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 AprMayJun

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 272829 30 31

Constitutional AI Alignment

RogerDearnaley27 May 2026 22:29 UTC

27 points

9 comments47 min readLW link

LLMs Through the Eyes of Vinge

Gordon Seidoh Worley27 May 2026 20:20 UTC

52 points

2 comments4 min readLW link

(www.uncertainupdates.com)

Biologically Plausible SGD Is Hard

Elliot Callender27 May 2026 19:34 UTC

9 points

0 comments1 min readLW link

Eval Cooperativeness May Be a Scalable Mitigation for Eval Gaming

Jasmine Li and Alex Turner

27 May 2026 19:33 UTC

73 points

5 comments10 min readLW link

(turntrout.com)

no, Magnifica Humanitas is not AI-written

bhauth27 May 2026 19:26 UTC

−13 points

18 comments3 min readLW link

Albuquerque ACX Meetup

Mary27 May 2026 18:27 UTC

2 points

0 comments1 min readLW link

Full automation of AI R&D probably yields a large speed up even without a software-only singularity

ryan_greenblatt27 May 2026 18:16 UTC

67 points

17 comments3 min readLW link

Not Prosthetics

Elliot Callender27 May 2026 17:22 UTC

11 points

0 comments2 min readLW link

BCI Cognition Enhancement is Possible

Elliot Callender27 May 2026 17:19 UTC

17 points

0 comments1 min readLW link

The ballad of TIGIT

Abhishaike Mahajan27 May 2026 17:04 UTC

84 points

1 comment9 min readLW link

Leveraging Introspection for Alignment

Yotam27 May 2026 16:54 UTC

25 points

3 comments7 min readLW link

Announcing Geodesic Research

Puria, Cam, Alexandra Narin, Edward James Young and Kyle O’Brien

27 May 2026 16:40 UTC

74 points

1 comment5 min readLW link

AI as a Social Technology, by Henry Farell

TheManxLoiner27 May 2026 13:41 UTC

15 points

0 comments3 min readLW link

(lovkush.substack.com)

More capable AI, less money raised

Shoshannah Tekofsky27 May 2026 12:57 UTC

28 points

2 comments3 min readLW link

(theaidigest.org)

Quantitative AI risk assessment: a starting point

Henry Papadatos, jakub_krys, malcolmmurray and Renn Karageorgieva

27 May 2026 9:42 UTC

38 points

7 comments11 min readLW link

(www.safer-ai.org)

[paper] Training on Documents About Monitoring Leads to CoT Obfuscation

Reilly Haskins, bilalchughtai and Josh Engels

27 May 2026 9:39 UTC

31 points

1 comment4 min readLW link

(arxiv.org)

No frontier model has acceptable levels of compliance with the EU AI Act and privacy legislation.

Daan Henselmans, Arno Libert, Amber Koelfat and LennardZ

27 May 2026 7:35 UTC

29 points

0 comments9 min readLW link

Thinking outside the box? LLM analysis of simplified cooperative poker

Dentosal27 May 2026 7:28 UTC

15 points

0 comments4 min readLW link

Standard deviations from just two values

kqr27 May 2026 5:01 UTC

41 points

2 comments3 min readLW link

(entropicthoughts.com)

Contra Wentworth on Physical Attractiveness for Men

Gretta Duleba26 May 2026 23:20 UTC

123 points

25 comments8 min readLW link

Training Language Models for Controlled Stochasticity

Sruthi Kuriakose and Davide Baldelli

26 May 2026 22:17 UTC

18 points

0 comments5 min readLW link

Are Mythos’ Cyber Capabilities Overstated? - Yes and No

Muhan Luo26 May 2026 22:17 UTC

7 points

1 comment10 min readLW link

Should we train LLMs to be human?

Hubert Plisiecki26 May 2026 22:16 UTC

3 points

0 comments2 min readLW link

Steering Directions Are Explanations, Not Handles

JackYoung2726 May 2026 22:15 UTC

8 points

0 comments7 min readLW link

You Can’t Tell a Conscience From a Leash by Watching

GenericHousewife_B26 May 2026 22:14 UTC

6 points

2 comments3 min readLW link

Finding the Mole: Bayesianism is Hard

laniakea26 May 2026 21:55 UTC

35 points

0 comments5 min readLW link

Simplifying Alignment by Expanding Scope

Adam Chlipala26 May 2026 21:42 UTC

3 points

0 comments7 min readLW link

Practical Learnings from Synthetic Document Finetuning

Axel Højmark and Jérémy Scheurer

26 May 2026 19:22 UTC

80 points

6 comments8 min readLW link

Claude, Author of the Humanitas

Linch26 May 2026 16:05 UTC

118 points

42 comments16 min readLW link

When does debate help a weak judge? Evidence from code, logic and math

ethanelasky and frank_b_n

26 May 2026 14:36 UTC

16 points

4 comments5 min readLW link

ACX Atlanta June 2026 Meetup

Steve French26 May 2026 13:59 UTC

2 points

0 comments1 min readLW link

RTMH: Pope Leo’s Magnifica Humanitas on AI

Zvi26 May 2026 13:20 UTC

36 points

5 comments29 min readLW link

(thezvi.wordpress.com)

The Fatal AGI Hardware Gap

jrincayc26 May 2026 12:55 UTC

4 points

5 comments1 min readLW link

Many portions of Magnifica Humanitas appear to be AI-written

DanielFilan26 May 2026 7:40 UTC

78 points

51 comments6 min readLW link

(danielfilan.com)

Brain transfers might be the easiest path to life extension

Semi-Pseudonymous26 May 2026 6:23 UTC

11 points

15 comments4 min readLW link

Some Thoughts on Bengio’s Scientist AI

Matthew Khoriaty26 May 2026 3:05 UTC

21 points

4 comments2 min readLW link

Brackets Are a Bad Way to Regulate

Hide26 May 2026 3:01 UTC

75 points

15 comments5 min readLW link

(hidefromit.substack.com)

Donating 80% While It Still Counts

jefftk26 May 2026 1:30 UTC

123 points

8 comments6 min readLW link

(www.jefftk.com)

Notes on Fourier Analysis

Menotim26 May 2026 0:39 UTC

32 points

5 comments23 min readLW link

Improving Petri scheming audits with environment blueprints

Jannes Elstner26 May 2026 0:31 UTC

12 points

0 comments6 min readLW link

Pope Leo’s First AI Encyclical – Summary and Commentary

John-Clark Levin25 May 2026 23:48 UTC

26 points

8 comments39 min readLW link

Cognitive Security as an AI Safety Cause Area

jsteinhardt25 May 2026 18:30 UTC

156 points

18 comments2 min readLW link

Sentient Welfare Across Three Futures

MichaelDickens25 May 2026 16:22 UTC

13 points

2 comments2 min readLW link

Linkpost: New Vatican Encyclical on AI Governance

Jackson Wagner25 May 2026 15:40 UTC

58 points

7 comments1 min readLW link

How AI Will Save Prediction Markets

alexjaniak25 May 2026 14:24 UTC

11 points

18 comments6 min readLW link

(x.com)

There should be a discussion about LW’s policy to allow calls for violence

Mikhail Samin25 May 2026 13:51 UTC

−5 points

21 comments10 min readLW link

Character-trained models can struggle to generalise

Nathaniel Mitrani25 May 2026 12:58 UTC

22 points

4 comments4 min readLW link

Applications open for the Secure Program Synthesis Fellowship

eitan sprejer25 May 2026 10:04 UTC

8 points

0 comments1 min readLW link

Announcing the Frontier Biodefense Fellowship (deadline 7 June)

Tobias H25 May 2026 7:58 UTC

5 points

0 comments3 min readLW link

Taxing Small Cars To Improve MPG

jefftk24 May 2026 21:50 UTC

91 points

11 comments2 min readLW link

(www.jefftk.com)