The polynomial fit on the Germany map haha. Congrats on the bioML model, this space is going to explode in the next few years.
RESEARCH
-
Knowledge Cutoff Will Persist in All Future AI Models
By
–
Knowledge cutoff is a thing. It will be in every model in the future, forever. So we good 🙂
-
Model specialization: temporary feature or fundamental architectural design?
By
–
The specialization angle is interesting. I wonder if this changes fast with each model update or if it's a more fundamental architectural difference.
-
Nature Publication V2 Reasoning Models Advancement Potential
By
–
Congrats on the Nature publication! The fact that V1 was built before reasoning models even existed makes the potential of V2 with current models pretty incredible.
-
China’s Open-Source AI Model Ecosystem: Insights and Developments
By
–
94K views on a China AI deep dive, people are hungry for real on-the-ground insights. Would love to hear more about the open-source model ecosystem there specifically.
-
Evaluation loops critical for self-improving AI agents
By
–
Self-improving agents are exciting but the evaluation loop is everything. How do you make sure it's actually improving and not just drifting? 🙂
-
Meta Launches TRIBE v2 Foundation Model for Brain Activity Prediction
By
–
Meta is back: Meta just dropped TRIBE v2, a foundation model that predicts how your brain responds to sight, sound, and language.
— Chubby♨️ (@kimmonismus) 26 mars 2026
Trained on 500+ hours of fMRI data from 700+ people, it can predict a new person's brain activity without any retraining, and its predictions are… https://t.co/kjdqTadDiC pic.twitter.com/jmvN4XTYiYMeta is back: Meta just dropped TRIBE v2, a foundation model that predicts how your brain responds to sight, sound, and language. Trained on 500+ hours of fMRI data from 700+ people, it can predict a new person's brain activity without any retraining, and its predictions are
-
OpenAI’s Product Strategy: Spud and Pre-training Focus
By
–
Curious to see if spud lives up to the hype. OpenAI's been juggling a lot of (questionable) products lately (e.g. Sora, which might be a good idea but way too early to be fully adopted), focusing on a big pre-train push might be what they need to go back ahead.
-
Sanctions Law and Open Research in Machine Learning
By
–
This is a really important and underappreciated issue. The intersection of sanctions law and open research feels like it could have huge consequences for the ML community.
-

GTR Framework Prevents AI Agents Getting Stuck in Reasoning Loops
By
–
How can we stop AI agents from getting 'stuck' in their own thoughts? Researchers from Tsinghua University, Tencent, and Peking University introduce GTR (Guided Thought Reinforcement). This novel framework uses an automated corrector to evaluate and refine an agent's reasoning