10 agent evals for AI engineers:(explained with usage)1) golden set→ a fixed set of cases you never edit, run on every single change.→ use as the baseline that tells you whether anything moved at all.2) llm as judge→ a second model scores the output against a written rubric.→ use when the answer is open-ended and there is no string to match against.3) rubric scoring→ one number per dimension: correctness, tone, safety, cost.→ use when a single score hides which part actually got worse.4) traject

your AI agent is slowly poisoning itself with its own memorynothing crashes. the agent keeps answering like everything is fine.it just starts treating old preferences, dead requirements and rewritten summaries as current facts.that's why this article matters.everyone is racing to give agents permanent memory. almost nobody is asking who decides what gets saved, whether it is still true and when it should disappear.every memory needs a source, a reason to stay and a rule for when it dies.without