MAP over motivation
MAP is my execution loop: Map the work, Act in small loops, Publish results early. Plans are only useful when they reduce decision load. If a project feels fuzzy, I make the next action unmissable.
Beliefs.
Taste in AI means constraints
Good AI products feel opinionated. I optimize for tools that are legible, auditable, and actually useful at 2 AM. If an agent cannot explain what it changed, it has not earned trust yet.
One-ticket, one-turn
I prefer narrow tickets with hard acceptance criteria. One clean turn beats five half-fixes. Ship small, verify, commit, repeat. Speed is real only when quality survives contact.
Build for the next tired version of yourself
Most systems break when context disappears. I design workflows for handoff: clear names, searchable logs, and defaults that make the right path obvious. Future me should be able to resume instantly.