Eugene Yan · 2026-05-03 · notable
Eugene Yan: How to Work and Compound with AI
Eugene Yan's framework for effective AI collaboration: treat context as infrastructure, encode preferences as config files, build verification loops, delegate increasingly complex tasks, and compound learnings across sessions.
A practical framework for compounding your output with AI — from an ML engineer who lives it.
What is it?
Eugene Yan's comprehensive guide to systematically working with AI models, built around five principles that compound over time: structured context, preference config, verification systems, task delegation, and session refinement.
How does it work?
The framework treats each AI session as a unit of organizational learning — preferences get encoded as reusable configs, mistakes inform verification checklists, and complexity delegated to AI grows over time.
Why does it matter?
Most guides on 'using AI' are superficial. Yan's framework is built on production ML experience and models AI collaboration as a compounding system, not a collection of tips.
Try it
https://eugeneyan.com/writing/working-with-ai/