Block's New AI Tool Now Writes 15% of All Its Production Code Changes
"The best way to think about Builderbot is as the missing layer between AI coding tools and how engineering actually works at scale."
Block announced Builderbot on June 17, 2026, an AI orchestration layer built to solve a specific problem the company says it kept hitting despite two years of AI investment across its engineering organization.
Coding tools work well inside a single repository. None of them could operate across hundreds of millions of lines of code, hundreds of services, and the full complexity of how Block actually builds software, according to the press release.
Builderbot runs inside Slack. An engineer tags @builderbot with a short description of what they need, a bug fix, a migration across services, a new feature, and the system gets to work directly in the thread.
Multiple team members can collaborate with it in real time, watching it research, plan, and implement while they steer the direction, without switching to a separate development environment.
"The best way to think about Builderbot is as the missing layer between AI coding tools and how engineering actually works at scale," said Brad Axen, Head of AI Capabilities at Block. "It handles the orchestration, the context, the environment, so our engineers can focus on the problems worth solving."
What Makes It Different
The distinguishing factor is scope rather than capability. Builderbot understands the full context of Block's codebase and can contribute to any piece of code across the company.
An engineer working on Cash App can use it to make a change inside a Square service they have never touched, because the system already understands how that service works.
Builderbot picks up tickets directly from Linear and Jira, creates the branch, writes the code, opens the pull request, watches continuous integration, and iterates based on feedback, according to the press release.
The system operates only on source code and system configuration, it does not access or process customer data, payment information, or personally identifiable information. This is a deliberate boundary given Block's position as a financial services and payments company.
Axen pointed to a concrete outcome on the Square side, where a list of seller-requested features that had been waiting for months shipped in days.
"Builderbot handled the scaffolding and the repetitive work, and our engineers made the decisions that shaped the product," he said. "It means an idea can go from backlog to live in front of millions of customers in days instead of months."
Builderbot is built on goose, the open source AI agent framework Block developed and contributed to the Agentic AI Foundation under the Linux Foundation, according to the press release.
The integration challenges Block encountered while building Goose directly inspired its collaboration with Anthropic on the Model Context Protocol, which has since become an industry standard for connecting AI agents to tools and data sources. Block says 100% of its engineers now regularly use AI in their work.
Key Takeaways
- Block's Builderbot now generates 15% of all production code changes, enhancing coding efficiency.
- Utilize Builderbot for real-time collaboration within Slack, streamlining engineering processes.
- Leverage Builderbot's contextual understanding to make cross-service code changes seamlessly.
- Integrate Builderbot with Linear and Jira for automated ticket handling and code management.
- Recognize Builderbot as a vital orchestration layer bridging AI tools and large-scale engineering.