The Last Human Code Review: Building Trust in AI-Generated Code — Itamar Friedman, Qodo
Source
youtube.com
Author
AI Engineer
Date
Context
Automated code review needs knowledge of the system's obligations to recognize changes that break them. In Itamar Friedman's Qodo talk, the proposed context windowThe maximum amount of text a model can consider at once — its working memory for the current conversation or task.Full definition → includes architectural relationships, service contracts and lessons from past outages. Much of that knowledge currently lives in developers' heads or scattered discussions. Making it maintainable by people and usable by agents would let a reviewer assess how changes interact across the system, including several changes under development at once. His argument shifts the review problem toward capturing engineering knowledge that a code diff alone does not contain.
Terms in this piece · Glossary
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.