Ep 108: Agents on Canvas

Davy and PJ try Figma’s on-canvas agent and compared to a quick gut check with Paper. and get into what these tools still need to do to fit into our daily work.

Davy and PJ dig into the messy state of software quality in 2026… from visual regression tests that look wrong to the growing need for shared skills, layered checks, and actual human judgment when everyone is shipping faster than they can review.

Are designers working in code, or directing agents that write it for them? Davy and PJ dig into the shifting role of Figma, the skills designers need to guide AI-assisted work, and why changing one color token can turn into an afternoon of CI, docs, and downstream consequences.

We discuss, you guess it, whatever ai-native means in the world of design systems, including that start of a new phase of PJ and Davy's design system spec that originated from this episode!

Davy and PJ debate how automation, visible chaos, and human judgment can help teams move fast without turning the design system into an expensive junk drawer.

Davy returns from Config with a backpack full of new Figma features. Also our annual discussion of were these releases for us?

From the comical wizard prompts of 2024 to the heavy-lifting reality of Figma MCP, Davy and PJ chart the compressed timeline of design system and canvas automation.

Why do we build in public when corporate nature tells us to hoard our work? Davy and PJ break down the unglamorous reality of sharing open-source tools, micro-contributions, and why 90s music is a hill worth dying on.

We hit the Centennial! In our 100th episode, Davy and PJ look back on 4 years of our design systems chatter, before diving deep into the future of design systems: a community-driven, machine-readable schema that moves us past the "Figma-as-Source-of-Truth" era.

Davy talks about trading a week of manual Figma updates for a "code-to-canvas" workflow that gets us "close enough" while PJ talks about running multiple worker-based agents to do production and qa work in tandem. Are these burner tasks?