Anthropic's Labs Team Drives Product Innovation Ahead of IPO
A rotating 20-person incubator built Claude Code and other breakthrough tools by staying close to AI research advances.

Anthropic's rapid product development stems from an unconventional internal structure: a 20-person rotating team called Labs that functions as a startup incubator within the AI company.
Led by Anthropic cofounder Ben Mann, the Labs team operates on two-week cycles, rapidly prototyping new product ideas and killing most of them. The group has produced some of Anthropic's most significant commercial offerings, including Claude Code, the Model Context Protocol for connecting AI agents to data sources, and Claude Design. According to Business Insider, which first reported these details, the team maintains a 20% to 30% success rate with its experimental projects.
How the incubator model works
Labs functions through a "persevere or pivot" review process every two weeks. Failed concepts get discarded or merged into other prototypes, while successful ideas "graduate" into standalone teams within Anthropic. Team members rotate frequently—projects typically move out of Labs once they exceed four people, creating constant churn as employees shift to other divisions or new staffers join.
Mann drew inspiration from Bell Labs and Google's Area 120, where he previously worked. The model gives Anthropic a structural advantage: product developers hear about AI model improvements directly from researchers, enabling them to build tools before capabilities become widely known.
This proximity to research proved crucial for Claude Code's development. Mann said his team learned that new models showed strength in agentic coding well before late 2024, when engineer Boris Cherny built an internal prototype. After Anthropic released it as a preview in February 2025, subsequent model improvements transformed Claude Code into what Mann called an "uber-popular product" that accelerated the company's growth.
Why it matters
As Anthropic prepares for a public offering, the Labs model represents both an innovation engine and a strategic challenge. The team must balance rapid experimentation with the demands of public market investors expecting predictable revenue growth. Additionally, products like Claude Design put Anthropic in direct competition with established software companies including Adobe and Figma—a dynamic that could complicate partnerships with enterprise customers who use multiple tools.
Mann acknowledged pushing employees to think bigger. When Cherny initially proposed a code analysis tool, Mann rejected it as insufficiently ambitious. "A lot of what the agents can do, nobody knows," Mann explained.
The Labs approach has spread to other Anthropic engineering teams, which now run their own "bets" groups. Mann said he wants Labs to tackle applications in biological research and clean energy storage, aligning product development more closely with Anthropic's stated mission of benefiting humanity through AI systems.
The team includes notable figures like Instagram cofounder Mike Krieger. Mann noted that Labs emerged from his earlier efforts to convince Anthropic leadership to release products at all, rather than rely solely on philanthropic funding while pursuing advanced AI research.
These details were first reported by Business Insider.
This is an original analysis by the Omega editorial team. Source reporting: AI Watch.
Want systems like this working for your business?
Book a Call
