Failure Modes

What goes wrong: mature codebases that break agents, silent failures that compile but misbehave, and harness assumptions that decay as models improve.


The Day-50 Problem. Agents work on greenfield, break on mature projects.

Silent Failures. When code compiles but behavior is wrong.

The Omakase Tradeoff. Why control matters more than defaults.

Harness Assumptions Decay. Every component encodes a model limitation that may already be stale.