Block Launches AI-Native Tool Builderbot to Handle 15% of Production Code Changes
Jack Dorsey's financial services company Block launched the AI-native tool Builderbot on Wednesday, which currently executes about 15% of the company's production code changes, completing over 200,000 operations daily and merging approximately 1,500 pull requests weekly.
Brad Axen, Head of AI Capabilities at Block, stated that Builderbot fills the gap between AI coding tools and large-scale operations of engineering teams, reducing work that previously took months to just days.
As an orchestration layer, Builderbot coordinates multiple AI agents to understand Block's complete codebase, services, APIs, and conventions, allowing engineers to modify at any system location. This marks a shift from AI-assisted coding to AI-native engineering.
Source: Public Information
ABAB AI Insight
Block, previously known as the parent company of Square and Cash App, has long relied on internal engineering teams to maintain a large fintech codebase. Under Jack Dorsey's push for a Bitcoin and open technology strategy, Builderbot is built on its open-source Goose framework, reflecting the company's leap from tool adoption to autonomous AI engineering capabilities.
On the capital front, Block is shifting engineering resources from repetitive coding to high-value innovation, accelerating iterations through AI agent orchestration. The motivation is to maintain a competitive edge in financial services and the Bitcoin ecosystem, compressing months of work into days, directly enhancing delivery speed and capital efficiency.
Similar to the evolution of OpenAI Codex or GitHub Copilot towards enterprise-level agents, as well as other tech giants' internal AI coding systems, the software engineering field is undergoing a transformation from auxiliary tools to full-stack AI-native development, with Block taking a leading position.
Essentially, this represents an acceleration of technological substitution, where the mechanism of multi-agent coordination breaks the limitations of a single codebase, achieving cross-system automation and shifting pricing power towards companies that master AI engineering infrastructure, driving a structural reconstruction of code productivity across the industry.
ABAB News · Cognitive Law
Code is not written; it is orchestrated. The more AI agents there are, the more engineers resemble conductors.
Time compression equates to capital amplification, from months to days, with efficiency leverage determining survival boundaries.
A 10% improvement in auxiliary tools leads to a 100% reshaping of native systems; the critical point of qualitative change from quantitative change has arrived.