Full-stack build·2026
Agentic regression harness
Playwright suite driven by an agent that triages its own failures and files reproducible bug reports.
TypeScriptPlaywrightNode
I build products end to end and keep them honest with agentic tests. Currently focused on where the two meet: AI augmented development with a QA spine.
Two builds I would show first.
Playwright suite driven by an agent that triages its own failures and files reproducible bug reports.
Typed-route blog and news front end with dark mode and CSS-variable theming.
Notes on testing, agents and shipping alone.
Six months of running an agent against a Playwright suite — what broke, what held, and the guardrails I keep.
Flakiness is rarely random. It is usually a race you have not named yet.
Prompting is scaffolding. The interesting part is what you refuse to accept back.
One email when something ships or something breaks interestingly. Nothing else.