I ate bear fat with honey and salt flakes, to prove a point

aggliu4 Nov 2025 2:00 UTC
255 points
34 comments5 min readLW link
(signoregalilei.com)

Leg­ible vs. Illeg­ible AI Safety Problems

Wei Dai4 Nov 2025 21:39 UTC
203 points
44 comments2 min readLW link

You’re always stressed, your mind is always busy, you never have enough time

mingyuan1 Nov 2025 22:07 UTC
180 points
6 comments3 min readLW link
(mingyuan.substack.com)

Lack of So­cial Grace is a Lack of Skill

Screwtape3 Nov 2025 4:43 UTC
147 points
21 comments6 min readLW link

What’s up with An­thropic pre­dict­ing AGI by early 2027?

ryan_greenblatt3 Nov 2025 16:45 UTC
147 points
12 comments20 min readLW link

Pub­lish­ing aca­demic pa­pers on trans­for­ma­tive AI is a nightmare

Jakub Growiec3 Nov 2025 13:04 UTC
134 points
6 comments4 min readLW link

The Un­rea­son­able Effec­tive­ness of Fiction

Raelifin3 Nov 2025 15:35 UTC
121 points
11 comments8 min readLW link
(raelifin.substack.com)

Re-rol­ling environment

Raemon1 Nov 2025 21:46 UTC
120 points
2 comments2 min readLW link

Com­par­a­tive ad­van­tage & AI

Simon Lermen3 Nov 2025 21:50 UTC
114 points
25 comments4 min readLW link

Leav­ing Open Philan­thropy, go­ing to Anthropic

Joe Carlsmith3 Nov 2025 17:38 UTC
104 points
23 comments18 min readLW link

Ilya Sutskever De­po­si­tion Transcript

anaguma2 Nov 2025 21:06 UTC
96 points
1 comment28 min readLW link

Eras­mus: So­cial Eng­ineer­ing at Scale

Martin Sustrik3 Nov 2025 5:20 UTC
87 points
7 comments4 min readLW link
(www.250bpm.com)

Heroic Responsibility

johnswentworth4 Nov 2025 23:26 UTC
78 points
28 comments2 min readLW link

LLM-gen­er­ated text is not testimony

TsviBT1 Nov 2025 14:47 UTC
77 points
74 comments11 min readLW link

Peo­ple Seem Funny In The Head About Sub­tle Signals

johnswentworth6 Nov 2025 4:03 UTC
76 points
17 comments5 min readLW link

An­thropic Com­mits To Model Weight Preservation

Zvi5 Nov 2025 21:30 UTC
72 points
9 comments14 min readLW link
(thezvi.wordpress.com)

The Tale of the Top-Tier Intellect

Eliezer Yudkowsky3 Nov 2025 20:21 UTC
71 points
49 comments35 min readLW link

Try­ing to un­der­stand my own cog­ni­tive edge

Wei Dai3 Nov 2025 8:49 UTC
63 points
13 comments4 min readLW link

The Zen Of Max­ent As A Gen­er­al­iza­tion Of Bayes Updates

4 Nov 2025 0:02 UTC
61 points
8 comments7 min readLW link

Hu­man Values ≠ Goodness

johnswentworth2 Nov 2025 19:24 UTC
60 points
29 comments6 min readLW link

“What’s hard about this? What can I do about that?”

Raemon3 Nov 2025 5:30 UTC
59 points
0 comments9 min readLW link

Re­search Reflections

abramdemski4 Nov 2025 4:33 UTC
59 points
2 comments3 min readLW link

A 2032 Take­off Story

romeo6 Nov 2025 0:20 UTC
55 points
7 comments34 min readLW link

Crime and Pu­n­ish­ment #1

Zvi3 Nov 2025 15:30 UTC
49 points
4 comments45 min readLW link
(thezvi.wordpress.com)

Why Is Print­ing So Bad?

johnswentworth1 Nov 2025 21:37 UTC
45 points
23 comments2 min readLW link

Halfhaven halftime

Viliam2 Nov 2025 21:29 UTC
43 points
8 comments4 min readLW link

Halfway to Anywhere

Screwtape6 Nov 2025 4:27 UTC
42 points
0 comments6 min readLW link

Be­ing “Use­fully Con­crete”

Raemon4 Nov 2025 22:15 UTC
41 points
4 comments4 min readLW link

A prayer for en­gag­ing in conflict

TsviBT4 Nov 2025 8:19 UTC
41 points
0 comments2 min readLW link

Why and how you should make your home smart (it’s cheap and se­cure!)

Mikhail Samin3 Nov 2025 3:27 UTC
39 points
3 comments8 min readLW link
(mikhailsamin.substack.com)

GDM: Con­sis­tency Train­ing Helps Limit Sy­co­phancy and Jailbreaks in Gem­ini 2.5 Flash

4 Nov 2025 16:25 UTC
38 points
2 comments6 min readLW link
(arxiv.org)

Build the life you ac­tu­ally want

mingyuan4 Nov 2025 4:50 UTC
37 points
2 comments3 min readLW link
(mingyuan.substack.com)

OpenAI: The Bat­tle of the Board: Ilya’s Testimony

Zvi4 Nov 2025 19:30 UTC
37 points
1 comment5 min readLW link
(thezvi.wordpress.com)

Meta-agen­tic Pri­soner’s Dilemmas

TsviBT5 Nov 2025 16:44 UTC
37 points
1 comment5 min readLW link

Model­ing the geopoli­tics of AI development

4 Nov 2025 17:31 UTC
37 points
0 comments2 min readLW link
(ai-scenarios.com)

My YC Pitch

Tomás B.2 Nov 2025 10:27 UTC
36 points
1 comment2 min readLW link

Thoughts by a non-economist on AI and economics

boazbarak4 Nov 2025 17:06 UTC
35 points
0 comments14 min readLW link

New home­page for AI safety re­sources – AISafety.com redesign

5 Nov 2025 10:33 UTC
34 points
2 comments1 min readLW link

FTL travel and sci­en­tific realism

Adam Scherlis2 Nov 2025 6:03 UTC
34 points
5 comments4 min readLW link
(adam.scherl.is)

A glimpse of the other side

mingyuan3 Nov 2025 4:00 UTC
33 points
5 comments2 min readLW link
(mingyuan.substack.com)

Reflec­tions on 4 years of meta-honesty

GradientDissenter2 Nov 2025 5:29 UTC
32 points
6 comments6 min readLW link

A/​B test­ing could lead LLMs to re­tain users in­stead of helping them

Daniel Paleka4 Nov 2025 19:30 UTC
28 points
0 comments4 min readLW link
(newsletter.danielpaleka.com)

Weak-To-Strong Generalization

abramdemski2 Nov 2025 2:45 UTC
28 points
0 comments9 min readLW link

How to be con­vinc­ing when talk­ing to peo­ple about ex­is­ten­tial threat from AI

Mikhail Samin5 Nov 2025 7:01 UTC
27 points
2 comments5 min readLW link

Maxwell’s De­mon and the Ar­row of Time

Adam Scherlis5 Nov 2025 7:35 UTC
26 points
2 comments6 min readLW link
(adam.scherl.is)

AI #141: Give Us The Money

Zvi6 Nov 2025 14:50 UTC
25 points
1 comment48 min readLW link
(thezvi.wordpress.com)

Not Over Or Un­der Indexed

Screwtape4 Nov 2025 22:54 UTC
25 points
0 comments6 min readLW link

How to sur­vive un­til AGI

Nikola Jurkovic5 Nov 2025 1:17 UTC
25 points
3 comments3 min readLW link
(nikolajurkovic.substack.com)

Met­formin 1000mg/​day upon symp­tom on­set may re­duce your risk of long covid by 10-30%

Drake Thomas2 Nov 2025 4:57 UTC
24 points
1 comment8 min readLW link

Freewrit­ing in my head, and over­com­ing the “twinge of start­ing”

ParrotRobot1 Nov 2025 1:12 UTC
23 points
1 comment6 min readLW link