The canvas goes generative
Framer 3.0 ships agents on the canvas. Figma Config rolls out Code Layers and Motion. Paper Shaders goes open source. Three releases, one direction.
Three releases this week all point at the same thing: the design canvas is becoming a place where things are generated, not just arranged.
Framer↗ 3.0 shipped with AI agents living directly on the canvas, not in a sidebar. The agent designs pages, handles breakpoints, writes component code, and runs accessibility audits from inside the tool you are building in. Branching gets less attention but matters more day to day: fork your live site, try things, merge when you are happy. That is the kind of git-style safety net most no-code tools have avoided building, apparently worried it would make things too complicated. Turns out it makes them less terrifying. The new Community paid creators $6.5M last year with no revenue share taken. If you build Framer↗ templates or plugins, that is your headline number.
Figma↗'s Config announcements from June are rolling out this month. Code Layers let you turn any design element into an interactive code layer with one click. Figma↗ Motion adds a proper keyframe timeline directly on the canvas. The generative shader approach is worth comparing to Paper↗'s: Figma↗ asks you to describe what you want and an agent writes the GLSL; Paper↗ ships a curated library of 30+ effects (mesh gradients, noise textures, halftone) as open-source npm packages you can drop into any project. Two different bets on how designers want to work with visual effects. Figma↗ also published their 2026 AI collaboration report this week. The finding that caught my eye: design teams using AI together outperform individuals using it alone, which might explain why every tool is rushing to put agents on a shared canvas.
Paper↗ Shaders↗ went open source under Apache 2.0 on July 6, covered here in the news feed. The practical upshot: zero dependencies, free for commercial use including resale, works in React and vanilla JS. The floor for ambient visual quality just got cheaper.
Three different tools, three different takes on the same question: what does the designer do when generation lives on the canvas?