Anthropic shakes up its leadership to bet on an internal experimental lab

When an Instagram co-founder voluntarily gives up a C-level title at one of the hottest AI companies on the planet to become a plain "member of technical staff," that isn't a demotion. It's a signal that the company has found something it considers more important than the corporate hierarchy.

Mike Krieger, who joined Anthropic two years ago as chief product officer, is leaving that post. He is not leaving the company. Instead, he becomes co-technical lead of Anthropic Labs, the internal incubator the company has decided to scale up sharply. He will lead the team alongside Ben Mann, who previously ran product engineering. Ami Vora, currently "head of product," will take over the vacated role of leading the product organization. Krieger will report directly to Anthropic president Daniela Amodei, which shows how much strategic weight the new assignment carries.

The scale of this decision needs context. Anthropic Labs started in mid-2024 as a tiny initiative: literally two people working on experimental projects at the intersection of research and product. A year and a half later, the company's leadership decided those experiments deserve one of Silicon Valley's most experienced product leaders. Krieger isn't just a manager. Together with Kevin Systrom, he built Instagram from zero to a billion users. Moving him to the lab isn't exile to a quiet backwater; it concentrates one of the company's most valuable resources on an area Anthropic considers strategic.

Labs focuses on "experimental products." That vague label hides a fairly concrete logic. Until now, Anthropic has made most of its money from the Claude API and its consumer chat interface. But the large language model market is commoditizing fast: OpenAI, Google, Meta, xAI, and dozens of startups offer models of comparable quality, and competing purely on the underlying technology keeps getting harder. The internal incubator is an attempt to find new product forms for AI that go beyond the familiar chatbot and API access. In essence, Anthropic wants to figure out what next-generation AI applications will look like and build them before its competitors do.

The strategy isn't unique, but the execution is unusual. Google has long experimented with AI products through DeepMind and internal labs, but it does so inside a giant corporate structure where innovation often drowns in bureaucracy. OpenAI, by contrast, ships new products aggressively, but from a single product hub. Anthropic is taking a third path: it carves the experimental team out into a separate unit with direct access to leadership, but keeps it inside a relatively compact organization. That lets it keep startup speed while freeing the experimenters from pressure to hit current product metrics.

There's also a personnel angle that can't be ignored. C-suite reshuffles always raise questions about internal politics. Still, the fact that Krieger stays at the company and gets a direct line to Daniela Amodei points more to a deliberate strategic move than to a conflict. Ami Vora, who is taking over product leadership, is an experienced leader, and handing her operational control of mature products looks like a sensible division of labor: one person owns what makes money today, the other owns what will make money tomorrow.

For the industry, Anthropic's move is further evidence that the large language model race is entering a new phase. Raw model power is no longer the main differentiator. The winners will be those who find the right product wrappers for AI: interfaces, use cases, and integrations users haven't even imagined yet. Creating a dedicated lab with top-caliber leadership is a bet that AI's future will be decided not by benchmarks but by product breakthroughs.

Anthropic isn't disclosing which experiments Anthropic Labs is running. But the fact that it is reshaping its top leadership for this effort says more than any press release could. Anthropic has clearly spotted something on the horizon, and decided it needs someone who has already changed once how hundreds of millions of people interact with technology.

*Meta is designated an extremist organization and banned in Russia.