What Is AI Context Engineering? Giving Coding Agents the Right Context

Chroma tested 18 models and found performance dropped as input length grew, even when the task stayed the same, according to research the company published on context rot. The finding challenges the common practice of giving coding agents more instructions when they break team rules, since additional material can degrade agent performance.

When a coding agent breaks a rule your team agreed on, the usual instinct is to give it more instructions, but more material can make the agent worse. Chroma tested 18 models https://www.trychroma.com/research/context-rot and found performance dropped as input length grew, even when the task stayed the same.