16 Mar 2026 23:38 UTC

15 points

3 comments2 min readLW link

(fulcruminc.substack.com)

Types of Handoff to AIs

Daniel Kokotajlo16 Mar 2026 22:24 UTC

65 points

11 comments8 min readLW link

AICRAFT: DARPA-Funded AI Alignment Researchers — Applications Open

Mike Vaiana, Diogo de Lucena and Kvee

16 Mar 2026 21:44 UTC

67 points

8 comments4 min readLW link

You can’t imitation-learn how to continual-learn

Steven Byrnes16 Mar 2026 21:20 UTC

209 points

54 comments6 min readLW link

PSA: Predictions markets often have very low liquidity; be careful citing them.

Eye You16 Mar 2026 21:07 UTC

126 points

11 comments3 min readLW link

The Plan

Commander Zander16 Mar 2026 20:58 UTC

5 points

0 comments1 min readLW link

What Are My Values?

Corm16 Mar 2026 20:43 UTC

7 points

0 comments8 min readLW link

[Question] Seeking Suggestions for 2026 S-Process Recommenders

Ethan Ashkie16 Mar 2026 20:31 UTC

4 points

0 comments1 min readLW link

Carioca Rationalist meetup

Giskard16 Mar 2026 20:30 UTC

2 points

0 comments1 min readLW link

Three Properties for Alignment (and Why We’re Not Training Them)

Quentin FEUILLADE--MONTIXI16 Mar 2026 20:26 UTC

8 points

5 comments3 min readLW link

Do LLMs Have Stable Preferences?

Robert Gambee16 Mar 2026 20:09 UTC

8 points

0 comments7 min readLW link

(github.com)

The Fermi Paradox Implies Domination

Noam Makavy16 Mar 2026 20:04 UTC

1 point

3 comments2 min readLW link

Adding Typos Made Haiku’s Accuracy Go Up

bira16 Mar 2026 18:31 UTC

31 points

3 comments3 min readLW link

What are the best ways to publish rational fiction nowadays?

Ihor Kendiukhov16 Mar 2026 18:13 UTC

17 points

15 comments3 min readLW link

Compradorization

Benquo16 Mar 2026 16:10 UTC

110 points

11 comments18 min readLW link

(benjaminrosshoffman.com)

SFF-2026 S-Process Grant Round Application Announcement

Ethan Ashkie16 Mar 2026 15:57 UTC

18 points

0 comments11 min readLW link

(survivalandflourishing.fund)

Reasons to be pessimistic (and optimistic) on the future of biosecurity

Abhishaike Mahajan16 Mar 2026 15:45 UTC

29 points

2 comments41 min readLW link

(www.owlposting.com)

Customer Satisfaction Opportunities

Tomás B.16 Mar 2026 15:04 UTC

155 points

17 comments13 min readLW link

(open.substack.com)

We found an open weight model that games alignment honeypots

Thomas Read and Joseph Bloom

16 Mar 2026 12:57 UTC

80 points

2 comments10 min readLW link

Monthly Roundup #40: March 2026

Zvi16 Mar 2026 12:20 UTC

30 points

5 comments32 min readLW link

(thezvi.wordpress.com)

Will AI Progress Accelerate or Slow Down? Projecting METR Time Horizons

Alvin Ånestrand16 Mar 2026 11:54 UTC

13 points

0 comments10 min readLW link

(forecastingaifutures.substack.com)

Models differ in identity propensities

Jan_Kulveit, Raymond Douglas, vgel, owencb, David Duvenaud and Ondřej Havlíček

16 Mar 2026 10:45 UTC

61 points

0 comments14 min readLW link

Terrified Comments on Corrigibility in Claude’s Constitution

Zack_M_Davis16 Mar 2026 7:36 UTC

178 points

68 comments11 min readLW link

Hidden Role Games as a Trusted Model Eval

james.lucassen16 Mar 2026 4:46 UTC

14 points

1 comment9 min readLW link

(jlucassen.com)

Digital Dichotomy and Why it exists.

Yesh Chala16 Mar 2026 2:02 UTC

8 points

1 comment3 min readLW link

Brown math department postdoctoral position

Dmitry Vaintrob and Eric Larson

16 Mar 2026 1:09 UTC

21 points

0 comments1 min readLW link

Hello, World of Mechanistic Interpetability

ValueShift Research15 Mar 2026 23:36 UTC

8 points

4 comments5 min readLW link

(I am confused about) Non-linear utilitarian scaling

core15 Mar 2026 23:33 UTC

9 points

6 comments4 min readLW link

Futurekind Spring Fellowship 2026 - Applications Now Open

Khushbu Sainani15 Mar 2026 23:28 UTC

2 points

0 comments1 min readLW link

Schedule meetings using the Pareto principle

beyarkay (Boyd Kane)15 Mar 2026 21:18 UTC

2 points

5 comments2 min readLW link

(boydkane.com)

Was Anthropic that strategically incompetent?

StanislavKrym15 Mar 2026 20:11 UTC

14 points

0 comments4 min readLW link

(www.lesswrong.com)

What Are We Actually Evaluating When We Say a Belief “Tracks Truth”?

Alex Glaucon15 Mar 2026 19:59 UTC

2 points

4 comments6 min readLW link

Superintelligence Risk Education that Scales – Lens Academy

Luc Brinkman, meriton, Ouro, Chris-Lons and Alexandra R

15 Mar 2026 19:32 UTC

64 points

2 comments3 min readLW link

Emergent stigmergic coordination in AI agents?

David Africa15 Mar 2026 12:30 UTC

50 points

2 comments3 min readLW link

My Willing Complicity In “Human Rights Abuse”

AlphaAndOmega15 Mar 2026 10:42 UTC

258 points

37 comments11 min readLW link

Less Capable Misaligned ASIs Imply More Suffering

Ihor Kendiukhov15 Mar 2026 9:36 UTC

11 points

3 comments6 min readLW link

Rationalist Passover Seder in Maryland

Rivka15 Mar 2026 5:42 UTC

3 points

0 comments1 min readLW link

When do intuitions need to be reliable?

Anthony DiGiovanni15 Mar 2026 4:18 UTC

8 points

8 comments3 min readLW link

The Artificial Self

Jan_Kulveit, Raymond Douglas, vgel, owencb, David Duvenaud and Ondřej Havlíček

15 Mar 2026 1:37 UTC

132 points

13 comments29 min readLW link

Bridge Thinking and Wall Thinking

Jay Bailey15 Mar 2026 0:20 UTC

48 points

6 comments1 min readLW link

LLM Misalignment Can be One Gradient Step Away, and Blackbox Evaluation Cannot Detect It.

Yavuz Bakman15 Mar 2026 0:19 UTC

34 points

5 comments3 min readLW link

Walking Math

TickRate15 Mar 2026 0:16 UTC

15 points

2 comments6 min readLW link

How post-training shapes legal representations: probing SCOTUS opinions across model families

burnssa15 Mar 2026 0:15 UTC

7 points

0 comments8 min readLW link

Self-Recognition Finetuning can Reverse and Prevent Emergent Misalignment

Arush , Shawn Zhou, Jiaxin Wen and Shi

15 Mar 2026 0:11 UTC

48 points

24 comments7 min readLW link

Safe AI Germany (SAIGE)

Jessica Wang15 Mar 2026 0:10 UTC

6 points

1 comment7 min readLW link

Optimal (And Ethical?) Methods To Find “Optimal Running”

JenniferRM14 Mar 2026 23:16 UTC

9 points

0 comments10 min readLW link

‘Staying with it’ Done Wrong

Selfmaker66214 Mar 2026 22:38 UTC

18 points

0 comments1 min readLW link

(selfmaker.substack.com)

Mini-Munich Succeeds Where KidZania Fails

Novalis14 Mar 2026 22:24 UTC

37 points

0 comments3 min readLW link

(minicities.org)

Forecasting Dojo Meetup—postmortem discussion.

Vojtech Brynych14 Mar 2026 20:32 UTC

3 points

0 comments1 min readLW link

What concerns people about AI?

spencerg14 Mar 2026 19:24 UTC

34 points

2 comments3 min readLW link

(www.clearerthinking.org)