An incredible mise en abyme happened yesterday. A developer managed to get 2x improvement by letting DeepSeek-r1 write code for a specific part of llama.cpp. 

Surprisingly, 99% of the code in this PR is written by DeekSeek-R1. The only thing I do is to develop tests and write prompts (with some trials and errors)

According to the prompts shared by the developer, the model spent thinking 3 to 5 minutes per response.