Harmonique

Play radio

Notes

LLM writes low-level code and improves performance by 2x

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.

DHH's absurd rant on ADHD

I got pissed off a bit too easily reading one more absurd take on ADHD. This time, it comes from DHH who blessed us with his weakly-informed opinion. Apparently, we ADHDers aren't crippled enough in our daily life that we have to suffer arrogant apes throwing feces blog posts at us. ADHD is a serious condition and society at large must face it instead of dismissing it.

ADHD is a subject of research since at least 1775

Continue reading →

I was wrong about Prediction Markets

Donald Trump just got elected, and while polls showed a very close race, prediction markets such as Polymarket were confident he would win. But why? I couldn't understand how they could be so certain when regular polling showed such unclear results.

I watched an interview with a French trader who invested millions of dollars betting on Trump's victory. This helped me understand what was really happening. Instead of writing about his research like a journalist would, this trader did careful analysis and put his money behind his findings. He spotted the difference between public perception and reality – and made a lot of money for his work.

Continue reading →