Andrej karpathy could have charged $2,000 for this course.
He put it on YouTube. The full training stack. Tokenization. Neural network internals. Hallucinations. Tool use. Reinforcement learning. RLHF. DeepSeek. AlphaGo. 3 hours of the most comprehensive LLM education that
Another local video test and here MiniMax H3 vs Wan 2.2 wins
The film has 30 scenes, each made image-to-video from one of Jurek's photos. MiniMax received a description of the action, so he did the character movement and lip-sync himself. The odd scenes were played in parallel on one 3090, the even ones on one
Anthropic just revealed something wild about the AI race.
Some Chinese AI companies were allegedly using Claude behind the scenes to improve their own models. And in some cases, users didn’t even know Claude was involved. Anthropic’s latest threat report says it uncovered
When I created a dashboard for DeepSeek v4.1 Flash on A100x4, I was able to visualize points that should be further optimized.
Claude
You have reached the end of the archive
All of deepseek