Agents on the Canvas in tldraw
summary
Tldraw reports a one‑month experiment, Fairydraw (December 2025), that places three fairies as agents directly on the infinite canvas; agents elect a leader, build a to‑do list and delegate tasks, and sometimes converge on a working diagram. Engineers prototype an electron wrapper exposing an HTTP endpoint so agents can run JS against a local canvas, enabling interactivity but also enabling risky script injection; demos succeed intermittently and reveal surprising capabilities.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 1:41Early AI canvas experiments
- 4:53Single-prompt AI agents
- 7:41Multi-step agent loops
- 9:49Agents on the canvas
- 13:41Fairies demo and orchestration
- 14:22Desktop app and sandboxed execution
- 16:28UI generation and script injection
- 18:12Local-first and agent empowerment
- 19:13Closing
