Builder notes on the AI tools I run on real deals.https://hermanchanai.com/https://hermanchanai.com/posts/every-health-check-was-green/https://hermanchanai.com/posts/every-health-check-was-green/Model evals don't catch a broken data layer. The fix is the same discipline pointed at the database — checks on the numbers people actually act on.Wed, 10 Jun 2026 00:00:00 GMThttps://hermanchanai.com/posts/the-feature-that-tested-well/https://hermanchanai.com/posts/the-feature-that-tested-well/How the eval caught a leaked metric — and then told me how to make the feature actually useful.Mon, 08 Jun 2026 00:00:00 GMThttps://hermanchanai.com/posts/cost-of-being-wrong/https://hermanchanai.com/posts/cost-of-being-wrong/Not all classification errors cost the same. The eval reports two numbers and ranks every failure by business impact.Fri, 05 Jun 2026 00:00:00 GMThttps://hermanchanai.com/posts/grading-an-llm-against-its-own-guesses/https://hermanchanai.com/posts/grading-an-llm-against-its-own-guesses/The fastest way to build a useless eval is to use the model's own output as the answer key.Tue, 02 Jun 2026 00:00:00 GMThttps://hermanchanai.com/posts/one-place-the-llm-lives/https://hermanchanai.com/posts/one-place-the-llm-lives/Dozens of Python tools, exactly one LLM call per document. The boundary is the design decision.Mon, 25 May 2026 00:00:00 GMThttps://hermanchanai.com/posts/rent-comps-t12-skills/https://hermanchanai.com/posts/rent-comps-t12-skills/Three problems running spreadsheet workflows in Claude chat that pushed me to build dedicated skills.Fri, 22 May 2026 00:00:00 GMThttps://hermanchanai.com/posts/what-this-site-is/https://hermanchanai.com/posts/what-this-site-is/An introduction to the site — who I am, what I build, and what to expect here.Sun, 17 May 2026 00:00:00 GMT