What Happened
In a post on X, Boris Cherny, a software engineer and former CEO of startup studio XO, shared a "little known fact" about his work at Anthropic. Cherny stated that the Anthropic Labs team—the team he specifically joined the AI company to be a part of—was responsible for shipping several foundational early products.
According to Cherny, this small team shipped:
- MCP (Model Context Protocol)
- Skills
- The Claude Desktop application
- Claude Code
He described the team's early ethos as "just a few of us, shipping fast, trying to keep pace with what the model was capable of." He recalled that the initial prototypes for desktop computer use, developed during the "Sonnet 3.6 days," felt "clunky and slow" but hinted at significant future potential.
Cherny concluded his post by announcing that the team is now releasing full computer use in Cowork and Dispatch, expressing excitement to see how users will leverage the new capability.
Context
The Anthropic Labs team appears to be an internal product development group focused on rapidly prototyping and shipping tools that extend Claude's capabilities into user-facing applications and developer platforms. The mention of "Sonnet 3.6 days" references an earlier iteration of Anthropic's Claude 3 model family, situating this team's early work in late 2023 or early 2024.
The products credited to this team are significant:
- MCP (Model Context Protocol): A protocol that allows Claude to connect to external data sources and tools, effectively giving the model access to a user's specific context (like code repositories, databases, or APIs).
- Skills: Likely refers to specialized capabilities or fine-tuned behaviors for Claude, potentially exposed via an API or within products.
- Claude Desktop: The standalone desktop application for interacting with Claude.
- Claude Code: A specialized tool or interface for using Claude in software development workflows.
The new release of "full computer use in Cowork and Dispatch" suggests a significant expansion of Claude's agentic capabilities. Cowork is Anthropic's collaborative workspace product, and Dispatch is likely a related tool for task management or automation. "Full computer use" implies Claude can now perform a wider range of direct, autonomous actions within a user's digital environment, moving beyond simple chat.
gentic.news Analysis
Cherny's post is notable not for revealing a new technical benchmark, but for providing a rare glimpse into Anthropic's product development philosophy. The existence of a small, fast-moving "Labs" team separate from the core model research division is a classic but effective Silicon Valley strategy. It allows a company to rapidly explore product-market fit for new capabilities enabled by foundational research, without bogging down the core research team or the main product pipeline. This structure is reminiscent of Google's early "Gmail Labs" or various skunkworks projects at other tech giants.
The specific products named—MCP, Claude Desktop, Claude Code—are foundational to Anthropic's strategy of making Claude a practical, contextual tool rather than just a conversational chatbot. MCP, in particular, is a critical piece of infrastructure that turns Claude into a potential platform, allowing developers to connect it to their own data and systems. The announcement of "full computer use" is the logical culmination of this work, representing a shift from Claude as an assistant you talk to about your computer, to an agent that can operate your computer within defined boundaries.
For practitioners, the key takeaway is Anthropic's clear focus on practical integration and agentic workflows. While competitors may tout raw reasoning scores on academic benchmarks, Anthropic's Labs team has been focused on the harder, messier problem of building the pipes and interfaces that allow a powerful model to actually do useful work in a user's existing environment. The success of this approach will not be measured on a leaderboard, but by user adoption of Cowork, Dispatch, and the developer ecosystem around MCP. This is a bet on the primacy of the platform and the user experience, not just the model.
Frequently Asked Questions
What is the Anthropic Labs team?
The Anthropic Labs team is a small, fast-moving product development group within Anthropic, focused on rapidly prototyping and shipping user-facing applications and tools that leverage Claude's capabilities. According to Boris Cherny, it was responsible for early versions of MCP, Skills, Claude Desktop, and Claude Code.
What is MCP (Model Context Protocol)?
The Model Context Protocol is a system developed by Anthropic that allows Claude to securely connect to and interact with external data sources, tools, and a user's personal context (like codebases, databases, or company docs). It's a key piece of infrastructure that enables Claude to be more than a general-purpose chatbot and instead act as a contextual assistant integrated into a user's specific workflow.
What does "full computer use in Cowork and Dispatch" mean?
This announcement suggests Claude's capabilities within Anthropic's Cowork (collaborative workspace) and Dispatch (likely a task management tool) have been expanded. "Full computer use" implies Claude can now perform more autonomous, agentic actions within these environments, such as navigating interfaces, manipulating files, or executing multi-step workflows on a user's behalf, moving beyond text-based conversation.
Who is Boris Cherny?
Boris Cherny is a software engineer and entrepreneur who joined Anthropic to work on the Anthropic Labs team. Prior to Anthropic, he was the CEO and co-founder of XO, a startup studio, and is known for his work in the TypeScript and developer tooling ecosystem.
