Sonar CTO Andrea Malagodi ran cost per PR across his engineers' AI spend - but the engineers with 500+ PRs per week weren't always the ones delivering the most value. @SonarSource's CTO and I unpack his research and get into: - What engineering value delivery looks like with AI agents (and how cost per PR can mislead) - Keeping agent work small enough to inspect (a five-day agent session that writes 400,000 lines leaves you blind) - Rolling AI tools out at enterprise scale, from 30% token savings to $50-a-week budgets + much more on the craft of software engineering in an era of AI factories Chapters: 0:00 Human reviewers rubber-stamp big AI changes 2:43 Learning the craft alongside AI tools 9:09 From lines of code to PR counts 12:17 Spreading top engineers' setups org-wide 16:50 The guide, verify, solve agent loop 22:51 Specialized agent lanes and multitasking limits 26:51 Clear asks and guardrails for agent coordination 30:29 Building disagreement between agents 34:59 Mixing models instead of picking one 39:46 Rolling out AI tooling at enterprise scale 46:19 Defense in depth for AI-written code 50:52 Protecting time for incident investigations 54:08 Personalized software for teams and individuals Brought to you by: - @SvixHQ: reliable webhooks for startups and the Fortune 500. link.svix.com/cot - @WalrusProtocol: Walrus Memory, portable and verifiable memory for AI agents. walrus.xyz/cot - @inngest: durable execution for long-running agents. inngest.link/cot-x - @g2i_ai: reviewing RL environments, evals, and training data. fandf.co/3SFxVm6 Look up Chain of Thought on YouTube, Spotify, Apple Podcasts, or wherever you listen. Thanks, Andrea!