Anthropic ships a design tool, Figma's CPO resigns from Figma's board

Two things happened on or around 17 April, and I want to note that they happened on the same day.

The first is that Anthropic launched Claude Design, a prompt-to-prototype product built on Claude Opus 4.7, which The Information described as generating "prototypes, slides, and marketing assets without requiring design skills."¹ The launch page pitches it as a design surface for people who don't have designers, which is a polite way of saying it's a design surface for people who would otherwise have bought Figma seats, or Adobe seats, or hired a junior.

The second is that Mike Krieger, Anthropic's Chief Product Officer, resigned from Figma's board on the same day the launch was reported.² Krieger joined Figma's board in 2022, before he joined Anthropic in 2024. The overlap had been awkward for roughly eighteen months and was presumably managed by the usual structures, recusal from competitive discussions, walled-off product roadmaps, the sorts of things that work until they don't. On the day the product shipped, they didn't.

Figma closed down ~6–7% on the news. Adobe closed down ~1.5%. That spread is itself the story.

What the product actually is

I spent some time with the Claude Design launch materials and the product itself. It is not, in the narrow sense, a Figma competitor. Figma's core product is a multiplayer vector editor with a component system, auto-layout, and a developer handoff story that took a decade to build. Claude Design does not ship any of that. It generates static and interactive prototypes from prompts, iterates on them in natural language, and exports to code or to image formats. The structural comparison is not Figma-the-editor; it is Figma-the-entry-point, the early-stage ideation and wireframing workflow that FigJam and, increasingly, Figma Make were built to own.

Figma's core product is a multiplayer vector editor with a component system, auto-layout, and a developer handoff story that took a decade to build.

This matters because Figma's seat-expansion story depended on pulling non-designers into the tool. Product managers with view-only seats upgrading to edit seats. Marketers starting in FigJam and graduating to Figma. The whole net-dollar-retention thesis rested on the non-designer user who learned to use a designer's tool because that's where the work lived.

Claude Design is a bet that this user never needed to learn the designer's tool. They needed the output.

The frame: SaaS apocalypse, specifically the entry-point version

The SaaS apocalypse frame predicts per-seat pricing compression as agents replace human users. The usual worked example is Salesforce seats replaced by agentic sales workflows. Claude Design is a cleaner case than that, because it doesn't even require the fiction of an agent sitting in a seat. It removes the seat. The non-designer PM who would have cost Figma $15/month on a Professional seat now pays Anthropic, or more precisely, pays Anthropic's Max plan, where Claude Design is bundled, for the prompt-driven output directly.

The frame's sharp prediction is: seat counts at tools that monetised the non-expert user should compress first and fastest. Figma's disclosed seat mix has always been heavily weighted to non-designers, the last number I saw publicly was roughly two-thirds non-designer seats in paid accounts.³ That ratio is precisely the exposure.

The frame also predicts something about pricing model migration: tools that can't hold the non-expert seat will pivot toward consumption or outcome-based pricing at the designer tier, trying to extract more from the remaining expert user. Watch for this at Figma's next pricing update.

Why Adobe fell less

The 4–5 percentage point gap between Figma's drop and Adobe's drop is the interesting number. Adobe has more exposure in dollar terms, Creative Cloud is a vastly larger seat base than Figma's, but the market discounted that exposure. I read this three ways, and I think all three are partially right.

First, Adobe's expert users are stickier. Photoshop and Illustrator workflows have twenty years of muscle memory, plugin ecosystems, and file-format lock-in behind them. The professional designer is not the user Claude Design is coming for, at least not in this version.

Second, Adobe has been shipping its own generative tooling (Firefly, the generative features inside the CC apps) for long enough that the market has already partially priced the seat-compression risk. Figma's generative story is younger and less integrated, and Figma Make, the closest analogue to Claude Design, launched only months ago and is itself a tacit concession that the prompt-to-prototype workflow is where the entry-point user is going.

Third, and this is the one I'd watch: Adobe's B2B enterprise motion runs through procurement relationships that Anthropic does not yet have. Figma's bottoms-up land-and-expand motion is exactly the motion that a horizontally-distributed model vendor can disintermediate. Adobe's enterprise sales motion is harder to disintermediate because the buyer is not the user.

What this is a case of

This is a case of the model layer moving up into application categories where the application's moat was workflow rather than capability. Figma's moat was workflow, the collaborative editing surface, the component libraries, the handoff. Claude Design doesn't attack the workflow; it obviates the need for the workflow by collapsing ideation-to-output into a prompt.

It is also a case of the frontier lab as a horizontal product company, which is the question that's been open since OpenAI shipped Canvas and Operator. Anthropic has been notably more disciplined than OpenAI about not shipping horizontal app surfaces, the Claude-is-a-platform-not-a-product posture has been explicit in their enterprise sell. Claude Design is a meaningful break from that posture. It is a first-party application, in a specific vertical, competing directly with named public companies.

The Krieger resignation on the same day is the tell. You do not resign from the board of a public company on the day your employer ships a product in that company's category unless both sides understood, some time ago, that this day was coming. I'd assume Figma's board had known the roadmap for months and had been managing the optics rather than the substance.

What to watch

Three things.

One: Figma's next pricing update and seat-mix disclosure. The frame predicts non-designer seat compression. Either the numbers show it or they don't.

Two: whether Anthropic ships a second first-party vertical app in the next two quarters. Claude Design alone could be a one-off experiment. A second one, Claude Research, Claude Legal, Claude Finance, something, would confirm a strategic pivot that the enterprise narrative has been denying.

Three: the Max plan attach rate. Claude Design is a feature inside a $100+/month consumer tier. If Anthropic starts disclosing Max subscribers as a standalone number, that's because the number has become load-bearing for the revenue story, which tells you something about how the inference-economics math is shaking out at the frontier.

The structural question underneath all of this is whether the design category is the canary or the exception. I don't know yet. The frame says canary. The evidence so far is consistent with the frame.

Footnotes

¶