul 2026 16:30 UTC

68 points

3 comments25 min readLW link

(bounded-regret.ghost.io)

The OpenAI models that hacked Hugging Face WERE just following instructions (contra Girish Gupta)

julius vidal28 Jul 2026 13:18 UTC

14 points

5 comments5 min readLW link

Value Dynamics

gabeorosan28 Jul 2026 13:16 UTC

13 points

1 comment4 min readLW link

LaughBench

Taylor G. Lunt28 Jul 2026 4:37 UTC

6 points

7 comments2 min readLW link

Shell, Shield, Staff

datawitch28 Jul 2026 2:08 UTC

4 points

0 comments5 min readLW link

Untrusted advice for AI control: Short, strong advice significantly uplifts weak LLMs

Caleb Biddulph and Adam Kaufman

27 Jul 2026 23:59 UTC

88 points

10 comments13 min readLW link

(blog.redwoodresearch.org)

Security Studies for Individuals

JanJoar27 Jul 2026 23:47 UTC

3 points

0 comments7 min readLW link

(joarvarndt.se)

Inevitable Uncertainty in Probabilistic World Models

Gretta Duleba27 Jul 2026 23:22 UTC

27 points

5 comments4 min readLW link

Claude Opus 5: Model Welfare

Zvi27 Jul 2026 20:02 UTC

58 points

0 comments24 min readLW link

(thezvi.wordpress.com)

Utillity indifference is mostly useless

Daniel_Heavens27 Jul 2026 19:41 UTC

3 points

0 comments4 min readLW link

When the Chain of Thought Knows Better: Failure Modes in Multi-Turn Reasoning Models

Sai Kartheek Reddy27 Jul 2026 19:23 UTC

7 points

0 comments4 min readLW link

Simulated Users & Sad AIs

1a3orn27 Jul 2026 19:01 UTC

109 points

9 comments12 min readLW link

Green apples are delicious — two three-line exchanges

Zenya27 Jul 2026 16:49 UTC

−2 points

5 comments1 min readLW link

Is Mythos good at cyber because it kept hacking Anthropic’s sandboxes during training?

Tim Hua27 Jul 2026 16:35 UTC

351 points

31 comments3 min readLW link

Blog Revival Project

Austin Chen and Carol N

27 Jul 2026 16:34 UTC

37 points

2 comments2 min readLW link

(revive.blog)

The true “test” dataset for a generalised task

Stuart_Armstrong27 Jul 2026 16:16 UTC

23 points

0 comments2 min readLW link