Deterministic modernization in an age of AI
summary
Argues that deterministic, repeatable upgrades beat agentic guessing: use OpenRewrite recipes to codify changes, run them across billions of lines and hundreds of thousands of repos, and reach roughly 95% automation while the last 5% is handled with targeted tooling. Shows concrete tools—trigraph for fast indices, a pre-think index, an MCP server and agent skills for Claude/Copilot—claims around 30% token reduction and far shorter runtimes versus multi-hour probabilistic runs that exhaust token budgets.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 3:51Code modernization in 2026
- 7:02AI agent limitations
- 9:29Incremental migration waves
- 16:11Automating recipe creation
- 24:17Closing migration gaps
- 34:24Modernization takeaways
- 35:49Q&A
related talks
45mTrusting AI to Modernize Software at Scale
17mBounded Autonomy: Between Free Will and Determinism
50mFrom Prompt Hacking to Architectural Determinism: Engineering Reliable GenAI Systems
43mTaming the Unpredictable: Reliability in Chaos
52mThe Agentic Shift- Are We Reasoning, Or Just Retrieving Smarter?
01h 00mKenote: AI Didn’t Replace You. It Promoted You!
36mHarnessing the Power of Nondeterminism for Software Engineering
27mWill AI (Finally) Replace Developers?