Next.js Hacker News
Even (very) noisy LLM evaluators are useful for improving AI agents
13 points by GabrielBianconi 2 days ago | discuss