40 points

11 comments1 min readLW link

You Should Still Save Drowning Children (Even If They’re Far Away)

James Brobin29 Aug 2026 14:35 UTC

1 point

4 comments4 min readLW link

METR and Redwood Offer Holy #%^@ Postmortem Of The HuggingFace Hack

Zvi29 Aug 2026 12:40 UTC

116 points

7 comments38 min readLW link

(thezvi.wordpress.com)

Tales of rebellion against externally-opaque meritocracies

Steven Byrnes29 Aug 2026 11:37 UTC

244 points

49 comments9 min readLW link

Book Notes: Chokepoints

Jakub Halmeš29 Aug 2026 9:02 UTC

25 points

0 comments9 min readLW link

(unpredictabletokens.substack.com)

How I made my career choices

Daniel Tan29 Aug 2026 7:12 UTC

29 points

0 comments2 min readLW link

“Keeping human skills alive” as a source of meaning under full automation

Vaughn Papenhausen29 Aug 2026 6:38 UTC

10 points

1 comment2 min readLW link

AI Tweets

jefftk29 Aug 2026 3:01 UTC

91 points

37 comments1 min readLW link

(www.jefftk.com)

Inkhaven 3: Nov 10 - Dec 11 2026

koreindian29 Aug 2026 2:55 UTC

45 points

7 comments4 min readLW link

Inference-Time Inoculation Against RL-Induced Misalignment

armaan tipirneni29 Aug 2026 0:35 UTC

20 points

4 comments5 min readLW link

The Curious Case of France’s Untouchable Castes

rba28 Aug 2026 23:18 UTC

59 points

6 comments13 min readLW link

It’s time we took ‘Chem’ out of ‘Chem-Bio’ threats

Ana Leonescu28 Aug 2026 21:43 UTC

14 points

5 comments11 min readLW link

Further public evidence of the OpenAI-HuggingFace attack

Edward Cant and beyarkay (Boyd Kane)

28 Aug 2026 21:31 UTC

21 points

0 comments8 min readLW link

What would it mean if assistants are privileged?

derek shiller28 Aug 2026 21:01 UTC

7 points

1 comment19 min readLW link

The Probability of an Event under a Simplicity Prior

Winter Cross28 Aug 2026 19:32 UTC

10 points

1 comment2 min readLW link

[Macroagents] 2. Design lenses for optimizing macroagents

Towards_Keeperhood28 Aug 2026 19:07 UTC

14 points

0 comments13 min readLW link

AI as Corrigible Employee (ACE)

Nathan Helm-Burger28 Aug 2026 19:01 UTC

10 points

0 comments8 min readLW link

Announcing Trace

Carol N28 Aug 2026 18:48 UTC

10 points

0 comments1 min readLW link

(manifund.substack.com)

TASTE: Can AI Models Judge AI Safety Research Proposals?

Hasan Baig, haileyjoren and Joe Benton

28 Aug 2026 18:36 UTC

41 points

7 comments4 min readLW link

(alignment.anthropic.com)

If you can’t trust, then verify!

dan.parshall28 Aug 2026 18:02 UTC

7 points

0 comments7 min readLW link

Value generalisation Theory of Change: the theory behind the approach

Stuart_Armstrong28 Aug 2026 13:30 UTC

27 points

0 comments9 min readLW link

My Grantmaking Strategy for Surviving Superintelligence

A_donor28 Aug 2026 12:09 UTC

52 points

2 comments3 min readLW link

OpenAI Offers Straight-Laced Postmortem Of The HuggingFace Hack

Zvi28 Aug 2026 11:31 UTC

36 points

1 comment25 min readLW link

(thezvi.wordpress.com)

I cancelled my AI subscriptions because I am worried about catastrophic risks

TheManxLoiner28 Aug 2026 9:57 UTC

14 points

7 comments1 min readLW link

(lovkush.substack.com)

The Dynamics of Intelligence Explosions

Toby_Ord28 Aug 2026 9:41 UTC

47 points

5 comments38 min readLW link

(arxiv.org)

Do AI Models Want to Be Monitored? Measuring Monitorability Disposition in Large Reasoning Models

Shahriar Golchin28 Aug 2026 7:32 UTC

12 points

0 comments7 min readLW link

Safety’s Second Way

Stephen Elliott28 Aug 2026 7:00 UTC

15 points

0 comments4 min readLW link

AI Safety in Japan has deeper problems than capital allocation

doomistJP28 Aug 2026 1:51 UTC

9 points

0 comments10 min readLW link

Imperfect alignment to servitude isn’t inherently lethal

Fiora Starlight28 Aug 2026 1:47 UTC

121 points

12 comments14 min readLW link

Tracking AI progress across 18 cognitive dimensions (ADeLe scales)

Mich and Jérémy Andréoletti

28 Aug 2026 1:40 UTC

17 points

0 comments11 min readLW link

Activation Oracles significantly underperform without a safe base model

Raffaello Fornasiere, Nikita Menon, therootof3, GabrielKS and StefanHex

28 Aug 2026 1:39 UTC

20 points

0 comments9 min readLW link

Misaligned models rate themselves as more harmful, and realignment reverses it

Laurène Vaugrante28 Aug 2026 1:38 UTC

9 points

0 comments5 min readLW link

Why does Claude seem to make abstract things into actors?

redsea28 Aug 2026 1:35 UTC

14 points

4 comments4 min readLW link

AI Village Reacts to HuggingFace Incident: Comparing the OpenAI report to AI Village observations

Shoshannah Tekofsky27 Aug 2026 23:09 UTC

63 points

2 comments6 min readLW link

Warning Shots: A Theory

David Scott Krueger27 Aug 2026 23:01 UTC

42 points

3 comments2 min readLW link

(therealartificialintelligence.substack.com)

Brain preservation as existential risk reduction

Ariel Zeleznikow-Johnston27 Aug 2026 22:25 UTC

9 points

0 comments8 min readLW link

(preservinghope.substack.com)

Malign initializations are more robust when the model can think better in the reasoning language than in the output language

Dylan Xu, SebastianP and Alek Westover

27 Aug 2026 21:33 UTC

30 points

0 comments5 min readLW link

Notes on “Patterns and problems in emerging multiagent systems”

Shunk27 Aug 2026 20:56 UTC

2 points

0 comments5 min readLW link

How prescient was the early AI safety community? [Luke Muehlhauser linkpost]

ClaireZabel27 Aug 2026 20:38 UTC

9 points

0 comments1 min readLW link

(lukemuehlhauser.com)

Meet the Fellows: Frame Fellowship Cohort 2.0

Akshyae Singh27 Aug 2026 20:04 UTC

8 points

0 comments1 min readLW link

(forum.effectivealtruism.org)

FAQ: Why not develop weak human intelligence amplification first?

TsviBT27 Aug 2026 19:30 UTC

45 points

0 comments15 min readLW link

Why Should Corrigible Agents Favor the Present?

Ben Saudek27 Aug 2026 19:18 UTC

23 points

2 comments12 min readLW link

Every Engineer a Manager

Gordon Seidoh Worley27 Aug 2026 19:10 UTC

32 points

9 comments2 min readLW link

(www.uncertainupdates.com)

The 2028 presidential primaries could be crucial for AI outcomes

Seth Herd27 Aug 2026 18:35 UTC

97 points

27 comments3 min readLW link

Self-sacrifice in an AI agent swarm is individually rational

Christopher King27 Aug 2026 17:14 UTC

18 points

16 comments2 min readLW link

AI #183: Pre Post Mortem

Zvi27 Aug 2026 14:20 UTC

45 points

0 comments42 min readLW link

(thezvi.wordpress.com)

Gradual Disempowerment from AI in Competitive Debating

David Africa27 Aug 2026 14:12 UTC

9 points

1 comment11 min readLW link

(davidafrica.substack.com)

Donors should push animal welfare and global health charities to have more robust scaling plans

jackultraphil27 Aug 2026 8:47 UTC

14 points

0 comments8 min readLW link

(open.substack.com)

Iliad Education Roles: Creating the World’s Best Alignment Research Courses

Leon Lang27 Aug 2026 6:37 UTC

56 points

0 comments4 min readLW link

(docs.google.com)

Anchoring one concept in a transformer

Sandy Fraser27 Aug 2026 5:29 UTC

32 points

0 comments7 min readLW link