Config 2026: code just became a design material
Figma's biggest release in years puts code layers, motion, and shaders on the canvas. What it changes, and what it does not.
At Config 2026 in June, Figma shipped its most significant release since Auto Layout, and the most telling part was not a feature. It was the idea the whole keynote was built around: code is not the opposite of design, it is a material for design.
That framing does a lot of work, because it names the shift exactly. For a decade, design tools treated code as the thing that happens after design: downstream, in someone else's editor, behind a handoff. Config 2026 quietly removed that boundary.
What actually shipped
The headline is code layers: working code on the design canvas, created from a frame, generated by the agent, or pulled from a real repository. You can explore interaction directions side by side as running code, pull a code layer back into editable design layers, and push edits back with a click. Design to code and code to design, in the same file, in both directions.
Around it, a set of new materials: a real motion timeline with keyframes and reusable animation systems, so motion design no longer means leaving the tool. Agent-built shaders with parameters exposed as canvas controls. 3D transforms. Generative plugins you describe instead of program. And the Figma agent itself grew into something with skills, which are packaged workflow instructions, and connectors into Notion, GitHub, Slack, and the rest of the stack a product team already lives in.
On its own, each is a feature. Together, they are an argument: the canvas is becoming the shared workspace where design, motion, and implementation stop being separate rooms.
The keynote where these launches were announced. Source: Figma on YouTube.
What it changes for working designers
Three practical shifts stand out to me.
First, the quality ceiling on exploration just rose. When trying something as real code costs one prompt instead of one sprint, more decisions get made against working behaviour instead of static pictures. That is healthy, because static pictures lie about interaction in ways running code cannot.
Second, systems thinking got more valuable, not less. Every one of these materials, whether motion presets, parameterised shaders, or agent skills, is a reusable building block. The designers who benefit most will be the ones who already think in reusable decisions rather than one-off screens. A motion system on a timeline is leverage. Forty custom animations is a maintenance bill.
Third, the handoff line moved again. When a designer can hand a developer a code layer instead of a marked-up frame, the conversation changes from build this to review this. That is a different skill for both sides, and teams that redraw the boundary on purpose will outrun the ones who let it happen to them.
What it does not change
The honest caveats. Most of this is in beta, gated by plan and seat type, and rolling out gradually, so the demo and your Tuesday are not yet the same thing. AI credits will eventually meter the agent, which means workflow costs need watching. And no amount of canvas machinery answers the questions that make design hard: what to build, for whom, and what to leave out. Config made expression cheaper. Judgment costs exactly what it did in May.
Takeaways
Learn the motion timeline now. It is the most solid of the new materials and it is immediately portable. Treat agent skills as a design-system problem, and encode your team's conventions once, on purpose. And if you have built your seniority on being the person who bridges design and code, notice that the bridge is becoming a place everyone can stand. That means the thing that sets you apart moves, again, to the quality of the decisions you bring there.