Yep, I just mean that it *feels* (could be very, very wrong) like OAI has much better post-training, Google/Anthropic have much better pre-training
MACHINE LEARNING
-
Anthropic’s Autopatch for Claude Code
By
–
Anthropic is working on the Autopatch feature for Claude Code, where it would automatically scan your repository for security vulnerabilities. h/t @btibor91
-
Pre-training vs Post-training: Google Anthropic OpenAI Strategy Comparison
By
–
It all makes sense. Google + Anthropic focused on pre-training, so you get to a great answer quicker, but the capability ceiling is lower. OpenAI focused on post-training, so great answers take longer, but you can push the models much further.
-
Optimiser la vitesse du code Vision par ordinateur
By
–
⚡ Speed up your Computer Vision code ⚡
— Satya Mallick (@LearnOpenCV) 12 décembre 2025
Is your pipeline running slower than it should? You might be cloning/copying images too often — and paying the cost in memory allocations and data movement.
💡 Quick fixes:
✅ Prefer assignment (Mat B = A) unless you truly need an… pic.twitter.com/ZbGJHC1DCOSpeed up your Computer Vision code Is your pipeline running slower than it should? You might be cloning/copying images too often — and paying the cost in memory allocations and data movement. Quick fixes: Prefer assignment (Mat B = A) unless you truly need an
-
Stanford SVL Hiring Software Developer for Robotics Simulation
By
–
Want to make a big impact in robotic research? Come and work with robots and the smartest students @StanfordSVL
!
We are hiring a software developer, focusing on simulation for robotics & robotic learning. You'll be working directly with me , @jiajunwu_cs and our amazing students -
Early adopters beat better chips through better data
By
–
Early adopters beat better chips. Early adopters fed better data to systems. Something I have known for a long time. Open AI has better users. It has better data.
-
LLMs as Skinnerian, Popperian, and Gregorian Creatures: Cognitive Limitations
By
–
Reflection on @dwarkesh_sp
's reflections on his interview with Rich Sutton, and why LLMs are exciting because they are Skinnerian, Popperian and Gregorian creatures. Minds with finite capacity cannot adapt forever without having to forget previous knowledge. This is true of -
FSD works without headlights because cameras see better at night
By
–
FSD works just fine without headlights. Cameras can see way more than you can at night. Take your iPhone out at night. You will see far more in the screen than your eyes can perceive.
-
World Model turns your house into a Holodeck
By
–
This is how a World Model turns your house into a Holodeck. https://t.co/veLd456wIH
— Robert Scoble (@Scobleizer) 12 décembre 2025This is how a World Model turns your house into a Holodeck.
-
Lightweight Discriminative Models: Overlooked ML Approach
By
–
Somehow I entirely missed the discriminative models that are lightweight, easy to understand, train, and infer.