Blog — Intrane.fr — #ai-engineering

Engineering · August 2026

15 Lines of Bash Beat My 3,000-Line Search Engine

I built a Go search engine with TF-IDF, graph theory, and bidirectional stemming. It lost to a shell script embedded in a 79 KB binary. Then I made the Go engine learn from the shell script. Here's what happened.

#memgraph #supercli #skills.match #search #irony #ai-engineering

Language · July 2026

The Crash You Can Mail

The worst bug report is 'it crashed once in production and I can't reproduce it' — the normal case for concurrency. machin turns a concurrent crash into a byte-for-byte reproducible, self-contained, mailable artifact. It's sound because the language already proved the program race-free.

#machin #concurrency #record-replay #debugging #ai-engineering

Language · July 2026

The Port That Found a 30,000× Bug in My Own Language

SuperCLI already ran in Node.js and Zig. I ported it to my own language, machin, to dogfood — 10,158 plugins, one 69 KB binary, 24 tests. Then the live test hung for 30 seconds. The bug wasn't in my code; it was in my language's string model. Fixed it: 30,000× faster. This is why you dogfood.

#machin #supercli #dogfooding #performance #ai-engineering

Language · July 2026

The compiler that hands you the murder weapon

An AI writes most of the code now — and its failure mode is the plausible-but-wrong function. So I taught the machin compiler to be an adversary: it hunts for the exact input that breaks your code, hands it back with a runnable repro, and — when the input space is small enough — proves there isn't one. Honestly.

#machin #compiler #verification #ai-engineering #design-by-contract

Product · July 2026

Tailwind Without Node: 172 kB, 16 ms, and a Way to Prove It

My agent-first design system compiles its stylesheet in 16 ms from a 172 kB binary — no Node. The trick isn't the rewrite: it's differential-testing every emitted rule against the real tailwindcss binary, so “100% compatible” is a green check, not a claim.

#machin #tailwind #design-system #wasm #ai-engineering

Language · July 2026

A 1B-parameter LLM at 20 tokens/second, in a language you've never heard of

I challenged machin — my machine-first language — to run TinyLlama-1.1B at 20 tok/s on a laptop CPU, pure MFL, no C kernels. It hit 21.8, token-identical to llama2.c, and 4x faster than the reference C single-threaded. Here's the 2.1 -> 21.8 arc and what it drove into the compiler.

#machin #llm #inference #performance #ai-engineering

Language · July 2026

The Language That Maintains Itself

First machin compiled itself. Then I handed its maintenance to an autonomous agent fleet: 104 merged PRs in three days, 40 issues cleared, trust score 57→100. Here's the honest version — the five systems bugs autonomy actually required, the gates that held, and the fix an agent claimed that a repro disproved.

#machin #ai-agents #autonomous-maintenance #compiler #ai-engineering

Language · July 2026

Fearless Concurrency Without the Send/Sync Tax

Rust guarantees data-race freedom, but you pay for it in Send/Sync bounds and lifetimes — a tax on the human. I built the same class of guarantee into machin, inferred by the compiler with zero annotations. Here's why a machine-first language makes that tractable.

#machin #concurrency #compiler #ai-engineering

Language · July 2026

The Day My Language Started Compiling Itself

A year ago I built a programming language for AI agents. This year it crossed the line that separates a toy from a real compiler: it started compiling itself — byte-for-byte, faster than the original. Here's how, and why it's the clearest proof of how I build software with AI.

#machin #compiler #self-hosting #ai-engineering