How I Work

Building with AI agents

I use AI coding agents — Claude Code, mainly — as my primary way of writing software, not as an occasional assist. Most of the code in my shipped apps was written this way: I describe what I want, review what comes back, and iterate.

What that looks like day to day

• I stay in the driver's seat. The agent writes code; I set direction, review changes, and decide what ships.
• Small, reviewable steps. I work in scoped changes I can actually read and verify, not one giant AI-generated dump.
• Ship, then iterate. I'd rather get a real version in front of users and improve it than polish something no one has used yet.
• Quality still comes first. Using an AI agent doesn't lower the bar — if anything, it means more time to focus on the details that matter: UX, privacy, and reliability.

Why I share this

This is the workflow I teach through DanielsApps. It's not theoretical — it's how Horizon Drop, BookChat AI, and the rest of my apps actually got built.