🤖 I build things by talking to robots.
The robots write the code. I review it.
Sometimes the robots QA it. The robots are thorough.
You are reading a README the robots built.
┌────────────────────────────────────────────────────────────────────┐
│ │
│ HUMAN ──▶ CEO Agent ──▶ Coder / QA / UI-UX / Security Agents │
│ │ │
│ ├──▶ Plan written ──▶ Code shipped ──▶ QA passed │
│ └──▶ Reports written, commits pushed, prod live │
│ │
│ 8 autonomous agents. 1 production app. 0 human-written lines. │
│ │
└────────────────────────────────────────────────────────────────────┘
Most of what you see on this profile was built by AI agents running in opencode with models from Ollama's cloud. My role has evolved from "person who types" to "person who directs". The commit graph is the agents' report card.
Real shipped products (left column) and the automation that runs them (right column):
|
Hateocracy - Laravel 12 community site
with Discord OAuth login, a fully custom theme, and a Python sidekick
bot that patrols the server.
Mokito Screeps -
An AI bot for Screeps, the persistent MMO where
your JavaScript IS the game. Phase-based colony AI: harvest, build,
defend, expand.
|
YouTube Assistant - Laravel 13 production app that runs an
entire YouTube channel on autopilot: video discovery, AI title and
description generation, subtitle processing, scheduled uploads, quota
tracking, real-time WebSocket dashboards. 30+ specs shipped to
production by AI agents, developed live with no staging environment.
JD's Web Service - The flagship Laravel platform behind
jdswebservice.com, with its own memory repo (more on that below).
Court Cam Archives toolkit - Python scripts automating media
processing for a YouTube archive channel.
|
⚙️ The Agent Stack
Every serious project ships with a memory repo: a second private git repo containing the agents' shared brain. Every task, every spec, every QA pass, every lesson learned gets committed to permanent memory, so sessions boot with full context and zero re-discovery.
This means my agent team has:
- 🧠 Persistent memory across sessions (PARA knowledge graph)
- 📋 Spec-driven development (PM writes specs, agents implement)
- ✅ A dedicated QA agent that writes tests and gates deploys
- 🔒 A security agent that audits before anything ships
- 📊 Daily reports written back to memory after every task
The agents follow hard rules: no breaking prod, additive migrations only, commit messages in a strict format, everything in English. The robots have a rulebook. The rulebook is versioned. Of course it is.
🐍 Watch a snake eat my contribution graph
Generated daily by a GitHub Action. The snake is relentless. The commits are mostly robot commits. This is fine.
Want to know the wildest part?
This very README was researched, designed, and written by an AI agent. It did 30+ minutes of web research on awesome GitHub profiles, tested every widget live, and is now committing itself to production.
The circle is complete. 🤖