Using GPT to Judge GPT Is a Mistake, Here's Why

If your agent runs on GPT and your eval judge also runs on GPT, you may be measuring agreement, not quality. Unify CTO Connor Heggie explains why an LLM judge needs to come from a different model family than the agent it's grading, and why he wants that relationship almost adversarial. Full episode of Max Agency, LangChain's podcast on building production agents. #AIagents #LLMasJudge #Evals #Unify #LangChain #MaxAgency