AI is writing more code than ever before—but most of it never reaches production. The real challenge isn’t capability; it’s trust.
Modern applications are inherently distributed: mobile apps communicate with cloud services, which interact with other services, queues, and databases. This complexity creates edge cases that AI systems struggle to anticipate and traditional testing approaches often fail to uncover. These issues arise when critical system contracts remain implicit rather than explicitly defined and verifiable.
In this talk, we explore what it takes to trust AI-generated code in production environments. Through the development of a real-world distributed application using Generative AI, we will examine the patterns, tools, and infrastructure practices that enable AI-assisted development to scale safely and reliably. Rather than simply coding faster, developers must evolve their approach to building systems that remain understandable, testable, and dependable—even when AI writes much of the code.
Finally, we’ll look ahead at emerging research and frameworks that make entire classes of distributed system failures impossible to express. These innovations point toward a future where AI doesn’t just generate code more quickly—but produces distributed systems that are correct by design and provably reliable.
