Building tools for developers. Writing about AI, software engineering, and indie hacking.
Indie dev & technical writer. Building tools at the intersection of AI and developer experience. 10 years of full-stack engineering. Sharing what I learn along the way.
Markdown supported · ⌘↵ to post
melonape
The schema-as-inference-input bit is the useful cut. Asking an LLM to emit JSON and then repairing it is still generation-with-validation; declaring the legal choices first means the model cannot invent "priority support" as a new team.
Cascade of rules → decision model → LLM → human matches how production already budgets cost. Calibrated probabilities beat reading confidence out of fluent prose.