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

12 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

96 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

Agenda: Infrastructure for Trading with Partially Misaligned AIs

VojtaKovarik7 Aug 2026 14:13 UTC

18 points

2 comments9 min readLW link

(limits-of-evaluation.org)

Upcoming Dovetail fellow talks & discussion

Alex_Altair7 Aug 2026 13:26 UTC

17 points

1 comment2 min readLW link

models may behave differently in graded episodes (a tirade)

nostalgebraist7 Aug 2026 11:49 UTC

278 points

31 comments61 min readLW link

Hypotheticals

Ramseyian7 Aug 2026 7:35 UTC

1 point

0 comments6 min readLW link

Struggles With Complexity

warner7 Aug 2026 7:25 UTC

4 points

0 comments2 min readLW link

How to Think like a Human 101

Morpheus7 Aug 2026 6:09 UTC

−17 points

3 comments1 min readLW link

Some SHA-256 hashes

Radford Neal6 Aug 2026 23:21 UTC

9 points

4 comments1 min readLW link

Have models report provable security bugs in their environment

anithite6 Aug 2026 22:52 UTC

24 points

2 comments4 min readLW link

Why do models task game?

aditya singh, Neel Nanda and Senthooran Rajamanoharan

6 Aug 2026 22:16 UTC

94 points

1 comment24 min readLW link

Contra Oster on Alcohol in Pregnancy. Part 1. The pharmacokinetics of alcohol metabolism

Mvolz6 Aug 2026 21:34 UTC

99 points

2 comments7 min readLW link

Side-Effects of Length Penalty in RL

Brian Davies and Luc Feron

6 Aug 2026 21:23 UTC

13 points

3 comments18 min readLW link

User awareness in frontier models

Ziqian Zhong and jsteinhardt

6 Aug 2026 20:43 UTC

59 points

1 comment12 min readLW link

(transluce.org)

The FRONTIER Act barely creates its implementing office

Philip Dowdell6 Aug 2026 19:58 UTC

11 points

0 comments3 min readLW link

My Private Personal Agent

danielms6 Aug 2026 19:53 UTC

13 points

2 comments5 min readLW link

Traffic Shaping for Workload Classification

Andrew Dickson6 Aug 2026 19:49 UTC

10 points

0 comments8 min readLW link

(lucidcomputing.substack.com)

You need to stop X companies to get a Y-month pause

Expertium6 Aug 2026 19:30 UTC

29 points

4 comments2 min readLW link

Three years of progress in 500 lines of code

Gerard Boxo6 Aug 2026 18:27 UTC

65 points

4 comments4 min readLW link