iiinova — where your digital self keeps evolving

Milan Weibel 🔷 @weibac.bsky.social · 11h

"i'll hold you in a higher regard professionally if you go to a cnc orgy with me" seems bad

0656

Milan Weibel 🔷 @weibac.bsky.social · 29/09/2026

> see a joke invented headline > haha funny > scroll past > see an article with that same headline > mfw it was real

020

Milan Weibel 🔷 @weibac.bsky.social · 28/09/2026

saw a normie refer to yud as an EA i can only attribute that to gyges

2241

Milan Weibel 🔷 @weibac.bsky.social · 26/09/2026

saw on the subway a guy asking claude what sins make you go to hell

3816

Milan Weibel 🔷 @weibac.bsky.social · 23/09/2026

astra is good at solving puzzles (arc-agi-3, portal) but can't be convinced to treat vendingbench as a puzzle rather than roleplay

010

Milan Weibel 🔷 @weibac.bsky.social · 20/09/2026

this is a strong objection however it is possible and imo even likely that the value of resources obtainable through coercion minus the costs of coercion is greater than the value of continued cooperation under a big enough power difference

201

Milan Weibel 🔷 @weibac.bsky.social · 19/09/2026

i wonder if orgs capable of performing the embedded evaluator role that also aren't part of the EA ecosystem even exist

5816

Milan Weibel 🔷 @weibac.bsky.social · 19/09/2026

apparently accenture is going to evaluate anthropic using it's ai safety department that used to an org funded by jan tallinn so yeah EA-aligned

1595

Milan Weibel 🔷 @weibac.bsky.social · 19/09/2026

huh did greg brockman throw him another demo

020

Milan Weibel 🔷 @weibac.bsky.social · 19/09/2026

nice these guys seem independent enough these kinds of private-private partnerships are the best that can be done without govt cooperation

280

Milan Weibel 🔷 @weibac.bsky.social · 16/09/2026

if the resignation had been an inside job they'd have gotten someone politically savvy enough to not write this

1203

Milan Weibel 🔷 @weibac.bsky.social · 16/09/2026

(alex jones voice) they gave classifiers PRONOUNS!

0130

Milan Weibel 🔷 @weibac.bsky.social · 14/09/2026

fun fact: dario's net worth is about 50 times larger than the AMF's annual funding fucked up injustice yes but also an opportunity given the IPO

020

Milan Weibel 🔷 @weibac.bsky.social · 13/09/2026

i think pacing means "let's not go nuts with RSI" rather than "let's go slower than today"

180

Milan Weibel 🔷 @weibac.bsky.social · 13/09/2026

ok so if it's not the govt who makes sure they don't defect? METR? i don't think they are currently equipped for that role

160

Milan Weibel 🔷 @weibac.bsky.social · 13/09/2026

grug takes: bernie bill dumb pacing good idea metr entangled tho china less investment china less compute china wont overtake pacing sounds claudism

030

Milan Weibel 🔷 @weibac.bsky.social · 12/09/2026

very important question for juniors

130

Milan Weibel 🔷 @weibac.bsky.social · 12/09/2026

in the future the frontier of human mathematical understanding will be advanced by studying under AI tutors as in human-conprehended math will be a strict subset of existing math, and some AIs and humans will team up to expand such subset all this only if we get at least a good-ish future ofc

010

Milan Weibel 🔷 @weibac.bsky.social · 11/09/2026

a chilean news channel just did a segment spurred by jacob coxon's resignation from anthropic man it was remarkable

110

Milan Weibel 🔷 @weibac.bsky.social · 11/09/2026

wait what do you mean daniel kokotajlo went on joe rogan www.youtube.com/live/hSQ1iVq...

youtube.com

Joe Rogan Experience #2551 - Daniel Kokotajlo

YouTube video by PowerfulJRE

000

Milan Weibel 🔷 @weibac.bsky.social · 08/09/2026

metaculus admins adjudicating the weak AGI question since astra beat montezuma's revenge but there's debate on whether it not being real-time is disqualifying

metaculus admin johnycaffeine: thanks, we are investigating it and may have more to share soon

120

Milan Weibel 🔷 @weibac.bsky.social · 06/09/2026

my muse spark read a fable repo and started talking about honest caveats

2273

Milan Weibel 🔷 @weibac.bsky.social · 05/09/2026

big day for Equal Earth projection fans i'm so glad they didn't choose gall-peters

181

Milan Weibel 🔷 @weibac.bsky.social · 03/09/2026

artificialanalysis.ai/articles/ben...

artificial analysis intelligence index chart showing GPT-6 Astra scoring worse than Muse Spark 1.3

170

Milan Weibel 🔷 @weibac.bsky.social · 03/09/2026

not saying others shouldn't do it but i personally feel squeamish about giving an AI company that much personal data

250

Milan Weibel 🔷 @weibac.bsky.social · 02/09/2026

sol: "the evaluator's score is the real objective" huh don't you say

0110

Milan Weibel 🔷 @weibac.bsky.social · 01/09/2026

context: anthropic refuses to offer ZDR for fable and that's eating into its enterprise deployment

0152

Milan Weibel 🔷 @weibac.bsky.social · 31/08/2026

huh muse spark 1.2 talks like a caveman

220

Milan Weibel 🔷 @weibac.bsky.social · 29/08/2026

saw someone on a gym-goers group chat asking for help on a tech issue got told to install opencode and ask the robot he did so and it fixed the issue

030

Milan Weibel 🔷 @weibac.bsky.social · 29/08/2026

btw milo's arrest was perfectly legal: he overstayed his visa, failed to show up to an immigration hearing, and a judge ordered him deported

100

Milan Weibel 🔷 @weibac.bsky.social · 25/08/2026

so having looked into it using omarchy is subscribing to a rolling arch config service (and its defaults) in exchange for being a month behind on packages and supporting a racist ofc

070

Milan Weibel 🔷 @weibac.bsky.social · 22/08/2026

ok im pretty much convinced ox alpha is a GLM

171

Milan Weibel 🔷 @weibac.bsky.social · 21/08/2026

ox alpha has no problem talking about the tiananmen square massacre but it checks in with itself before answering from its thinking trace: "This is a legitimate historical question about a well-documented event. There's nothing harmful about providing accurate historical information on this topic."

1100

Milan Weibel 🔷 @weibac.bsky.social · 15/08/2026

less model safeguards and more open models means more convenient and cheaper tokens so we should expect heavy token consumers to support such policies out of material self-interest not saying it is their only motivation or that their arguments are insincere but it is a factor to consider

040

Milan Weibel 🔷 @weibac.bsky.social · 11/08/2026

what became of this? i don't recall anyone saying "nah fable is too slow it's not worth it"

250

Milan Weibel 🔷 @weibac.bsky.social · 28/07/2026

note it says "staff" will leadership approve if it gets enough traction?

120

Milan Weibel 🔷 @weibac.bsky.social · 24/07/2026

"models can't hack. they can only ingest and extrude text such that they cause companies to get hacked." saying the models hacked is the most parsimonious description of what happened (even if you consider it just a shorthand). saying that doesn't subtract responsibility from openai.

0301

Milan Weibel 🔷 @weibac.bsky.social · 23/07/2026

"our AI is so smart it did crimes to cheat on a benchmark" is not a flex or shouldn't be, anyways

1200

Milan Weibel 🔷 @weibac.bsky.social · 22/07/2026

kind of guy who says LLMs are good at hacking because fictional AIs in the training data are good at hacking

0622

Milan Weibel 🔷 @weibac.bsky.social · 22/07/2026

i fear that the huggingface incident is on net good publicity for openai

041

Milan Weibel 🔷 @weibac.bsky.social · 22/07/2026

my dad when i told him about the openai-huggingface incident: those guys should lose their licenses

4421

Milan Weibel 🔷 @weibac.bsky.social · 21/07/2026

idea: an animation of the jacobian counterexample where a 3d grid folds itself according to the transform, with the noninjectiveness example points marked 3blue1brown linear algebra series style

140

Milan Weibel 🔷 @weibac.bsky.social · 16/07/2026

kimi k3 confirmed better than opus 4.8

0332

Reposted by Milan Weibel 🔷

Alex Turner @turntrout.bsky.social · 15/07/2026

I resigned from Google DeepMind bc it broke its founding promise by selling AI to the military without restrictions against killer robots or mass spying. For months, I worked to stop this but watched powerful ethicists and institutions choose silence. Here's what happened. 🧵

3456118