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 10 11 121314 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

Free will is like temperature

Optimization Process12 Aug 2026 23:32 UTC

47 points

10 comments1 min readLW link

Impact markets made concrete

Carol N12 Aug 2026 22:45 UTC

6 points

0 comments6 min readLW link

(manifund.substack.com)

Unblocking AI’s Continual Learning: Hints From How Humans Learn

nimakeivan12 Aug 2026 18:28 UTC

3 points

23 comments10 min readLW link

Monthly Roundup #45: August 2026

Zvi12 Aug 2026 17:50 UTC

24 points

1 comment20 min readLW link

(thezvi.wordpress.com)

Finding the Seams of Perception

jimmy12 Aug 2026 17:33 UTC

30 points

0 comments2 min readLW link

(beneathpsychology.com)

Introducing the Conceptual Reasoning Index

Chi Nguyen, Emery Cooper, Caspar Oesterheld, Alex Kastner and Joe Benton

12 Aug 2026 17:08 UTC

83 points

16 comments5 min readLW link

(alignment.anthropic.com)

One attention head carries knight forks in a chess transformer, and here’s a new toolkit that found it.

David Litman12 Aug 2026 16:48 UTC

8 points

0 comments1 min readLW link

(github.com)

The Psychology of Cope: Rationality Is Not Reversed Irrationality

Chris_Leong12 Aug 2026 12:20 UTC

27 points

12 comments3 min readLW link

Demon Safety

Ben Pace12 Aug 2026 11:55 UTC

53 points

5 comments1 min readLW link

The Closure of the Internet (Research Linkpost)

Dean Valentine12 Aug 2026 8:22 UTC

34 points

4 comments1 min readLW link

(arctotherium.substack.com)

AI swarms are starting to pose indirect takeover risk

oakhu and Alex Mallen

12 Aug 2026 5:05 UTC

138 points

6 comments10 min readLW link

Did the alignment community underestimate its power?

StanislavKrym12 Aug 2026 2:56 UTC

23 points

0 comments10 min readLW link

The Age of Pluribus: One Consultant for Everyone

Dorothy Gale12 Aug 2026 1:39 UTC

9 points

0 comments4 min readLW link

We should consider how long monitoring is reliable for during RL

lachlan on a boat12 Aug 2026 1:22 UTC

20 points

1 comment4 min readLW link

When (and when not) LLMs can verbalize awareness of J-Space concept injections—Initial results

Ethan Garcia12 Aug 2026 1:21 UTC

8 points

0 comments15 min readLW link

(e-m-garcia.github.io)

Arguments for and against (me) dropping out

hersheys12 Aug 2026 1:12 UTC

30 points

3 comments8 min readLW link

Patient Zero

LoopGameScrollMonkey12 Aug 2026 1:10 UTC

17 points

0 comments12 min readLW link

An anytime algorithm for mixing the computable measures

Cole Wyeth12 Aug 2026 0:57 UTC

26 points

0 comments4 min readLW link

Various Reflections About What Happened With OpenAI’s Internal Models

Zvi11 Aug 2026 22:20 UTC

63 points

1 comment30 min readLW link

(thezvi.wordpress.com)

Claude Opus 5 Just Beat My Text-Based Adventure Game Benchmark

derelict543211 Aug 2026 20:19 UTC

22 points

4 comments5 min readLW link

[Webinar] Why AI Safety is a Capital Allocation Problem

Schizoid Rentoid11 Aug 2026 19:27 UTC

2 points

0 comments1 min readLW link

Misaligned AIs could use killer robots to take over

Omar Khursheed and TurnTrout

11 Aug 2026 19:03 UTC

134 points

10 comments6 min readLW link

(turntrout.com)

Measuring Spurious Correlations with Feature Strength

egan11 Aug 2026 18:05 UTC

36 points

1 comment10 min readLW link

AI governance work needs much better monitoring

jackultraphil11 Aug 2026 17:42 UTC

10 points

0 comments8 min readLW link

(fundinganthropalypse.com)

LLMs Are Starting To Noticeably Accelerate Our Work

johnswentworth11 Aug 2026 17:06 UTC

249 points

19 comments2 min readLW link

Software Is Hard

cylonator11 Aug 2026 16:26 UTC

5 points

0 comments1 min readLW link

How risky would it be to make powerful AI obey one or a few people?

cousin_it and Seth Herd

11 Aug 2026 16:15 UTC

93 points

24 comments8 min readLW link

Extreme concentration of power over ASI has non-obvious advantages

Seth Herd11 Aug 2026 16:13 UTC

43 points

4 comments19 min readLW link

Those Who Make History

Raelifin11 Aug 2026 13:59 UTC

83 points

1 comment8 min readLW link

(open.substack.com)

Seeing things through in the age of AI

alkjash11 Aug 2026 13:14 UTC

30 points

3 comments1 min readLW link

Productive Signaling: Competitive Software Development, Not Competitive Programming

Adam Chlipala11 Aug 2026 12:16 UTC

9 points

0 comments9 min readLW link

The Next Ecology

Eigenbraid11 Aug 2026 8:29 UTC

10 points

4 comments5 min readLW link

Redux: (∃ Stochastic Natural Latent) Implies (∃ Deterministic Natural Latent)

David Lorell11 Aug 2026 5:52 UTC

103 points

33 comments2 min readLW link

On using crises to shift political will for AI

clickyquack11 Aug 2026 4:54 UTC

17 points

0 comments4 min readLW link

Revived Lightweight Transit Predictions Page

jefftk11 Aug 2026 2:31 UTC

11 points

0 comments1 min readLW link

(www.jefftk.com)

Probing Knowledge Recovery in Unlearned Models

mehnoor11 Aug 2026 2:21 UTC

7 points

0 comments7 min readLW link

Before We Defer Research to AI: Measuring Apparent-Success-Seeking

Keira Leal11 Aug 2026 2:21 UTC

17 points

7 comments5 min readLW link

Canadian Nuclear Emancipation: Canada’s role amidst global dreams of energy security and nuclear development

tandemocracy11 Aug 2026 2:20 UTC

8 points

0 comments3 min readLW link

(worldwithoutwater.substack.com)

A Topic Detector, Not a Lie Detector: what J-space monitoring actually tracks

Melchior de Polignac11 Aug 2026 2:18 UTC

10 points

0 comments4 min readLW link

Models inherit the writer, not who the writer was imitating

0Chris5R11 Aug 2026 2:16 UTC

7 points

0 comments10 min readLW link

What Claude Saw Below

Luke Nicholls11 Aug 2026 2:14 UTC

60 points

1 comment27 min readLW link

Creative math research by AI as the latest sign of the end

Mitchell_Porter11 Aug 2026 1:37 UTC

57 points

10 comments2 min readLW link

A study on instability of LLM responses as a behavioral signature of self-Referential reports.

PARAS BALANI11 Aug 2026 1:01 UTC

7 points

0 comments1 min readLW link

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