“When you don’t have hundreds of billions of dollars of funding, you’re actually forced to work together and collaborate." https://t.co/Xkw2CjWFit

I’m co-organizing an “AI for Science: Algorithms to Atoms” social event during #NeurIPS2025 with Yann LeCun, Anima Anandkumar, Bill Dally, and Max Welling If you want to talk about AI Scientist, World Models and the future of AI-driven discovery, please come on Dec 5 3:30pm PT! https://t.co/0r7fHKwwUR

At the #BloombergNewEconomy Forum in Singapore, I discussed the future of AI, and why open-source is how rest of the world can work together to develop AI. When you don’t have hundreds of billions of dollars of funding, you’re forced to collaborate. Constraints breed innovation. https://t.co/YEvpvA9mND

@jacobrintamaki Have you seen the connection between scaling laws and induction heads? https://t.co/3UJCI4E1QU https://t.co/HIdvkcAiRj

@jacobrintamaki Looking at loss as a function of token index (as seen here) was very much influenced https://t.co/UmYHpmeZX5

More progress on Claude's alignment! https://t.co/zjfTQkpW75

⛵Marin 32B Base (mantis) is done training! It is the best open-source base model (beating OLMo 2 32B Base) and it’s even close to the best comparably-sized open-weight base models, Gemma 3 27B PT and Qwen 2.5 32B Base. Ranking across 19 benchmarks: https://t.co/LOOikMkruy

Another problem was that we saw this: https://t.co/fEHQsLeUnV

To get the details, here’s the full retrospective: https://t.co/AOX8rXivpW For even more details, here’s the GitHub issue with various links: https://t.co/RXYgzAIMoc including to the code for the full reproducible experiment: https://t.co/LhVqVVr4tY

The model is also on Hugging Face, though bear in mind it’s just a raw base model, so we wouldn’t recommend using it for production without additional fine-tuning: https://t.co/yZFD0JjCZ5

In the AI ecosystem, who supplies the data? the compute? the models? We just released a new tool on the AI Supply Chain. Our dataset reveals how AI models, data, compute, capital, and even talent change hands. Here’s why you should care 👇 https://t.co/bFspYmK6cG

Announcing Olmo 3, a leading fully open LM suite built for reasoning, chat, & tool use, and an open model flow—not just the final weights, but the entire training journey. Best fully open 32B reasoning model & best 32B base model. 🧵 https://t.co/vnGrArA44X

Inspired by @karpathy’s NanoChat, I created a simple experiment for running a full pre-mid-post-training run using Marin! The whole thing runs on a single v5p-8, so researchers can reproduce the whole thing at extremely low cost on the TPU Research Cloud! https://t.co/IXt6zXv5H2

It is one thing for everyone from @sama to @ylecun to @elonmusk and the Twitterverse to have attacked me literally for years for saying that scaling alone would not get us to AGI. Another to pretend now that that never happened — now that I have largely been proven correct. https://t.co/n77ChV5lrQ

Wow. Just wow. @ylecun taking credit for my March 2022 argument that scaling would hit a wall and that LLM would not bring us to AGI--after he initially attacked me for saying it and continued to promote them right up until ChatGPT ate has lunch--has to among the most intellectually dishonest things I have seen a scientist do in my entire life. @SchmidhuberAI was right about LeCun.

interesting post from @boazbaraktcs: https://t.co/3SXF6mvV8h

GPT-5.1 (Thinking High) is about 300 times cheaper per task than o3-preview (Low) while scoring only a few points lower on ARC-AGI-1. 1 year later intelligence has gotten 300 times cheaper. This is why I can’t stand people who say “wahh the models too expensive” it will become cheaper.

METR (50% accuracy): GPT-5.1-Codex-Max = 2 hours, 42 minutes This is 25 minutes longer than GPT-5. https://t.co/NgqG3E5LfB

3 years ago we could showcase AI's frontier w. a unicorn drawing. Today we do so w. AI outputs touching the scientific frontier: https://t.co/ALJvCFsaie Use the doc to judge for yourself the status of AI-aided science acceleration, and hopefully be inspired by a couple examples! https://t.co/5pxuUp9x3r

Princeton’s Breakthrough Qubit Could Finally Make Quantum Computing Practical https://t.co/4eAD8vKPti https://t.co/m0LhQo0tkS

Princeton’s Breakthrough Qubit Could Finally Make Quantum Computing Practical https://t.co/4eAD8vKPti https://t.co/m0LhQo0tkS

Submitted for your approval... 💥The Parallax Zone. 👉Where the 20-watt human campfire and the megawatt AI fusion reactor stare at the same world from opposite ends of reality. 👉Neither wins and the depth only appears in the angular difference between them. 👉This isn’t merger, this is stereoscopic cognition. More in @PsychToday 👇👇👇 https://t.co/odJW54VBa1

Ilya Sutskever: Humans aren’t AGI; they know little and learn constantly. Pretrained AGI on the other hand, overshoots by skipping that trial-and-error phase. Real intelligence comes from ongoing learning, not from starting out “finished.” https://t.co/VBzrs761ez

Should we change the name X back to Twitter ?!! Please be honest’ A. Yes B. No https://t.co/16itTmntWN

Should we change the name X back to Twitter ?!! Please be honest’ A. Yes B. No https://t.co/16itTmntWN

Vlad Tenev’s New AI Startup Harmonic Secures $120M At A $1.45B Valuation With Zero Revenue https://t.co/yrE1xISEgN

Vlad Tenev’s New AI Startup Harmonic Secures $120M At A $1.45B Valuation With Zero Revenue https://t.co/yrE1xISEgN

NEW: @JTLonsdale on @SquawkCNBC blasts AI safetyism, calls for one national standard. "You have states trying to ban AI in healthcare that are going to prevent us from making health care cheaper... That safetyism is what Europe does. If you want to be a poor civilization...then you should move to Europe. But America needs to stay free." Joe Lonsdale: This has been an issue we're working on all year, obviously close friends with @DavidSacks and @mkratsios47 running this in the administration. And listen, we are on the verge of something amazing for our civilization. We have tens of thousands of builders in our country. They're going to bring down the cost of health care. They're going to make houses cheaper. Permits should only take hours instead of months. Just like education is going to be better for all of our kids and personalized. They're going to fix our civilization. And yet on both the far left and the far right, you have a lot of populists who for, for really nefarious reasons, are opposing this. There are reasonable people, by the way, who have some concerns, and there's ways to compromise... But you have a lot of crazy populists. You have a patchwork of just really intense stuff they're doing that would break all of this, and we can't let that happen. Andrew Ross Sorkin: And to you, the thing that you can't break is this idea of states taking this into their own hands?  Joe: I'm a federalist, Andrew, on most issues. I have teams in 23 states with @InstituteCicero... I think it's reasonable for states to do a lot of things as laboratories of democracy... What's unreasonable is you had Biden NGOs, but pushed by the last administration. You have European-like rules, and you have populists trying to put in crazy new regulatory authorities that demonize the builders to demonize the tech. And if you force us to hop through regulatory agencies in every state, China is going to win and the builders are going to lose.  ...there's always an argument in our country between safety and liberty. And I think people, especially if they're not in the details, are right to be concerned and want to be safe. But if you're going to use safety to have arbitrary fines or arbitrary penalties where every state, including the radical blue states like New York, are going to be able to break our builders, that's not okay. And listen, you have states trying to ban the use of AI in health care right now in ways that are going to prevent us from making health care cheaper. You have New York State already banning it in education. This is insane. Public schools in New York are not allowing AI. They're not even in any kind of school in New York. You have to call yourself a homeschool to use AI right now in New York, which means we can't personalize it and lift up the lives of the kids who are not doing well in inner cities. This is crazy stuff, Andrew. That safetyism is what Europe does. If you want to be a poor civilization...then you should move to Europe. But America needs to stay free. ...a random legislator in New York state is not going to stop at the AI apocalypse. They're only going to harass and slow us down and make us lose to China.

Music fans enraged as AI-generated ‘singer’ tops the holiday charts: ‘Seriously?’ https://t.co/nWGjZEqvrv

Music fans enraged as AI-generated ‘singer’ tops the holiday charts: ‘Seriously?’ https://t.co/nWGjZEqvrv

‘Digital colonialism’: how AI companies are following the playbook of empire https://t.co/66EyuWw47F @ConversationUS @ConversationEDU

‘Digital colonialism’: how AI companies are following the playbook of empire https://t.co/66EyuWw47F @ConversationUS @ConversationEDU