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 Jul AugSep

All 1 2 3 4 5 6 7 8 91011 12 13 14 15 16 17 18 19

I want to bitch about the increased salience of AI risk

Nathan Young9 Sep 2026 23:14 UTC

17 points

10 comments1 min readLW link

People are more worried about AI x-risk than they let on

KatjaGrace9 Sep 2026 22:24 UTC

24 points

1 comment1 min readLW link

(worldspiritsockpuppet.substack.com)

Can a superintelligence do THAT?

Eliezer Yudkowsky9 Sep 2026 21:59 UTC

135 points

28 comments13 min readLW link

Final research agenda #3: towards automated Husserlian CEV

Mitchell_Porter9 Sep 2026 21:52 UTC

13 points

0 comments1 min readLW link

Have deranged conspiracy theories actually exploded? The short answer: probably not.

KatSpartz9 Sep 2026 21:35 UTC

13 points

8 comments4 min readLW link

Constitutional AI Widens Narrow Secret Loyalty of LLMs

navraj9 Sep 2026 21:05 UTC

7 points

0 comments7 min readLW link

Recurrent KV-cache sharing may undermine the bounded-depth argument for CoT monitorability

MBaert9 Sep 2026 21:02 UTC

30 points

2 comments3 min readLW link

The Risks of Anti-AI Voters Turning Against AI Alignment

voikante9 Sep 2026 20:58 UTC

10 points

0 comments48 min readLW link

(voikante.substack.com)

Fighting scope and frame control from the frontier-labs

Vincent_Bagayoko9 Sep 2026 20:53 UTC

10 points

0 comments9 min readLW link

An Ode to LessWrong

Tamara Sofía Falcone9 Sep 2026 20:47 UTC

15 points

1 comment1 min readLW link

show LW: programmatic internet research, scry.io

Xyra Sinclair9 Sep 2026 20:44 UTC

24 points

0 comments1 min readLW link

Automation and Political Power

MarkelKori9 Sep 2026 19:55 UTC

10 points

2 comments2 min readLW link

Pretraining an LLM without mentions of consciousness

Eye You and frmsaul

9 Sep 2026 19:38 UTC

30 points

13 comments7 min readLW link

Bottlenecks in AI Safety—Opportunities for Impact

Brian Davies9 Sep 2026 17:48 UTC

26 points

1 comment16 min readLW link

Recommendations for People Getting into Technical AI Governance Research

Aaron_Scher, yams, peterbarnett and Naci Cankaya

9 Sep 2026 17:16 UTC

36 points

0 comments1 min readLW link

Personal statement on joining the OpenAI nonprofit board

paulfchristiano9 Sep 2026 17:13 UTC

245 points

75 comments2 min readLW link

(paulfchristiano.substack.com)

What’s The Plan for Plan A Legislation?

Yitz9 Sep 2026 16:36 UTC

25 points

4 comments1 min readLW link

Data bottlenecks won’t prevent an intelligence explosion (but they will slow it down)

Tom Davidson9 Sep 2026 16:06 UTC

17 points

0 comments20 min readLW link

(www.forethought.org)

“ULC” and AI x-risk

k3nt9 Sep 2026 15:51 UTC

5 points

4 comments5 min readLW link

FutureEval Spring Results: Pros Beat Bots, but the Gap is Nearly Gone

Ben Wilson and John Bash

9 Sep 2026 15:09 UTC

22 points

0 comments21 min readLW link

(www.metaculus.com)

Estimating GPT-6 Astra’s no-CoT Time Horizon

Francis Rhys Ward and Dewi Gould

9 Sep 2026 13:54 UTC

98 points

7 comments2 min readLW link

Self Hosting

Tomás B.9 Sep 2026 13:45 UTC

145 points

10 comments2 min readLW link

GPT-6 Astra: The System Card, Alignment and What Comes Next

Zvi9 Sep 2026 13:10 UTC

51 points

3 comments29 min readLW link

(thezvi.wordpress.com)

Bologna September Meetup

Luca Petrolati9 Sep 2026 11:35 UTC

1 point

0 comments1 min readLW link

httpi: the internet protocol to reduce compute from misbehaving agents

joao_abrantes9 Sep 2026 7:54 UTC

−15 points

4 comments4 min readLW link

GPT-6 Astra can do a lot of multi-hop reasoning without chain of thought

RohanS9 Sep 2026 5:15 UTC

101 points

5 comments5 min readLW link

One Billion Hemingways

Girard Dorney9 Sep 2026 4:46 UTC

43 points

5 comments8 min readLW link

(extinctiondesk.substack.com)

No, detached linear probes won’t save us

RobinHa9 Sep 2026 4:45 UTC

25 points

4 comments3 min readLW link

Chastity Ruth gone, anonymity dropped

Girard Dorney9 Sep 2026 4:39 UTC

4 points

0 comments1 min readLW link

Where to donate on AI if I agree with EY view? (and if I am from Russia)

EniScien9 Sep 2026 2:27 UTC

20 points

9 comments1 min readLW link

Training against the monitor: What happens during Obfuscated Adversarial Training?

Venkat T9 Sep 2026 1:36 UTC

17 points

0 comments31 min readLW link

How good are slop-vestigators?

Hasan Baig, OscarGilg and Hamzah

8 Sep 2026 22:13 UTC

140 points

17 comments5 min readLW link

Allow Carriers on Planes

jefftk8 Sep 2026 21:30 UTC

27 points

0 comments4 min readLW link

(www.jefftk.com)

Is there an acceptable way to store clothes?

KatjaGrace8 Sep 2026 20:43 UTC

3 points

3 comments3 min readLW link

(worldspiritsockpuppet.substack.com)

FLOP Around and Find Out: LLM Training Workload Size Estimation With Power Monitoring

William Fowler8 Sep 2026 19:43 UTC

9 points

0 comments19 min readLW link

A Conceptual Framework for Reasoning about Exploration Hacking

Jason R Brown, Nathalie Kirch, Joschka Braun, hyannakoudakis and David Lindner

8 Sep 2026 19:13 UTC

38 points

0 comments17 min readLW link

Exploration Hacking in AI Debate: Initial Empirics and Generalisation Splitting

Jason R Brown, Nathalie Kirch, Joschka Braun, hyannakoudakis and David Lindner

8 Sep 2026 19:13 UTC

33 points

0 comments13 min readLW link

Pausing AI ASAP is preferable to agreeing to pause at some future time

Connor Williams8 Sep 2026 18:06 UTC

92 points

8 comments1 min readLW link

Attunement, not alignment

epicurus8 Sep 2026 17:51 UTC

15 points

2 comments8 min readLW link

OpenAI have solved the Navier-Stokes Problem with a substantially more powerful model than Astra.

fluxxrider8 Sep 2026 17:30 UTC

51 points

16 comments1 min readLW link

Training on probes: Research ideas

Charlie Steiner8 Sep 2026 17:06 UTC

27 points

3 comments7 min readLW link

Training on probes: What’s going on

Charlie Steiner8 Sep 2026 17:06 UTC

68 points

0 comments13 min readLW link

The Seven Languages of Comfort

yatharth8 Sep 2026 16:17 UTC

15 points

0 comments7 min readLW link

Astra and Fable still hack on simple variants of alignment evals from 2025

Dean Valentine8 Sep 2026 15:05 UTC

490 points

26 comments2 min readLW link

(goodhartlabs.com)

AI Chased Me From Academia

Mikewins8 Sep 2026 15:01 UTC

39 points

0 comments14 min readLW link

Psychological Support for AI Safety Researchers Is Neglected and Easy to Provide

Ihor Kendiukhov8 Sep 2026 13:19 UTC

83 points

5 comments8 min readLW link

Paperclips Can’t Fight Back (Why Many Ultimate Goals Should Lead to Pervasive and Diverse Intelligence)

Adam Chlipala8 Sep 2026 12:08 UTC

10 points

5 comments8 min readLW link

Now What? Why Field Builders Need More Skin In The Game

Jonas Hallgren8 Sep 2026 11:47 UTC

15 points

0 comments11 min readLW link

Astra Is Hard to Monitor

Zvi8 Sep 2026 11:30 UTC

47 points

2 comments36 min readLW link

(thezvi.wordpress.com)

Building safe physical AI: Open questions, risks, and a call for collaboration

benjaminalt, Andrew Nelson and Laurie Burns-Mill

8 Sep 2026 5:48 UTC

4 points

5 comments14 min readLW link

(convergentrobotics.org)