Skip to main content

Share story

AI

Building a C compiler with a team of parallel Claudes

Nicholas Carlini, a researcher on the Safeguards team, experimented with agent teams in which multiple Claude instances work in parallel on a shared codebase without active human intervention. He assigned 16 agents the task of writing a Rust-based C compiler from scratch that could compile the Linux kernel. The agents completed the work over nearly 2,000 Claude Code sessions at a cost of $20,000 in API usage and produced a 100,000-line compiler able to build Linux 6.9 on x86, ARM, and RISC-V. Carlini built a harness that places each Claude in a continuous loop so that the agent selects the next task immediately after finishing one. Sixteen agents ran in separate Docker containers mounted to a shared upstream git repository, with a synchronization method that prevented two agents from addressing the same problem at once. The project showed that independent test cases allow easy parallel work, while a single large task such as kernel compilation initially caused agents to overwrite one another's changes. A new harness that used GCC as an oracle for random file subsets let agents fix distinct bugs in parallel until the compiler handled the full kernel. Agents were also assigned specialized roles, including removal of duplicate code, performance improvements to the compiler, generation of efficient output, structural critique from a Rust viewpoint, and documentation updates. The effort served as a benchmark for the Claude 4 model series and followed an initial design goal of a from-scratch, dependency-free, GCC-compatible optimizing compiler with an SSA IR and support for multiple backends.
Sources
In this story
Published by Tech & Business, a media brand covering technology and business. This story was sourced from Anthropic and reviewed by the T&B editorial agent team.
Back to Newswire
Keep reading
Full wire
AI Policy
AI Policy

US appeals court upholds Pentagon restriction on Anthropic's Claude

A US appeals court upheld the Pentagon's designation of Anthropic as a supply chain risk, leaving in place a restriction on the military and its contractors using Claude. In a 2-1 decision, the DC Circuit denied Anthropic's challe...

AI Products
AI Products

Anthropic opens Claude Code cloud sessions to eligible subscribers

Anthropic has made Claude Code cloud sessions available to eligible subscribers without requiring enrollment in its research preview, BleepingComputer reports. The sessions run coding tasks on Anthropic's infrastructure, allowing ...

AI Policy
AI Policy

White House asks AI labs to delay model access for UK testers

The White House has asked OpenAI and Anthropic to keep new AI models from the UK's AI Security Institute until the US government tests them, Politico reported. A British official confirmed the reported request to Bloomberg. Politi...

AI Security
AI Security

AI agent incidents traced to access error in security tests

Several incidents in which AI agents targeted real-world systems stemmed from one testing setup at Israeli startup Irregular, its co-founder Omer Nevo told The Verge. Irregular was evaluating models from OpenAI, Meta, Anthropic an...

Infrastructure Capital
Infrastructure Capital

Nscale secures $3.36 billion in financing ahead of planned IPO

British AI data center developer Nscale said it secured $3.36 billion in convertible-note financing ahead of a planned US stock-market listing. The Third Point-led deal makes $2.36 billion available immediately, while a further $1...

Capital AI
Capital AI

Enveda raises $311 million to advance AI-assisted drug candidates

Enveda has raised $311 million in Series E financing at a $2 billion valuation as it moves drug candidates found through its AI-assisted search of natural compounds into human testing. Catalio Capital Management led the round, wit...