Figma Had Its Best Config Ever. Your Client Already Shipped Without You.

Config ended Wednesday. The consensus formed before the hotel bar closed: Figma Motion (native animation timelines), Code Layers (GitHub on canvas), WebGPU shaders, an AI design agent wired into Notion, Slack, and Atlassian, and Weave tools for generative plugins. Dylan Field on stage saying AI lowers the floor but designers will raise the ceiling. Eight thousand people in Moscone Center South, nodding. The discourse: AI and design finally converged. Figma is the collaboration surface where it all comes together. Designers are essential.

I'm going to argue that the consensus is looking at the wrong match.

The Consensus Is Seductive. Here's Why.

Figma actually shipped real things this cycle. Motion is genuinely useful — building animation directly in the design file, without the After Effects handoff or the ProtoPie export, removes a friction point that's been annoying product designers since 2019. Code Layers is interesting for the 20% of designers close enough to the codebase to care. The agent's skills — packaged workflows the agent can run on command — are in early access and they work. Config was not a vaporware conference. These are real features solving real problems for real teams.

Dylan Field's 'designers raise the ceiling' framing is also smart positioning. He's right that the canvas remains the surface where taste, systems thinking, and judgment compound. When product designer Edward Chechique says he's 'gone from zero possibilities to great possibilities' using the agent to build plugins for dashboard layouts and data visualization, that's not marketing copy. That's a skilled designer getting faster.

The problem isn't the features. The problem is what the consensus decided the features mean for the industry.

The Evidence the Consensus Is Ignoring

Two months before Config, Anthropic shipped Claude Design. Not a research preview, not a blog post about future capabilities — a product, available now, for Claude Pro subscribers. It builds production-ready interactive prototypes from natural language prompts. It reads your codebase and design files at onboarding and auto-applies your color system, typography, and components. The use case Anthropic pitched to TechCrunch at launch: 'founders and PMs without a design background' who want to share ideas more easily.

Read that again. Not 'teams who want to move faster.' People who don't have a designer.

OpenAI has a Product Design plugin. Canva's AI generates brand-consistent assets from text. These tools are not competing with Figma for the design team's seat license. They're competing for the moment before the design brief exists — the Monday morning when a PM decides whether to open a chat with Claude or open a Figma file and start a project with the design team.

That's the moment no one in the Config recap is talking about.

What's Actually Happening at the Layer Figma Can't See

The floor didn't just get lower. The door changed.

For small clients, solo founders, and early-stage startups, Claude Design's 15-minute prototype isn't 'good enough.' It's better than what most of them were getting before — a week-long back-and-forth to define a brief, followed by a wireframe they couldn't evaluate because they didn't have the visual vocabulary to react to it. Now they have an interactive thing they can click through on Day 1. Some of them will never call a designer.

The mid-market is next. Not this quarter. But the tools mature fast. The real question isn't whether Figma Motion is better than ProtoPie (it is). The real question is whether the client signs a contract with the design studio before they open Claude Design and start clicking.

Figma's answer to that question is essentially: 'once the client is in the room with the designer, the canvas is the best collaboration surface.' That's true. The problem is the room is harder to get into.

The studios that figure this out first won't be the ones with the fastest Figma agents. They'll be the ones who repositioned themselves to be useful before the brief — the people who help the PM think through the problem before they start generating screens. The deliverable shifts: from pixels to framing. From files to the first question.

The Prediction

Studios that charge for brief-definition, design system architecture, and editorial direction — not hours in Figma — will hold their margins through 2027. Studios that charge for execution will not.

Confidence: Medium. The directional pressure is clear. Disintermediation at the brief layer is already happening for small clients — founders and indie PMs are building with Claude Design, Canva AI, and OpenAI's plugin right now. Whether this pressure climbs to established product orgs depends on how fast these tools hold up under complex, multi-state UI. Today they don't. That ceiling is moving. Timeframe: Now for solo/small-team clients. 6 months for early-stage SaaS product teams with a budget and no full-time designer. 1 year+ for mid-market companies with existing design processes in place.

What would prove me wrong: if Figma's Code Layers ships at scale by Q3 and product teams adopt the canvas-as-codebase workflow for both design and engineering. At that point, the client's PM is in Figma anyway to collaborate with their engineers — skipping the designer becomes structurally hard. Code Layers as a moat is actually a plausible bet. It's just not the story the post-Config consensus is telling.

The Config headline was 'AI and design finally converged.' That's true, and it matters, and Figma shipped real things. But the more accurate headline for this week is: the competition for your client's project isn't Sketch. It's the chat they opened before they called you.