AMD AI director says Claude Code is becoming dumber and lazier since update

Build a regression testing framework specifically for agentic code tools that tracks 'laziness' and 'correctness' decay over time. By capturing model responses to a standard suite of complex refactoring tasks, developers can quantify performance regressions following model updates.

Suggested repo: AgentWatch

"Stop wondering if your AI is getting lazier—measure it with automated regression suites."

Estimated effort: 40h