https://joelstevick.com/blog Engineering write-ups on building with AI. en https://joelstevick.com/blog/what-the-golden-set-found https://joelstevick.com/blog/what-the-golden-set-found Wed, 11 Nov 2026 12:00:00 GMT Thirty-two questions, three runs and under two dollars. The book's index had never built, code questions never got their code, and my scorer was older than my rules. Evals Ask the book https://joelstevick.com/blog/first-sentence https://joelstevick.com/blog/first-sentence Fri, 30 Oct 2026 12:00:00 GMT The demo checks every answer against its sources, so I taught it to type one checked sentence at a time. The eval set caught a version losing 8 of 26 answers. Tuning the demo https://joelstevick.com/blog/catching-up https://joelstevick.com/blog/catching-up Mon, 26 Oct 2026 12:00:00 GMT Every five minutes my site works out what is due and does each thing it can, as me. It tells me only what needs my hands, one line with one button. Running the shop https://joelstevick.com/blog/finding-prospects https://joelstevick.com/blog/finding-prospects Sun, 25 Oct 2026 12:00:00 GMT One button on my admin page searches the web for companies that may need my services, checks each find in code, and learns from earlier finds. Finding clients https://joelstevick.com/blog/widening https://joelstevick.com/blog/widening Wed, 21 Oct 2026 12:00:00 GMT Adding each passage's neighbours answered all 5 two-passage questions, up from 4, at 60% more a question, over my 50% limit. I switched it on anyway. Tuning the demo https://joelstevick.com/blog/tuning-with-evals https://joelstevick.com/blog/tuning-with-evals Wed, 21 Oct 2026 12:00:00 GMT My demo chatbot said my résumé listed no jobs. How I fixed it test-first, then tuned the pipeline with a 25-question eval on live models, for $1.96. Tuning the demo