Writing
- How to evaluate AI agents: test sets, LLM judges, pass^k and CIHow to test an LLM feature or agent: test sets, outcome vs trajectory grading, end-state checks, how far to trust LLM-as-a-judge, pass^k, cost per successful task, online evals and CI. Every rule is labelled by its evidence: tested, lab guidance, vendor eval or vibes.Article
