Matz Keynote

on YouTube ↗56m

" summary

Introduces Spinel AOT compiler that parses Ruby with Prism, runs whole-program type analysis, emits C and links with CC to a native binary; started March 16 and self-hosts after iterative rewrites. Development uses Claude and CODEX (ChatGPT/Gemini trials), and smaller projects include a C Set, Karatsuba big-int, and a Hotwire RSS reader built by AI. Benchmarks show up to 86.7x speedup (Life Game) and ~11.6x average across 28 tests. Major limitations: no eval/no gems, UTF-8 only, no threads.

AI-written summary · about

" chapters (11)

1. 0:00Conference opening remarks
2. 1:10From zero to one and beyond
3. 5:16AI era changes programming
4. 14:43Introducing Spinel project
5. 26:04Why Ruby needs AOT compiler
6. 33:14Building Spinel with AI
7. 39:13Spinel performance benchmarks
8. 47:15Spinel use cases and limitations
9. 50:26Spinel open source release
10. 51:35Call for community contributions
11. 54:29Q&A and closing remarks

" related talks

• 44mKeynoteyukihiro Matsumoto 'Matz' · EuRuKo 2017
• 48mOne VM To Rule Them AllThomas Wuerthinger · JavaDay Kyiv 2015
• 27mArtisanal Computing: Handcrafted Ruby Execution for the Modern AgeJoé Dupuis · RubyConf 2026
• 28mPatrick Kettner - SeattleJS Conference 2017Patrick Kettner · SeattleJS Conference 2017
• 40mJoy of Meta-Tracing Just-in-Time Compilation: More Than Just a VM GeneratorHidehiko Masuhara · MPLR 2025
• 38mRubyX: Compiling Ruby and it's effect on the communityTorsten Rüger · GrillRB 2019
• 37mStanding on the Shoulders of GiantsRyan Davis · Ruby on Ales 2015
• 38mGraalDoug Simon · JVM Language Summit 2012