SE Radio 732: Jason Gorman on The Effective Use of AI For Software Development

AI in software development functions most effectively as a tool for specific, constrained tasks rather than as an autonomous replacement for human engineers. Large language models suffer from reliability issues and context dilution, making them unsuitable for long-horizon, agentic coding. Instead, integrating AI into established agile practices—such as test-driven development and small, frequent feedback loops—allows developers to manage the technology’s inherent unpredictability. Jason Gorman, a veteran software architect, highlights that AI acts as a stress test for existing workflows; teams lacking robust processes often experience increased technical debt and production incidents. While AI can accelerate code generation, the real bottleneck in software development remains human comprehension and architectural design. Over-reliance on automated tools risks eroding critical cognitive skills, underscoring the necessity of maintaining human oversight and foundational engineering discipline to ensure long-term project stability and quality.