Neal Ford and Sam Newman Discuss Agentic AI

Watch the entire Superstream: https://learning.oreilly.com/videos/software-architecture-superstream/0642572277543/0642572277543-video403958/?utm_medium=social&utm_source=youtube&utm_campaign=free+trial&utm_content=sa+superstream As agentic AI coding tools crank out more and more code, the software architects tasked with managing it are turning to a surprisingly old playbook. Speaking at the Software Architecture Superstream, Sam Newman and Neal Ford make the case that modular programming principles from the 1960s are exactly the right framework for bounding AI agents and verifying their output. You'll want to check this one out for Newman's breakdown of why larger context files are actively degrading AI output quality, how breaking your problem space into smaller units solves the verification problem and the context problem at once, and what "engineering through specification" actually looks like in practice. Neal and Sam close with a sharp exchange on AI-assisted code reviews that cuts right to a concern many engineering leaders haven't caught up with yet: "I saw somebody saying our code reviews are much faster now with AI," Sam notes. "But that's because your commits are so huge that no one's reviewing them." Neal's response: "But the AI is reviewing them. So what could go wrong?" Follow O'Reilly on: LinkedIn: https://www.linkedin.com/company/oreilly/ Facebook: http://facebook.com/OReilly Instagram: https://www.instagram.com/oreillymedia BlueSky: https://bsky.app/profile/oreilly.bsky.social