Mike Krieger, the Anthropic chief product officer and Instagram co-founder, resigned from Figma's board on April 14 after joining less than a year ago. TechCrunch reports the resignation comes because Anthropic's Opus 4.7 release, which shipped this morning, includes design tools aimed at the same market Figma serves. Krieger did not release a statement. Figma's stock rose 5 percent on the news, which is market signal that investors believe Figma can survive the direct competition but were worried about the insider-conflict optics of the board position. The move is the clearest signal yet that AI labs intend to ship vertical products rather than just sell API access, and it is the first time Anthropic has publicly moved to compete with one of its larger customer integrations.

I covered Opus 4.7 earlier today and noted the "new website-design tool" as a secondary product in that release. The Krieger resignation reframes the small mention as deliberate strategic positioning rather than a one-off add-on. Anthropic has the raw capability to build a competitive design product: the model can generate images, code, and structured output; the Artifacts surface in Claude gives them a canvas; and the commercial pressure to build revenue above the API line is already visible in their pricing and enterprise positioning. The missing pieces are domain expertise (Figma has two decades of workflow depth, team-collaboration features, file formats, and plugin ecosystem), designer trust, and the specific UI affordances that separate a demo from a daily tool. Whether Anthropic is seriously attempting to replace Figma or carving off a casual-user segment is the operational question the reporting does not yet answer.

This is the SAASpocalypse thesis arriving in concrete form. The thesis says that foundation-model labs with ultra-capable models can move up the stack and disintermediate the SaaS companies that currently wrap models into workflow products, because the labs own the core capability and can price predatorily against any layer of value-add. The counter-thesis says established software brands have domain depth, customer relationships, regulatory work, and integration ecosystems that foundation-model labs cannot replicate in a reasonable timeframe. Both theses are partially right. What Anthropic shipping a Figma-adjacent product establishes is that the first-order SAASpocalypse thesis is being tested in production against a specific competitor in a specific vertical. Canva AI 2.0, Adobe Firefly, and now Anthropic direct-to-design are compressing the same market simultaneously. The winner is going to be decided by which one executes on workflow rather than which one has the best underlying model, and Anthropic is starting from behind on workflow.

For anyone building on Anthropic's API in a vertical that an AI lab could plausibly compete with, the strategic question got sharper today. If you are a SaaS company layering Claude on top of a workflow that does not have deep proprietary data, strong customer switching costs, or regulatory moats, assume the lab can ship a direct competitor within twelve to eighteen months. That does not mean you should stop building; it means the moat needs to be something other than the model itself. If you are a builder of vertical applications in markets where the labs have not yet moved (healthcare-specific workflows, legal discovery, industry-specific compliance, regulated finance), you now have a clearer signal that the SAASpocalypse move starts in markets with low regulatory drag and high consumer visibility (design, productivity, general coding). Regulated verticals get disintermediated later, not never. For Figma specifically, the test is whether two decades of workflow depth survives the distribution advantage of Anthropic bundling design tools into Claude. My read is Figma survives but takes a 20 to 40 percent margin hit over the next two years as casual-design volume migrates toward whoever ships the lowest-friction model-native experience. That is the cost of sitting between customers and the model provider when the model provider decides to go direct.