Issue 21, Oct 10, 2026
The Unified Diff Trap
Senior AI Engineer interview at Google, and the interviewer asks:
“Your coding agent has exactly one tool: bash. It can grep, sed, patch, and run pytest. So why does it keep botching file edits, and what edit interface would you give it instead of unified diffs with line numbers?”
Don’t say: “Use a stronger model”
Why line-number drift and shell-escaping quietly destroy your coding agent's edit success rate, and how content-anchored search/replace fixes it.
Share this trapShare on LinkedIn
The full answer, with the mechanism and the arithmetic, is for paid subscribers on Substack.