Independent engineer in Taiwan, building AI engineering infrastructure that survives N agents mutating one repo at the same time.
- AI-Atomic-Framework — A deterministic governance brake for multi-agent coding. Gives every change a 5D CID identity, runs it through a 7-stage lifecycle, and seals each task with a replayable SHA256 evidence packet. Atom maps retire legacy modules through a gated
draft → shadow → canary → active → legacy-retiredreplacement lifecycle. - 3KLife — An AI-driven Three Kingdoms game project. My day-job test bed for ATM: five coding agents working a single repo every day.
When N agents concurrently mutate a single repo, how to leverage the Atomic Foundry and 5D CID to forge a rock-solid system governance foundation. Seven-stage lifecycle, SHA256 evidence chain, INV-ATM red lines, atom maps for legacy retirement.
AI Learning Notes — bilingual engineering notes (EN + 繁中) on multi-agent coordination, Coding Agent governance, prompt engineering, harness engineering, and shipping production AI systems.
If anything I build or write resonates, I'd love to hear from you — start a Discussion or open an issue on the relevant repo.