Blog
Engineering in the new paradigm.
- (1)Taking agent workflows apartComparing ArDD, Superpowers, and Spec Kit led me to a more useful question: what does each module do, how does it take effect, and who makes sure it runs?
- (2)The bugs I could only find by running the thingA follow-up to the discipline-tool post. I ran ArDD on real work and it broke in ways I couldn't have reasoned about — worktree state committed in the wrong place, a git config flipped out from under me, and the redesign that made a dead agent a no-op instead of a mess.
- (3)I'm bad at discipline, so I built a tool that does itCoding agents imitate a healthy codebase well but bring no judgment to an empty one — so I built a spec-driven framework to supply the discipline I won't. A postmortem, a pivot, a retrofit test, and whether you actually need any of it.