The Deal
On February 25, 2026, Anthropic announced it had acquired Vercept, a Seattle-based AI startup that had built a cloud computer-use agent. TechCrunch reported it was Anthropic’s second major acquisition in three months, following the December 2025 purchase of Bun, a coding agent engine.
Financial terms were not disclosed. Co-founder Oren Etzioni confirmed to GeekWire that the deal provided a positive return for investors — a measured reassurance given the company’s short runway.
What Vercept Built
Vercept was founded in 2024 by five former Allen Institute for AI (AI2) researchers: Matt Deitke, Kiana Ehsani, Ross Girshick, Luca Weihs, and Oren Etzioni. The team brought deep expertise in machine learning and computer vision — a combination that made them credible candidates to solve one of AI’s harder practical problems: getting models to reliably operate real software.
Their flagship product, Vy, was a cloud-based computer-use agent that could control a remote Apple MacBook via natural language. Users could direct it to navigate applications, manage workflows, and perform multi-step tasks across software that wasn’t built for AI integration. As Forbes described it, Vercept was building AI that could use computers the way people do.
Vercept raised $16 million in a seed round in January 2025 at a $67M post-money valuation, with former Google CEO Eric Schmidt among early backers. Total funding raised exceeded $50 million by the time of the acquisition, per TechCrunch.
Who Joins Anthropic — and Who Didn’t
Three of Vercept’s five co-founders — Kiana Ehsani, Luca Weihs, and Ross Girshick — will join Anthropic as part of the acquisition, according to TechCrunch. Their computer vision and agentic AI backgrounds are directly complementary to Claude’s existing strengths in reasoning.
The other two co-founders are not part of the deal:
- Oren Etzioni — former Executive Director of the Allen Institute for AI, perhaps the most prominent name attached to Vercept — did not join Anthropic. (More on his public comments below.)
- Matt Deitke — took a significantly more lucrative path.
The $250M Meta Defection
Before the acquisition was announced, Matt Deitke had already departed Vercept for Meta’s Superintelligence Lab. The compensation package Meta offered him became one of the more striking data points of the 2026 AI talent war.
Deitke was 24 years old. Meta’s initial offer was $125 million over four years; he turned it down. Mark Zuckerberg then met with him personally and doubled the offer to approximately $250 million, with roughly $100 million payable in the first year. Deitke accepted.
MLQ reported that the deal may be one of the largest employment compensation packages in corporate history. Before Vercept, Deitke had led development of Molmo, an open multimodal AI model at AI2 capable of processing text, images, and audio.
The New York Times described the moment as a signal of how extreme AI talent competition has become. At 24, Deitke’s trajectory — AI2 researcher → startup co-founder → $250M Meta recruit — compressed a decade of career milestones into roughly two years.
The 30-Day Product Shutdown
As part of the acquisition, Anthropic announced Vy would be shut down on March 25, 2026 — giving existing customers a 30-day migration window. This is standard for acqui-hire deals where the acquiring company wants the team and technology but not the customer relationships.
The technology itself will feed into Anthropic’s computer use capabilities, which allow Claude to interact with live applications — clicking, typing, navigating spreadsheets, managing multi-step workflows across tools — instead of just generating text about those tasks.
Etzioni’s “Throwing in the Towel”
The most candid commentary on the deal came from Oren Etzioni himself. He described the acquisition to GeekWire as “throwing in the towel” despite Vercept’s technical traction — a striking admission of the challenges of competing as an independent agent startup against well-funded incumbents.
A public dispute followed on LinkedIn between Etzioni and Seth Bannon, the lead investor, with Etzioni citing the company’s failure to hire business-side talent as a key factor in the early exit. The disagreement offered a rare window into the internal tensions that can end a startup even when the technology works. Silicon Republic covered the dispute in additional detail.
The honest framing from Etzioni is worth noting: it wasn’t a “strategic fit” press release. He called it what it was.
Why This Matters for Anthropic
Computer use is one of the more contested capabilities in the current agentic AI cycle. OpenAI has Operator, Google has Project Mariner, and Microsoft’s Copilot ecosystem is being extended for screen-level control. Anthropic launched its own computer use feature in October 2024 but acquiring a team that had built exclusively in this space — with AI2-caliber machine learning depth — accelerates the roadmap.
The strategic logic: Vercept’s team spent months building infrastructure for real-world computer control. That institutional knowledge doesn’t translate from a GitHub repository. Bringing Ehsani, Weihs, and Girshick inside Anthropic means Claude’s computer use feature has dedicated founders who’ve shipped a commercial product in exactly this domain.
Assessment
This is a well-structured acqui-hire at an early but credible stage. The Vy product never reached scale, but the team pedigree — multiple AI2 alums, proven in vision and agentic tasks — is the real acquisition target. Anthropic gets specialists; Vercept’s investors get a positive return; customers get a 30-day exit window. The only losers in the traditional sense are users who had built workflows on Vy.
The Matt Deitke subplot is the most culturally significant part of the story. A 24-year-old turned down $125M and received $250M instead — and this was considered newsworthy but not implausible in February 2026. That says more about the state of AI compensation than anything else.
Etzioni’s candor is genuinely refreshing. The AI startup space is full of acquisitions dressed up as synergistic partnerships. Calling it “throwing in the towel” is honest, and the investor-founder dispute that followed suggests the early exit was a calculated escape from a harder path, not a celebratory outcome.
Rating: 4/5 — A strategically sound acqui-hire with notable subplots: an unprecedented $250M talent raid and a founder willing to say out loud what most acquisition announcements obscure.