I bring manufacturing discipline to software built by AI.
Seven years in electronics manufacturing taught me one thing: quality comes from the process, not from heroes. Now I build software with AI agents and run it the same way. Every change takes the same road, and every rule has a check that holds it.
What I build
yokoten · 横展: stops your AI coding agent from making the mistake you can't undo. One command, no setup. A guard at the bridge (force push, wiped folder, leaked key, release without the owner), a guide for the road (five stages from "safety net" to "a team of agents"), and CI checks that explain every finding. Zero dependencies, MIT. npx yokoten · yokoten.app · under yokoten-labs Built in the open with its own process: 520+ merged pull requests, 1,900+ tests on Linux, Windows and macOS, 52 decision records, a review by a different AI model on every change; a change to a hard line (rules, workflows, release) merges only with my approval. cool-workflow: a saved, cited report from your AI agent. On npm. How I work
Issue → pull request → checks → independent review → merge → release. When a problem shows up twice, I fix its cause, not the next case. (That is 横展, yokoten: one line's lesson spreads to every line.)
Open to remote roles and contract work:
AI delivery and developer-experience setup: guardrails, CI and review flow for AI agents Release and quality engineering for AI teams Technical project management for AI products