matz · 2026-04-24 · notable
Spinel: Matz Releases a Self-Hosting Ruby AOT Compiler — 11.6× Faster, Built With Claude
Ruby creator Yukihiro Matsumoto released Spinel, a self-hosting AOT compiler that turns Ruby into standalone native binaries with no runtime deps. Built in ~1 month with Claude. 11.6× geometric mean speedup over CRuby across 28 benchmarks; 86.7× for Game of Life. Debuted at RubyKaigi 2026.
Ruby's creator built a self-hosting AOT compiler in one month with Claude — it runs Ruby 11.6× faster as a standalone native binary.
Key specs
| GitHub stars | 1.1k |
|---|---|
| Geometric mean speedup vs cruby | 11.6× |
| Game of life speedup | 86.7× |
| Benchmarks measured | 28 |
What is it?
Spinel is an ahead-of-time compiler for Ruby released April 24, 2026 by Yukihiro 'Matz' Matsumoto and debuted at RubyKaigi 2026. Written in Ruby (self-hosting — the 21,000-line backend compiles itself), it converts Ruby source to standalone native executables linked only against libc and libm. Supported: classes, inheritance, mixins, blocks, exceptions, pattern matching, fibers, BigInteger, Regexp. Excluded by design: eval, send, method_missing, threads — trades dynamism for static analysis tractability. Matz built it in approximately one month using Claude as a coding partner.
How does it work?
Spinel parses Ruby via libprism (the official Ruby parser), serializes the AST, then runs whole-program type inference followed by C code generation. The generated C is compiled with a standard C compiler into a native binary. The bootstrapping is iterative: Spinel compiles itself, verifying self-hosting at each step. The type inference is whole-program — it sees all method calls before generating any code — which enables tighter optimization than JIT approaches that compile hot paths incrementally.
Why does it matter?
Spinel is the most dramatic Ruby performance project since YJIT, and notable on two levels. Technically, the 11.6× geometric mean speedup (86.7× for Game of Life) shows what's possible when Ruby's dynamic features are traded for static compilation — directly relevant for compute-heavy Ruby workloads like scientific computing or CLI tooling. Culturally, Matz building a 21,000-line compiler in a month with AI assistance is a striking demonstration of AI-accelerated systems programming from one of the field's most respected language designers.
Who is it for?
Ruby developers running compute-intensive workloads, CLI tool authors, systems programmers
Try it
git clone https://github.com/matz/spinel