In­tro­duc­ing Lunette: au­dit­ing agents for evals and environments

15 Dec 2025 23:17 UTC
23 points
0 comments1 min readLW link
(fulcrumresearch.ai)

Pri­vate AI clouds are the fu­ture of inference

perfectfwd15 Dec 2025 23:04 UTC
3 points
0 comments9 min readLW link
(perfectforward.substack.com)

Naming

CTA15 Dec 2025 23:00 UTC
3 points
0 comments4 min readLW link

View­ing an­i­mals as eco­nomic agents

foodforthought15 Dec 2025 18:13 UTC
10 points
2 comments5 min readLW link

Digi­tal Free­dom Fund open for grant ap­pli­ca­tions (Dead­line: 17th Fe­bru­ary)

gergogaspar15 Dec 2025 16:25 UTC
8 points
0 comments1 min readLW link

Луна Лавгуд и Комната Тайн, Часть 9

15 Dec 2025 16:01 UTC
2 points
0 comments1 min readLW link

Defend­ing Against Model Weight Exfil­tra­tion Through In­fer­ence Verification

15 Dec 2025 15:26 UTC
119 points
15 comments8 min readLW link

Ro­ta­tions in Superposition

15 Dec 2025 14:58 UTC
54 points
6 comments11 min readLW link

What is an eval­u­a­tion, and why this defi­ni­tion matters

Igor Ivanov15 Dec 2025 14:53 UTC
33 points
1 comment7 min readLW link

Con­scious stars

Alexandre Variengien15 Dec 2025 14:49 UTC
7 points
0 comments4 min readLW link
(alexandrevariengien.com)

A Case for Model Per­sona Research

15 Dec 2025 13:35 UTC
110 points
11 comments4 min readLW link

GPT-5.2 Is Fron­tier Only For The Frontier

Zvi15 Dec 2025 13:20 UTC
32 points
1 comment19 min readLW link
(thezvi.wordpress.com)

[Question] How to ac­count for mis­in­for­ma­tion when look­ing for effec­tive al­tru­ist causes?

SpectrumDT15 Dec 2025 13:13 UTC
8 points
2 comments1 min readLW link

Why even a sin­gle ASI might not nec­es­sar­ily have a sin­gle, per­ma­nent objective

Horosphere15 Dec 2025 12:21 UTC
1 point
0 comments2 min readLW link

Do you love Berkeley, or do you just love Lighthaven con­fer­ences?

Screwtape15 Dec 2025 7:48 UTC
82 points
4 comments5 min readLW link

When bits of op­ti­miza­tion im­ply bits of mod­el­ing: the Touchette-Lloyd theorem

15 Dec 2025 4:21 UTC
23 points
0 comments11 min readLW link

Notes on Soft­ware-Based Com­pute-Usage Verification

Alek Westover15 Dec 2025 3:40 UTC
8 points
0 comments12 min readLW link

De­sign­ing a Job Dis­place­ment Model

claywren14 Dec 2025 22:23 UTC
22 points
0 comments19 min readLW link

A high in­tegrity/​epistemics poli­ti­cal coal­i­tion?

Raemon14 Dec 2025 22:21 UTC
148 points
34 comments13 min readLW link

Fan­ning Radiators

jefftk14 Dec 2025 21:10 UTC
14 points
0 comments1 min readLW link
(www.jefftk.com)

Ab­strac­tion as a gen­er­al­iza­tion of al­gorith­mic Markov condition

Daniel C14 Dec 2025 18:55 UTC
8 points
0 comments7 min readLW link

No, Amer­i­cans Don’t Think For­eign Aid Is 26% of the Budget

Julius14 Dec 2025 18:47 UTC
67 points
17 comments5 min readLW link
(thegreymatter.substack.com)

A Life That Can­not Be A Failure

Bentham's Bulldog14 Dec 2025 16:40 UTC
−7 points
0 comments5 min readLW link

Should LLMs ac­cept in­vites to Ep­stein’s is­land?

Lukas Petersson14 Dec 2025 15:21 UTC
5 points
0 comments1 min readLW link
(lukaspetersson.com)

The Ax­iom of Choice is Not Controversial

GenericModel14 Dec 2025 4:08 UTC
42 points
29 comments7 min readLW link
(enrichedjamsham.substack.com)

Open Source Repli­ca­tion of the Au­dit­ing Game Model Organism

abhayesian14 Dec 2025 2:10 UTC
20 points
0 comments1 min readLW link
(alignment.anthropic.com)

Why did I be­lieve Oliver Sacks?

Eye You13 Dec 2025 23:39 UTC
68 points
17 comments1 min readLW link

In Fa­vor of Inkhaven-But-Less

Alice Blair13 Dec 2025 23:16 UTC
26 points
6 comments2 min readLW link

Micro-vi­sions for AI-pow­ered on­line content

Alexandre Variengien13 Dec 2025 23:05 UTC
11 points
0 comments8 min readLW link
(alexandrevariengien.com)

When is it Worth Work­ing?

foodforthought13 Dec 2025 21:40 UTC
23 points
1 comment6 min readLW link

[Question] What does “lat­tice of ab­strac­tion” mean?

Adam Zerner13 Dec 2025 21:19 UTC
11 points
8 comments1 min readLW link

Filler to­kens don’t al­low se­quen­tial reasoning

Brendan Long13 Dec 2025 20:22 UTC
74 points
5 comments1 min readLW link

Hog­warts’ 2025 Win­ter Sec­u­lar Sols­tice Celebration

Espedair Street13 Dec 2025 19:55 UTC
3 points
0 comments1 min readLW link

Toss a bit­coin to your Light­cone – LW + Lighthaven’s 2026 fundraiser

habryka13 Dec 2025 19:32 UTC
310 points
129 comments52 min readLW link

“Smarter brains run on sparsely con­nected neu­rons”

Pjain13 Dec 2025 18:27 UTC
12 points
0 comments1 min readLW link
(www.bio.psy.ruhr-uni-bochum.de)

Whack-a-mole: gen­er­al­i­sa­tion re­sis­tance could be fa­cil­i­tated by train­ing-dis­tri­bu­tion imprintation

lennie13 Dec 2025 17:46 UTC
23 points
0 comments14 min readLW link

Za­greb meetup: win­ter 2025

dominicq13 Dec 2025 17:10 UTC
5 points
0 comments1 min readLW link

Why the con­cept of AI al­ign­ment as it is cur­rently for­mu­lated is morally trou­bling

Horosphere13 Dec 2025 14:37 UTC
5 points
3 comments5 min readLW link

You Can Just Buy Far-UVC

jefftk13 Dec 2025 13:10 UTC
123 points
26 comments1 min readLW link
(www.jefftk.com)

How I stopped be­ing sure LLMs are just mak­ing up their in­ter­nal ex­pe­rience (but the topic is still con­fus­ing)

Kaj_Sotala13 Dec 2025 12:38 UTC
199 points
67 comments29 min readLW link

The Inevitable Evolu­tion of AI Agents

Steven McCulloch13 Dec 2025 11:51 UTC
32 points
1 comment9 min readLW link

Per­ma­nently Padding a Suitcase

jefftk13 Dec 2025 3:10 UTC
11 points
0 comments1 min readLW link
(www.jefftk.com)

Is the Con­structible Uni­verse All There Is?

GenericModel13 Dec 2025 2:29 UTC
18 points
2 comments8 min readLW link
(enrichedjamsham.substack.com)

Wages un­der superintelligence

Zachary Brown13 Dec 2025 0:06 UTC
33 points
1 comment1 min readLW link
(beforeporcelain.substack.com)

Trust is Nei­ther Scalar Nor a Snapshot

phoenix13 Dec 2025 0:05 UTC
6 points
3 comments3 min readLW link

Su­per­weight Da­m­age Re­pair in OLMo-1B uti­liz­ing a Sin­gle Row Patch (CPU-only Ex­per­i­ment)

sunmoonron13 Dec 2025 0:03 UTC
12 points
2 comments2 min readLW link

What’s guaran­teed in Life?

shanzson12 Dec 2025 21:27 UTC
−6 points
3 comments1 min readLW link

Eval­u­at­ing LLM hy­poth­e­sis gen­er­a­tion in biol­ogy is hard.

Austin Morrissey12 Dec 2025 20:48 UTC
1 point
0 comments6 min readLW link

Book Re­view: The Age of Fight­ing Sail

Suspended Reason12 Dec 2025 20:15 UTC
63 points
3 comments12 min readLW link

U.S. Democ­racy Threat In­dex: $10,000 in Fore­cast­ing Prizes

ChristianWilliams12 Dec 2025 19:43 UTC
4 points
0 comments1 min readLW link