https://marioruiz.net/blog/ Commands, old machines, and experiments. Agents and APIs, and the history and language around the tools. en Fri, 30 Oct 2026 12:00:00 GMT https://marioruiz.net/blog/agent-reads-the-docs/ https://marioruiz.net/blog/agent-reads-the-docs/ Fri, 30 Oct 2026 12:00:00 GMT Every chat starts by reading the same file. If that file is long and stale, every answer starts a little wrong. https://marioruiz.net/blog/tests-are-the-spec/ https://marioruiz.net/blog/tests-are-the-spec/ Fri, 23 Oct 2026 12:00:00 GMT A ticket in prose is advice. A failing test is an order. When an agent writes the code, only the second one holds. https://marioruiz.net/blog/code-and-poetry/ https://marioruiz.net/blog/code-and-poetry/ Tue, 20 Oct 2026 12:00:00 GMT Writing poetry taught me that every line has to earn its place. Code is the same craft, and with AI writing the words for us, we are quietly letting it go. https://marioruiz.net/blog/generated-tests-are-a-diff/ https://marioruiz.net/blog/generated-tests-are-a-diff/ Tue, 13 Oct 2026 12:00:00 GMT Which line in that diff is worth reading, and why the agent must not be the one who updates the suite. https://marioruiz.net/blog/boards-from-childhood/ https://marioruiz.net/blog/boards-from-childhood/ Sun, 05 Oct 2026 12:00:00 GMT Childhood boards, a Nokia mancala, and the experiments I still make for fun. Online rooms with voice and camera for the boards that need a link. https://marioruiz.net/blog/one-payload-three-clients/ https://marioruiz.net/blog/one-payload-three-clients/ Tue, 30 Sep 2026 12:00:00 GMT Slack, the web, and the IDE share one JSON payload for health checks. If an agent paraphrases it, a failed check becomes mostly fine.