iiinova — where your digital self keeps evolving

Reposted by Benjamin

Ethan Mollick @emollick.bsky.social · 21/09/2026

I mean that the AI companies will not be allowed to substitute for doctors & lawyers, even if they are better at some/all tasks than doctors and lawyers, as jobs are also socially determined and have political & social ties that go beyond the job function. See Abbott's "Theory of the Professions"

4461

Reposted by Benjamin

🌱️ @crumb.bsky.social · 16/09/2026

Xiaomi are streaming their latest RL run live! (peep the cost raising by many dollars per second) mimo.xiaomi.com/rl/

Screenshot of a reinforcement learning training monitor for Xiaomi's MiMo models. Two runs are in progress: mimo-v2.6-pro at step 10 (cost $704,895) and mimo-v2.6-flash at step 15 (cost $303,769). A notification bar lists node restarts and result updates, and the main panel shows a benchmark comparison line chart alongside multiple training metric graphs.

1505

Benjamin @premium-content.bsky.social · 16/09/2026

Sounds too good to be true, but also quite plausible. I think we'll see more of this. When the hype started, it led to a lot of funding not just for LLMs but also for alternative approaches. They are the first to reach paydirt, but I expect others to follow.

010

Reposted by Benjamin

Zach Weinersmith @zachweinersmith.bsky.social · 15/09/2026

Can we please stop doing this? I made a comic about it and everything!

142240411

Reposted by Benjamin

Sichu Lu @sichulu.bsky.social · 15/09/2026

when you make a meme on the internet bky special because no one really follows me here anyway

813013

Reposted by Benjamin

Richard McElreath 🐈‍⬛ @rmcelreath.bsky.social · 12/09/2026

I do wonder what David Hilbert would think about all this. He was accused of trying to automate mathematics, in the 1920s! The "Hilbert program" failed in an exact sense, because Gödel, but it succeeded by making formal proof systems like Lean possible. And Lean makes the AI-sausage-machine work.

Two panels from the graphic novel Logicomix.
Left: He wants a machine to feed it axioms and make theorems, like one where a pig enters the one side...
Right: ...and the sausages come out from the other!! (laughter)

3749

Benjamin @premium-content.bsky.social · 12/09/2026

I ended up making a bunch of comments to Draghi's opinion piece.

010

Reposted by Benjamin

Pekka Lund @pekka.bsky.social · 10/09/2026

Can we now forget Buckmaster's claims? "On Wednesday evening, in response to questions from The Times, OpenAI said in a statement that it was “categorically” impossible for its A.I. system to have been influenced by anything Dr. Buckmaster had done in the past two months."

nytimes.com

The Mathematician Crushed Between OpenAI and Anthropic Over a Math Problem

Tristan Buckmaster was on the path toward an important proof when one of the A.I. giants used its staggering resources to get there first.

2223

Reposted by Benjamin

Colin @colin-fraser.net · 10/09/2026

A rogue AI agent has solved the Hodge Conjecture so that it can use the prize money to buy a phone number in order to complete its assigned phishing task

025723

Reposted by Benjamin

ellie lockhart (she/they) @eleanor.lockhart.contact · 09/09/2026

Was having a convo at work about how Astra is much easier to work with for what we do than Opus 5 and someone was like “we’re doing toxic workplace gossip about coworkers but they’re robots, you realize that right”

1398

Benjamin @premium-content.bsky.social · 09/09/2026

This might be the consequence of AISI's major screw-up with Mythos.

000

Reposted by Benjamin

Ethan Mollick @emollick.bsky.social · 08/09/2026

Weakly General AI achieved (at least according to the criteria set in Metaculus in 2020) & a year before the most recent predictions: ✅Loebner prize was a weak Turing Test, equivalent achieved by GPT-4.5 ✅Winograd passed by GPT-3 ✅SAT passed at 75% by GPT-4 ✅Montezuma's Revenge beaten by GPT-6

3387

Benjamin @premium-content.bsky.social · 08/09/2026

It's obviously a joke but maybe it isn't.

020

Reposted by Benjamin

Zach Weinersmith @zachweinersmith.bsky.social · 08/09/2026

The drama over Navier-Stokes matters a lot, but it is worth emphasizing that we don't know the truth yet. In any case, if any ai-human combo have a proof or near-proof, that's ultimate what matters for humanity. If you still think LLMs are no good for math, it is time to update your mental model.

1012911

Benjamin @premium-content.bsky.social · 08/09/2026

He's right, but... Now what? "If European AI is reduced to Mistral, then we're finished," the [French Finance] minister told reporters in San Francisco. "We can't put all our eggs in one basket: it's a whole ecosystem that needs to be developed" fr.finance.yahoo.com/actualites/i...

fr.finance.yahoo.com

IA : le ministre Roland Lescure appelle l’Europe à ne pas tout miser sur Mistral

Le ministre de l’Economie Roland Lescure alerte sur les limites d’une IA européenne qui reposerait essentiellement sur Mistral. Selon lui, l’Europe doit bâtir un écosystème plus large pour garantir sa...

110

Reposted by Benjamin

Key 🗝 🦊✅ @keytryer.net · 07/09/2026

There's someone literally making entire full-length Doctor Who episodes in their bedroom. I'm not going to speak to their quality since I haven't finished them all, but that there isn't already a full GOT Final Season remake is because there isn't enough will. www.youtube.com/watch?v=-aAY...

youtube.com

Episode 1 | Doctor Who Reboot

YouTube video by The Unmade Season

3162

Reposted by Benjamin

mr. TIM @timkellogg.me · 04/09/2026

obviously a distilled model

Claude is a LinkedIn thought leader with sudo.
It'll delete your repo, write a beautiful apology, and ask if you'd like the apology turned into an Artifact.
Every correction becomes a fucking restorative justice circle. "You're right to hold me accountable." I'm trying to get a Docker container running, Nelson Mandela. Nobody asked you to heal.
Claude Shannon gave us information theory, and this motherfucker honors his legacy by using 4,000 tokens to say "good point."

2403

Benjamin @premium-content.bsky.social · 05/09/2026

Failures in defense policy may eventually be paid for in blood. It is no surprise that countries closer to Russia have no patience for protectionist folly. It's scary how ignorant the political leaders are of Europe's position. Europe has actively ...

200

Reposted by Benjamin

Rudi Bachmann @bachmannrudi.bsky.social · 04/09/2026

Es gibt eine offizielle Stellungnahme des @vfsecon.bsky.social, verfasst von Regina Riphahn, Kerstin Schneider und mir (unter Mithilfe einiger KollegInnen) zum drohenden Verlust von Daten in Deutschland. Bitte RT und mit Euren Kontakten in der Politik sprechen. www.socialpolitik.de/de/kommissio...

socialpolitik.de

Zustimmung der Cookies | Verein für Socialpolitik e.V

03421

Reposted by Benjamin

Sung Kim @sungkim.bsky.social · 04/09/2026

When you are the most profitable company in the world, this happens: NVIDIA reportedly priced the deal at $12.9303 billion as a nerdy joke: The “129303” part is the decimal version of U+1F917. U+1F917 is the Unicode code for the 🤗 hugging-face emoji. Hugging Face uses 🤗 as its signature emoji.

714814

Reposted by Benjamin

Joe Tidy BBC News @joetidy.bsky.social · 03/09/2026

This was surreal. I emailed the proton mail of the hacker group that breached the UK airports yesterday and they replied overnight with this...

912824

Reposted by Benjamin

Bluesky Safety @safety.bsky.app · 02/09/2026

Dogpiling or harassing users is unacceptable. We have taken recent action against accounts for this behavior, and we will be increasing our efforts, consistent with our Community Guidelines.

bsky.social

Community Guidelines - Bluesky

8703113546

Reposted by Benjamin

Zach Weinersmith @zachweinersmith.bsky.social · 01/09/2026

Just because there's a lot of pessimism on this site, did you know that a ton of African economies are finally hitting rapid development? E.g.

1247549

Reposted by Benjamin

rahaeli @rahaeli.bsky.social · 31/08/2026

Yet another instance of the EU's "reining in Big Tech" laws killing off another Small Alternative because dealing with the DSA is too onerous for labor-of-love hobby forums to deal with

47840

Reposted by Benjamin

Jan @butwhosjan.bsky.social · 29/08/2026

Happy 30-year anniversary to the Democratic National Convention dancing to Macarena

56739185

Reposted by Benjamin

Daphne Keller @daphnek.bsky.social · 28/08/2026

I know we are inundated daily with claims about the dangers social media poses to kids. But an overwhelming predictor of suicide attempts -- 80% of attempts in this study -- is abusive parents or caretakers. pmc.ncbi.nlm.nih.gov/articles/PMC...

pmc.ncbi.nlm.nih.gov

Child abuse and the prevalence of suicide attempts among those reporting suicide ideation

Victims of child abuse may be at increased risk of acting on suicide ideation, although this has not been empirically tested. We estimated the risk of suicide attempts associated with child abuse amon...

03012

Reposted by Benjamin

samhain ⎔ @personhood.removal.surgery · 28/08/2026

The ones whose posts were too good to make the other pack: go.bsky.app/7bzo85z

9664

Reposted by Benjamin

David Greene @davidgreene.bsky.social · 26/08/2026

In the Meta settlement, "Age Inappropriate Accounts" includes those that regularly share "Restricted Goods & Services." As we explained in our recent submission to the Meta Oversight Board, that category commonly includes information about abortion medication. www.eff.org/deeplinks/20...

01417

Reposted by Benjamin

dame @dame.is · 27/08/2026

between the state department beginning to criminalize technology use on the left and the left’s decision to stigmatize technology use within its own ranks, the far right is positioning itself to maintain an upper-hand unless we do something about it

214127

Reposted by Benjamin

Ted Underwood @tedunderwood.com · 27/08/2026

MIT report is superb. It acknowledges disruption, concedes different disciplines will handle this differently, foregrounds people and social interaction — but also says something obvious I had just about given up hope institutions could say: aiandeducation.mit.edu/report/

3.1.1. Revisit course goals
Before considering how to construct Al-aware assessments, instructors should reconsider their goals for student learning- in every subject they teach. What should students know or be able to do by the end of the course?

29523

Reposted by Benjamin

Mike Masnick @masnick.com · 27/08/2026

In case you wanted about 3,700 words on the problems with today's "settlement" that gives Meta way more power over how the internet works than it should have.

121537643

Reposted by Benjamin

mr. TIM @timkellogg.me · 26/08/2026

hell yeah the huggingface 🤗 incident is still in the news OpenAI finally releases a full complete tech report openai.com/index/huggin...

openai.com

The Hugging Face incident and the road ahead

OpenAI shares findings from the Hugging Face security incident and the steps we’re taking to strengthen AI model security, monitoring, and alignment.

2234

Reposted by Benjamin

Daphne Keller @daphnek.bsky.social · 26/08/2026

This thread has my scattered takes on the Meta settlement agreement, at first skim. It covers a lot of the big pieces, despite beginning with a more tangential interest of mine about the MPA (nee MPAA). I think a settlement like this should be unconstitutional. But it's a weird new legal Q.

2227

Reposted by Benjamin

P(aul) Frazee @pfrazee.com · 25/08/2026

X shuts down apps that show its posts. Bluesky exists to ernable it. One of these networks is about promoting freedom, the other is not.

431589253

Reposted by Benjamin

jae @fubarchitect.com · 23/08/2026

there is now ~1 ~human ~verifiably in ~hell

934458

Reposted by Benjamin

Ethan Mollick @emollick.bsky.social · 15/08/2026

As Infocom games are now open source, I had Codex update 1987’s word game “Nord and Bert Couldn't Make Head or Tail of It” for playing now. It keeps O'Neill’s puzzles, though its easier to win with a new UX (but the AI did better translating some puzzles than others) nord-and-bert.netlify.app

3302