Simon Willison: Engineering practices that make coding agents work - The Pragmatic Summit

The discussion centers on leveraging AI coding agents to enhance developer productivity, with Simon Willison sharing his experiences and strategies. He details a shift from occasional AI assistance to a workflow where agents write the majority of code, emphasizing the importance of test-driven development (TDD) and manual testing to ensure code reliability. Willison introduces "ShowBoat," a tool for documenting manual tests, and explores "conformance-driven development," using existing test suites for implementation. Addressing risks, he highlights "prompt injection" and the "lethal trifecta" of private data access, malicious instructions, and exfiltration vectors, advocating for sandboxing and careful data handling. He also touches on the evolving role of open source in an AI-driven development landscape.