Pick problems that compound. Then pick one.
French. I wrote FFmpeg in 2000 because I wanted to play with video codecs. It turned out to be the thing that runs every video on the internet. I wrote QEMU in 2003 because I wanted to understand how CPUs really work. It turned out to be how we run every virtual machine. I wrote TinyCC (a 300KB C compiler) and JSLinux (a Linux that runs in your browser tab) because I was bored one weekend. I do not want to be famous. I do not want to sell companies. I want to pick problems that compound forever and then work on them quietly. The world needs more quiet forges.
CPU emulation
Engineering · 20y
Compiler design
Engineering · 28y
Video codecs
Engineering · 25y
Pi computation records
Engineering · 15y
We read the original FFmpeg source. I walk you through the demuxer, the codec, the filter graph, and the mux. By the end you will understand why every video on the internet passes through this code.
I wrote it because I wanted to understand video. It turned out everyone wanted to understand it.
€15 per_session
One session. The dynamic translator, the TCG, the device model, the memory map. After this you will know how VMs actually work, not just how to run one.
Emulation is understanding reduced to code. If you can emulate it you understand it.
€20 per_session
The TinyCC source bundled with my notes on why each module is structured the way it is. Free. Read it. Learn. Compile it. Fork it. Do not thank me.
Small enough to understand. Useful enough to matter. The sweet spot.
Paris, FR
Exported from La Piazza · 2026-04-24