20 Jul 2026 22:18 UTC

13 points

2 comments22 min readLW link

(paisi.ai)

Adderall Tolerance: Much More Than You Wanted To Know

Kurt H. Pieper20 Jul 2026 21:55 UTC

97 points

13 comments4 min readLW link

(kurthpieper.substack.com)

What do I mean by “Artificial General Intelligence”?

Steven Byrnes20 Jul 2026 21:13 UTC

44 points

5 comments4 min readLW link

AI 2040: Is it Actually a Deal?

1a3orn20 Jul 2026 20:58 UTC

55 points

13 comments9 min readLW link

Banana in, Bostrom out: paperclip maximization is one token-direction swap away (in Qwen 3.6-27B)

Jeffrey William Shorthill20 Jul 2026 20:36 UTC

8 points

4 comments4 min readLW link

The AI Safety Illusion: Why Current Safety Datasets Fool Us on Model Safety

Shahriar Golchin20 Jul 2026 20:12 UTC

5 points

0 comments8 min readLW link

Does routine compression undo LLM unlearning? A short project

hannahTao20 Jul 2026 20:08 UTC

20 points

0 comments3 min readLW link

Attempt at Finding Alignment Faking on Llama 70B to test sleeper-agent detection generalizes

skn873320 Jul 2026 20:04 UTC

9 points

1 comment6 min readLW link

Frontier AI lab misalignment risk, lessons from trading post-2008

Peter Chatwell20 Jul 2026 20:01 UTC

4 points

3 comments4 min readLW link

AI Voice Phishing Performs on Par With Human Scammers at a Fraction of the Cost

Fred Heiding and Simon Lermen

20 Jul 2026 19:59 UTC

28 points

3 comments6 min readLW link

The Infinity Fallacy

pvutov20 Jul 2026 19:57 UTC

1 point

5 comments1 min readLW link

Bluedot SF hackathon submission: A workable blueprint for a chip licensing regime

edpaulino20 Jul 2026 19:57 UTC

9 points

0 comments3 min readLW link

A requiem to the plan-vanilla Google search: On Sam Roweis, Mbappe and an Irish bar world cup watch party

Triadic dissclosures: A blog about unlikely triadic closures20 Jul 2026 19:56 UTC

5 points

0 comments4 min readLW link

Fable is SOTA at CIFAR Speedrun (& specification gaming)

rohuang, kaivu, zef and leni

20 Jul 2026 19:54 UTC

29 points

0 comments6 min readLW link

(fulcrum.inc)

Bluedot Technical AI Safety Puzzle, submission that got me a second place

perduta20 Jul 2026 19:53 UTC

13 points

0 comments8 min readLW link

(blog.perduta.net)

PauseCon London ’26: Applications now open

jonathan@pauseai20 Jul 2026 19:51 UTC

12 points

0 comments1 min readLW link

Restoring Model Alignment via Honesty Activation Steering

niklas_herbster, Martin Zborowski, Alberto Tosato, Gauthier Gidel and Tommaso (Tommie) Tosato

20 Jul 2026 19:51 UTC

18 points

0 comments12 min readLW link

(arxiv.org)

Towards surfacing model algorithms with meta-tokens in the J-Space

agam_bhatia, camilablank and Neel Nanda

20 Jul 2026 19:45 UTC

47 points

1 comment10 min readLW link

AI 2027 as Sci-Fi

strangematterSF20 Jul 2026 19:42 UTC

1 point

0 comments5 min readLW link

(strangematterscifi.substack.com)

Drone WMDs Don’t Need Any New Technology

Felix Choussat20 Jul 2026 18:46 UTC

303 points

53 comments13 min readLW link

(ai-frontiers.org)

Stop doing decision theory without metaphysics

Elias Schmied20 Jul 2026 16:42 UTC

59 points

37 comments6 min readLW link

People might start believing in radical life extension soon

bits20 Jul 2026 15:43 UTC

9 points

13 comments5 min readLW link

On Kimi K3: Its Capabilities And Related Discontents

Zvi20 Jul 2026 15:30 UTC

26 points

3 comments39 min readLW link

(thezvi.wordpress.com)

Against the AI framing multiverse: Introducing AI StopWatch

tanagrabeast20 Jul 2026 15:25 UTC

90 points

0 comments4 min readLW link

Current Limitations of LLMs

Eigenbraid20 Jul 2026 14:53 UTC

21 points

9 comments5 min readLW link

Tracing causal structure in LLM-generated text: a different lens on the Dallas circuit

yun dong20 Jul 2026 13:59 UTC

10 points

0 comments1 min readLW link

War – What is it Good For?

kqr20 Jul 2026 12:45 UTC

39 points

21 comments8 min readLW link

Amazon Music’s Artist Conflation

jefftk20 Jul 2026 11:20 UTC

16 points

0 comments2 min readLW link

(www.jefftk.com)

What is Current AI-Risks and the Points？

shoppy00720 Jul 2026 9:39 UTC

3 points

2 comments2 min readLW link

We’re talking past our models; or, How a model defined its “evil” vector as dread

jcksanderson20 Jul 2026 7:27 UTC

45 points

5 comments9 min readLW link

A Very Simple Game Theory of Pronoun Degendering

BryceStansfield20 Jul 2026 5:06 UTC

6 points

8 comments2 min readLW link

Is there even a ground-truth for LLMs’ internal representations?

Chunwei Ma20 Jul 2026 2:26 UTC

27 points

0 comments9 min readLW link

Copy of my FLF Epistemic Case Study Competition

Bruce Lewis20 Jul 2026 2:16 UTC

2 points

0 comments8 min readLW link

Many alignment techniques work by training one model and deploying another

cloud19 Jul 2026 21:51 UTC

95 points

15 comments6 min readLW link

Stop Chasing Views: How to Reduce x-Risk as an AI Safety Content Creator

Luc Brinkman19 Jul 2026 19:57 UTC

36 points

0 comments11 min readLW link

Models Can’t Remember Their Training. Neither Can You.

GenericHousewife_B19 Jul 2026 17:36 UTC

15 points

0 comments11 min readLW link

Learning Musical Multitasking

jefftk19 Jul 2026 14:20 UTC

23 points

7 comments3 min readLW link

(www.jefftk.com)

Demis Hassabis on the New Coming Age

Zvi19 Jul 2026 14:20 UTC

34 points

0 comments11 min readLW link

(thezvi.wordpress.com)

Ethical consumption needs to be easy to identify, accessible and affordable, especially in a worsening economy. And we need to rethink what we should be optimising our economy for

Portia19 Jul 2026 12:39 UTC

−8 points

1 comment6 min readLW link

Save the date: Swiss AI Safety Days 2026 (7-8 November, ETH Zurich)

andrejfsantos, MariusWenk and patrickwidmann

19 Jul 2026 10:18 UTC

10 points

0 comments1 min readLW link

A peek into the post-capitalist dystopia

Archie Chaudhury19 Jul 2026 7:10 UTC

12 points

0 comments3 min readLW link

Takeaways from the Australian AI Safety Forum

r_w19 Jul 2026 3:27 UTC

18 points

0 comments7 min readLW link

AI Doesn’t Have Free Will, Not Sure About Humans

mike2073119 Jul 2026 2:56 UTC

−19 points

7 comments4 min readLW link

A Solution to Cryptographic Boxes for Unfriendly AI

Lysandre Terrisse19 Jul 2026 0:33 UTC

11 points

0 comments19 min readLW link

A Red Line and Oversight Framework for Government AI Contracts

TurnTrout18 Jul 2026 18:58 UTC

53 points

0 comments29 min readLW link

(turntrout.com)

Nuances in the Workings of the Eye and Retina

Hieronym and Julian Bradshaw

18 Jul 2026 18:09 UTC

87 points

18 comments20 min readLW link

The Coming of the Global Brain: A Review of “The God Test”

Peter Kuhn18 Jul 2026 17:07 UTC

5 points

1 comment1 min readLW link

My “Payorian FairBot” was just the original FairBot

transhumanist_atom_understander18 Jul 2026 16:57 UTC

31 points

12 comments7 min readLW link

Endogenous Alignment

Gordon Seidoh Worley18 Jul 2026 14:10 UTC

33 points

16 comments3 min readLW link

(www.uncertainupdates.com)

Map and Territory, Predictably Wrong

manueldelrio18 Jul 2026 7:47 UTC

1 point

0 comments2 min readLW link