Dorsey’s Block says new AI tool handles 15% of code work

The announcement that Block’s Builderbot can autonomously handle 15% of routine coding tasks marks more than just another incremental advance in AI-assisted development—it signals a quiet but profound shift in how software is built at scale. For years, AI coding tools have excelled at generating small snippets or boilerplate code, but their real-world utility has been limited by integration friction: compatibility with legacy systems, security constraints, and the need for human oversight in production environments. Builderbot’s claim suggests it bridges that gap by embedding AI directly into the engineering workflow, where policies, dependencies, and review processes already exist. That’s significant because it implies a future where AI doesn’t just assist developers but becomes a co-pilot for entire codebases, reducing cognitive load on teams while accelerating delivery. This isn’t happening in a vacuum. Block, the fintech giant formerly known as Square, has long prioritized internal tooling as a competitive advantage, investing heavily in automation to support its payments and financial infrastructure. The company’s push into AI-driven development reflects a broader trend among large-scale tech firms: treating engineering not as a cost center but as a strategic asset that can be scaled through automation. If Builderbot delivers on its promise, it could redefine productivity benchmarks in an industry where even a 1% efficiency gain can translate to millions in saved engineering hours. Yet the 15% figure—while notable—also raises questions about what’s excluded. Is this figure derived from internal benchmarks, or does it account for the hidden costs of debugging, security reviews, and integration testing that often dwarf initial development time? The bigger implication may lie in how this tool normalizes AI’s role in mission-critical systems. If Builderbot can safely generate and test code at scale, other companies may follow, accelerating a trend where AI doesn’t just write functions but participates in architectural decisions. But that future hinges on trust—specifically, whether developers and regulators will accept AI-generated code in systems where failure isn’t an option. The open question is whether Builderbot’s approach scales beyond Block’s proprietary environment, or if it remains a niche solution tailored to fintech’s stringent compliance needs. Either way, the story isn’t just about AI writing code; it’s about whether we’re ready for AI to help build the systems we all depend on.