Steve Ruiz, creator of tldraw, explains how he turned a free, open-source whiteboard into an SDK business built on a high-performance web canvas. The product lives at tldraw.dev and is already being used to build internal tools, collaborative interfaces, and now agent-facing canvases. This is not a pivot story. It is a deliberate architecture decision made before the AI wave arrived.
The most useful part of this conversation is Ruiz's breakdown of how SDK and infrastructure companies absorb the pressure of agentic software differently than SaaS companies do. When an LLM needs a surface to think on, an infinite canvas becomes infrastructure, not a feature. Ruiz has already shipped an agent starter kit at tldraw.dev and a demo called tldraw fairies that shows what happens when you hand a model an unbounded drawing surface.
The episode runs with a 15-minute bonus segment for Changelog++ members. Read the show notes for direct links to the agent kit and the fairies demo. The real reason to listen is the internal tooling discussion: Ruiz has a specific thesis about where agentic software creates demand for custom, embedded canvas experiences, and he names the constraints that will shape that market.
[READ ORIGINAL →]