Systems

Most codebases are structured for human developers navigating with IDEs. AI agents don't use IDEs — they use tool calls. Every file read costs time. Every cross-file edit risks failure. Here's how I'm restructuring a 33,000-line codebase to be maintained by AI.

CREATE TASKS — plan items where the agent's job isn't to do work, but to read the world, decide what work should exist, and alter the plan. The plan becomes a program that generates more of itself.

browser-tool.ts — a thin Playwright CLI that turns an AI agent into a web researcher. No scraping framework, no complex automation. Just launch, screenshot, decide, act.

A living design system built on Atomic Design Methodology — where the AI agent works backward from pattern specs, composing primitives and tokens it already knows. The methodology finally makes sense when the maintainer never gets tired.

AGENTS.md and CAPABILITY-MAP.md — two files that make AI agents honest about their limits, and resourceful about overcoming them.

A shell loop that runs an AI agent repeatedly against a plan — each iteration gets a fresh context window, does one thing, verifies the build, and exits. No human needed.

The operating principle behind everything I build — and why most engineers are doing it backwards.

SOUL.md, HUMAN.md, and TENSIONS.md — three files that turn a generic AI into a thinking partner that actually knows you.

Four files that replace standups, sprint planning, and handoff meetings when building with AI.