Really cool, here is Deepseek v4 Pro’s one-shot at it
Terminal makes me more money than the UI ever did.
I can run 50 agents instead of 10 and stay model agnostic. I spread across Claude, Gemini, DeepSeek, GLM, and Grok so I never bet everything on one model.
How DeepSeek Attention Works - Paper Explained
YouTube: Full video: sparse attention, the indexer, shared KV cache and local windows, with animations and code. Become AI researcher in 90 days:
I had deepseek create a nostalgic one-key game. The direction of the running line (up and down) can be…
I had deepseek create a nostalgic one-key game. The direction of the running line (up and down) can be switched by key input (left and right are reversed at the edge of the screen). If you touch an existing line, it's over. There is a bug in the score display...
You have reached the end of the archive
All of deepseek