AI Experiments
Apps I've been building and designing solo with AI lately.
How I build with AI
I'm using Cursor, Codex, and Claude Code depending on what each does best at the time. I find that one tool tends to pull ahead for a while, then others catch up, and eventually they converge and feel mostly the same. For consistency across agents, I keep a roadmap and handoff docs inside the repo for agents to follow. I find that without these, agents revisit stuff we've already covered or go off in directions I didn't ask for. I also use skills (e.g., frontend-design and web-design-guidelines for UI work), my own custom rules, and MCPs (e.g., Playwright for testing). Besides actual coding, I treat AI as a junior design partner, using it for research, brainstorming, testing directions, and consultations.