Members-Only
Recent Talks & Demos are for members only
You must be an AI Tinkerers active member to view these talks and demos.
Agent HQ: Gamifying AI Agents
See a live demo of Agent HQ, a game where AI coding agents work on real tasks in sandboxed workspaces. Learn how to manage stateful, crash-safe agents and optimize performance.
Agent HQ is a retro top-down office game where you hire real AI coding agents each employee at a desk is a live Claude Agent SDK or OpenAI Codex SDK session that actually reads, writes, and runs code in its own sandboxed workspace folder while you watch from a game world.
In the demo I’ll run the working system live: hire an agent in the office, give it a name and model, type a real task, and watch it walk to its desk and start working. Speech bubbles and the office feed stream its actual tool calls, assistant text, and results in real time. I’ll show the architecture (Phaser client ↔ WebSocket server ↔ provider runners), the per-agent workspace folders on disk, the persisted save file and JSON session transcripts, and the wire protocol that drives it all. Repo and logs will be on screen