Claude Code
My main agent for code, tests, reviews and deploys. Runs in the terminal.
I rarely write code by hand. I direct agents, review their work and ship it. These are the tools behind that.
The agents that write, design and work together.
My main agent for code, tests, reviews and deploys. Runs in the terminal.
Visual work: mockups, images and design specs that Claude then implements.
One agent per area, each reading its own memory first, plus a coordinator that tests them.
Tools that keep decisions, tasks and progress in order.
Repo-native memory for coding agents: decisions, lessons learned and open TODOs live in the repo, so agents pick up where they left off.
A shared board agents can read and update.
Feature branches and pull requests; agents never merge to main on their own.
Where I build, test and run things.
Everything runs in Linux on a Windows machine.
Agents screenshot pages on desktop and phone sizes before I see them.
Small build scripts instead of heavy frameworks.
The core tech I reach for most.
Static, fast pages like this one. No framework needed.
3D in the browser, and 2D games.
Programmatic video.
Where the site goes live.
Simple shared hosting with free SSL.