How I work
I build software with AI agents, and I keep it honest with tests. That is the whole method, and every project on this site was built that way.
The tools move fast. The discipline is what makes the speed safe. What follows is how I actually work: the loop I run, and the habits that keep agent-built software durable.
The RECL loop
Requirements, Execute, Check, Loop. It is the cycle I run on every project, and the reason agent speed does not turn into agent chaos.
Requirements
I write down what the thing has to do, in plain language, before any code. A short PRD, an AGENTS file, a task list. The clearer the requirements, the less the agent has to guess.
Execute
I let the agent build against those requirements, and I stay in the loop. I read the diff and steer rather than accept whatever comes back.
Check
Every change is checked against a test. Pytest for the Python projects and Vitest for the TypeScript ones, with cached fixtures and recorded payloads so the checks run in CI without a live service. If it is not tested, it is not done.
Loop
Then I loop. Refine the requirements, run it again, tighten the tests, until the result holds. The loop is short on purpose.
What keeps it durable
Verification first
Speed without verification is just a faster way to ship bugs. Tests are what let me move quickly and still trust the result. Every project here is guarded by them, from unit tests on the logic to CI that runs without a live service.
Requirements before code
Most of the leverage is upstream of the agent. A written PRD, an AGENTS file, and a task list turn a vague ask into something an agent can execute and a test can check.
The pairing that lasts
Agent fluency on its own is common now. What makes it durable is pairing it with ordinary engineering discipline: test-driven development, CI/CD, structured logging, version control. I was working this way well before I started using AI, and that engineering practice is what I wrote up in a 100-page manual for fellow analysts at work. AI adoption is what I advocate for there. The projects on this site I built with agents, on my own time.