Welcome

Describing yourself is always a somewhat artificial exercise. I'm not a ninja — even though the picture might suggest otherwise — but I do prefer working in the shadows.

If I had to sum up what I've been doing for thirty years: I encounter systems that don't work, and I try to understand why.

Sometimes it's a bit in the wrong place. Sometimes it's an undersized processor. Sometimes it's an assumption that no one questions anymore. And every time, understanding is what makes the difference.

I've been writing code since childhood. My first program was at six years old on a ZX81. At twelve, I wrote professional software that was used for thirty years. Since then, I never really stopped. C, C++, Go, JavaScript, Rust are part of my daily work. But I have a particular affinity for low-level layers: assembly (x86_64, SPARC, ARM, RISC-V, Z80...), processor architecture, operating systems. More recently, I've also been working with FPGAs (Verilog, VHDL).

I contributed to Plan 9, the successor to Unix designed at Bell Labs by the creators of C and Unix. First with hardware drivers — Matrox, ATI, VIA Gigabit Ethernet — then by identifying and fixing a bug in the 9vx scheduler, a component that had been mathematically validated. The fix, confirmed by the community, still stands to this day.

I've worked on systems where failure is not an option. A math library in assembly for flight computers, with precision required down to 10⁻¹⁵. A memory protection mechanism deployed on ten million machines — zero crashes. An application virtualization solution to sandbox programs in isolated environments. A signature-less antivirus, built for the French Navy and the Ministry of Justice. What connects these projects is the same reflex: understand what's really happening, measure, don't guess.

Today, I'm building a local LLM inference engine, optimized for CPU, in Rust and hand-written assembly (AVX-512, NEON). My focus is on where performance actually breaks: memory bandwidth, latency, cache hierarchy, execution ports. I document this work in a series of technical articles. The goal: a sovereign engine, resource-efficient, controlled end-to-end.

In parallel, I share what I do: writing, open source, and technical videos — no staging, just code, systems and ideas.

Beyond that, I love science, reading papers, understanding, digging deeper. And staying close to what truly matters.

— Philippe Anel