Thorsten Ball: devs still judge model code by human readability — when does that stop?

steipete · x · 2026-09-17

Thorsten Ball (who leads Amp) observes that people still judge model-generated code by how hard it is for humans to understand or modify, and wonders when that metric will stop mattering. The remark touches a core debate in the coding-agent era: whether human readability remains the right yardstick when code is increasingly written and consumed by AI.

Original post →

More from coding & agent

coding & agent channel →