This new Michael Levin interview from Balázs Kégl is truly captivating. youtu.be/3zqI0iG428c?si… Kudos, Balazs, on your interviewing technique, and your own honest questions and pushback.

The narrative around LLMs is that they got better purely by scaling up pretraining *compute*. In reality, they got better by scaling up pretraining *data*, while compute is only a means to the end of cramming more data into the model. Data is the fundamental bottleneck. You can't

How is it possible that Claude Sonnet 4.5 is able to work for 30 hours to build an app like Slack?! The system prompts have been leaked and Sonnet 4.5's reveals its secret sauce! Here’s how the prompt enables Sonnet 4.5 to autonomously grind out something

What's your first reaction?

This is insane. New AI model from Samsung, 10,000x smaller than DeepSeek and Gemini 2.5 Pro just beat them on ARC-AGI 1 and 2 Samsung’s Tiny Recursive Model (TRM) is about 10,000x smaller than typical LLMs yet smarter because it thinks recursively instead of just predicting

An unscripted and wide-ranging conversation that ends with more questions than answers. In other words, the best kind of conversation. I’ve been following Balázs on X for almost 2 years. To be able to sit down in conversation with him has been an amazing opportunity. And

Please watch Geoffrey Hinton’s talk before replying. Please read Schrodinger’s book “what is life?” as well as the follow up book with the same title by another Nobel Laureate. I also highly recommend Daniel Dennett’s books. Or ask your favourite LLM 😉 Let’s please increase

📢 New paper alert !! How to use Policy Gradient methods without explicit rewards? We address this question in our new work "From Data to Rewards: a Bilevel Optimization Perspective on Maximum Likelihood Estimation" 📜 huggingface.co/papers/2510.07… 🖥️ github.com/abenechehab/nl… 1/🧵

🤝 Co-led with our brilliant intern: Gabriel Singer ❤️ Huge thanks to our wonderful collaborators at Huawei Noah's Ark Paris (Corentin LEGER,Youssef Attia El Hili,Albert Thomas @albertcthomas.bsky.social,Balázs Kégl ), Cognizant AI Lab (Giuseppe Paolo), and KAUST (Maurizio Filippone).

This is somewhat of a half-baked post, but I find it fascinating that given all of the scientific breakthroughs we’ve made, we still lack a compelling unifying theory of emotions The majority of research tends to optimize for ‘regulation’ aka supression + coping strategies, as

Once you accept that consciousness is real, and denying it leads to a bad performative contradiction, you have no choice but to accept woowoo. The only choice you have is where to put it.

Dr John Dee, Elizabeth I's court magician, thought that manipulating the right symbols and gazing into a black mirror was a way of summoning nonhuman entities We on the other hand are much less superstitious

Is there a structural difference between the belief that a complex but essentially random word generator has will, and astrology (a complex but essentially random system has causal power on the world)? Genuine question, asking for all the four quadrants.