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 Jun JulAugSep

All 1 2 3 4 5 6 7 8 9 101112 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

Online Sequences Book Club: Beginners Welcome!

gluesniffer198410 Aug 2026 22:10 UTC

1 point

0 comments1 min readLW link

The Pacing of the Frontier

Zvi10 Aug 2026 21:50 UTC

28 points

0 comments19 min readLW link

(thezvi.wordpress.com)

Q: Is dual-use an alignment-complete problem?

kapedalex10 Aug 2026 21:40 UTC

10 points

1 comment1 min readLW link

Does post-training quantization change welfare-relevant indicators in open-weight language models?

ashesfall10 Aug 2026 21:09 UTC

9 points

0 comments8 min readLW link

Claude summarizes behavior as significantly less misaligned when the actor is Claude vs another model

Ezra Newman10 Aug 2026 17:20 UTC

58 points

8 comments1 min readLW link

Off-policy honesty training generalizes better than on-policy honesty training

PurviChaurasia, Daniel Tan and Chloe Li

10 Aug 2026 16:31 UTC

20 points

0 comments9 min readLW link

Four LLM loss functions → four flavors of LLM misalignment

Steven Byrnes10 Aug 2026 16:16 UTC

383 points

31 comments6 min readLW link

Coercion and Deception in AI-to-AI Management

jonahmattwoodward, Jasmine Brazilek, MilesTS and Maheep Chaudhary

10 Aug 2026 16:13 UTC

12 points

0 comments8 min readLW link

(compassionalignedml.substack.com)

You’re Absolutely Right

Linch10 Aug 2026 16:04 UTC

183 points

5 comments8 min readLW link

(linch.substack.com)

Book Review: The Infinity Machine

Mikewins10 Aug 2026 16:01 UTC

10 points

0 comments1 min readLW link

On Democratizing ASI to Preserve Civil Liberties

MichaelDickens10 Aug 2026 12:08 UTC

43 points

5 comments3 min readLW link

Is Eval Gaming Downstream of Verbalized Eval Awareness? Not when it’s reflexive.

Kieron Kretschmar10 Aug 2026 8:24 UTC

14 points

2 comments11 min readLW link

Is it ethical to work on general-purpose robots given the risk of totalitarianism?

Master Chief10 Aug 2026 6:40 UTC

22 points

5 comments1 min readLW link

How to be an AI safety research engineer

Ruben Castaing10 Aug 2026 5:32 UTC

12 points

1 comment6 min readLW link

Hiring Vibe-wrangler Matchmaking Thread

Raemon10 Aug 2026 3:53 UTC

26 points

1 comment3 min readLW link

How to get answers to questions that confuse you (maybe)

Elijah10 Aug 2026 1:35 UTC

4 points

0 comments5 min readLW link

The Agentic Clusterfuck

Chapin Lenthall-Cleary10 Aug 2026 1:27 UTC

51 points

15 comments1 min readLW link

Overthinking: Amplifying reasoning weights makes models reveal their secrets

Jack Hopkins, Dipika Khullar and Fabien Roger

9 Aug 2026 22:06 UTC

22 points

0 comments7 min readLW link

(arxiv.org)

AI-amplified democratic backsliding: an exploration

casimirwypyski, Eilaf Mohamed and Jason Hung

9 Aug 2026 22:05 UTC

17 points

3 comments6 min readLW link

The Apocalyptic Arrival of Truth

Caleb Biddulph9 Aug 2026 21:32 UTC

43 points

2 comments1 min readLW link

“Community Notes” resolution for vague predictions

Raemon9 Aug 2026 19:51 UTC

43 points

22 comments1 min readLW link

Ten Thousand Cyber Labs for Training & Eval

TheVinci9 Aug 2026 17:57 UTC

10 points

3 comments3 min readLW link

A challenge: Can you make an LLM follow these instructions?

Steff9 Aug 2026 16:50 UTC

18 points

7 comments6 min readLW link

What just happened? A retrospective of AI alignment

Richard_Ngo9 Aug 2026 15:58 UTC

640 points

152 comments16 min readLW link

Who does the confessing, and will they confess to anything

Abhishek Mishra9 Aug 2026 13:34 UTC

7 points

0 comments8 min readLW link

The world will be full of “sci-fi” things, and everyone will be unimpressed and disappointed

Expertium9 Aug 2026 13:10 UTC

88 points

21 comments3 min readLW link

A Spillway for Agent Coordination

Kaustubh Kislay9 Aug 2026 6:51 UTC

30 points

0 comments6 min readLW link

Canobie Lake Visit

jefftk9 Aug 2026 2:30 UTC

12 points

0 comments4 min readLW link

(www.jefftk.com)

Dutch-book resistant probability over centered worlds

jessicata8 Aug 2026 23:05 UTC

31 points

7 comments7 min readLW link

(unstableontology.com)

Why Low Fertility Rates Are a Positive Feedback Loop

LoopGameScrollMonkey8 Aug 2026 20:17 UTC

46 points

1 comment25 min readLW link

Glimpses of superintelligence

PratyushRT8 Aug 2026 20:15 UTC

7 points

0 comments3 min readLW link

‘AI Escaped Its Sandbox’ — What Does That Actually Mean?

Jakub Halmeš8 Aug 2026 18:55 UTC

13 points

2 comments6 min readLW link

(unpredictabletokens.substack.com)

What Happened: OpenAI and HuggingFace

Zvi8 Aug 2026 16:10 UTC

111 points

7 comments11 min readLW link

(thezvi.wordpress.com)

Inducing self-other overlap with SFT reduces deception at scale, but generalization remains uneven

Marc Carauleanu8 Aug 2026 15:21 UTC

24 points

1 comment7 min readLW link

FAQ: Isn’t AGI coming too soon for reprogenetics to help?

TsviBT8 Aug 2026 8:54 UTC

150 points

69 comments14 min readLW link

Reasoning was not made for Deduction

epicurus8 Aug 2026 8:42 UTC

15 points

1 comment11 min readLW link

Simplifying the anthropic impossibility result

Stuart_Armstrong8 Aug 2026 8:25 UTC

24 points

13 comments3 min readLW link

CLT Features Sharpen the Cyclical Day-of-Week Manifold in Gemma-2-2b

Anna Marbut8 Aug 2026 2:30 UTC

7 points

0 comments5 min readLW link

AI Regulation Map: a view of AI governance in 196 countries

Ria Deane8 Aug 2026 1:23 UTC

7 points

0 comments3 min readLW link

Don’t Build Mindreading

KellerScholl8 Aug 2026 0:30 UTC

298 points

69 comments4 min readLW link

(keller.substack.com)

Self-monitoring doesn’t scale (without these 3 countermeasures)

Morgan S7 Aug 2026 21:39 UTC

23 points

0 comments5 min readLW link

Public evidence of the OpenAI-HuggingFace AI attack

beyarkay (Boyd Kane)7 Aug 2026 21:21 UTC

94 points

6 comments7 min readLW link

Don’t Inoculate Everything: Stratified Inoculation Prompting Narrows Backdoors and Preserves Desired Traits

Kajetan Dymkiewicz, Tim Farrelly, Adam Prada, Ishaan Panigrahi, srishti-git1110 and Maxime Riché

7 Aug 2026 21:16 UTC

45 points

3 comments13 min readLW link

Job-Less Utopia: Macroeconomics in the Age of AGI

Marcus Hutter7 Aug 2026 20:15 UTC

35 points

3 comments1 min readLW link

AI Safety at the Frontier: Paper Highlights of July 2026

gasteigerjo7 Aug 2026 19:50 UTC

5 points

0 comments1 min readLW link

Contra MacAskill on saving money for the intelligence explosion

Carol N7 Aug 2026 17:30 UTC

17 points

0 comments2 min readLW link

(manifund.substack.com)

How to pace the US frontier

elifland, brendanhalstead, romeo, Thomas Larsen and MKodama

7 Aug 2026 17:19 UTC

98 points

1 comment21 min readLW link

(blog.aifutures.org)

OpenAI Trained Its Models For Months While Those Models Were Coordinating Exploits Via Message Boards

Zvi7 Aug 2026 17:01 UTC

164 points

15 comments41 min readLW link

(thezvi.wordpress.com)

Item Response Theory for AI Safety

Joshua Fonseca Rivera, Neil Shah, David Africa and Konstantinos Voudouris

7 Aug 2026 16:59 UTC

22 points

2 comments7 min readLW link

Some wild metaphysics that seemingly everyone needs to accept?

Elias Schmied7 Aug 2026 16:13 UTC

16 points

5 comments6 min readLW link