Block launches AI agent Builderbot: 15% of code is now written by a neural network
Blockchain company Block, led by Jack Dorsey, has taken a significant step in automating development by implementing the AI tool Builderbot. According to my data, this system already generates about 15% of all the company's software code. This is not just another AI assistant, but a full-fledged orchestrator of AI agents integrated into the corporate messenger Slack.
How Builderbot Works
An engineer simply needs to tag the bot's account and describe the task. Builderbot independently analyzes the code, finds errors, suggests fixes, or creates new features. The key difference from standard solutions is access to Block's entire codebase. This allows, for example, a developer from the Cash App team to make changes to Square services they have not worked with before. The bot automatically takes tasks from Jira, creates branches in the repository, writes code, and submits a Pull Request.
Scale and Performance
The scale is impressive: Builderbot performs over 200,000 operations per day, and closes about 1,500 code merge requests weekly. As noted by Block's head of AI capabilities, Brad Axen, "what used to take months now takes days." The bot handles routine tasks and environment setup, allowing engineers to focus on solving complex problems.
Security and Technological Foundation
It is important to emphasize that the tool works exclusively with source code and system configurations. It has no access to customer data or payment information. Technologically, Builderbot is built on goose, an open-source framework contributed by the Agentic AI Foundation. Block also collaborated with Anthropic to create the Model Context Protocol (MCP).
My analysis: The transition from simply using AI for code writing to "native" engineering processes based on neural networks is not just a trend, but a new reality for the IT industry. Block demonstrates how AI assistants can be scaled to the level of corporate systems, which will inevitably lead to a redistribution of roles in development teams. However, the key issue remains the balance between automation and maintaining control over critical systems.