Stop being skeptical about AI for development with Charity Majors
Summary
Charity Majors discusses the polarization between AI enthusiasts and skeptics, arguing that reliability is declining as engineers ship unread code. She advocates for adopting QA and Ops validation practices to manage non-deterministic AI systems and emphasizes that engineers must "own the loop" rather than blindly trusting tools. The episode covers the need for higher engineering discipline, the shift from editing to replacing code, and the importance of honest dialogue about AI's costs and benefits.
Topics discussed
Introduction: The two camps of AI in software engineering
Charity Majors' background: Linden Lab, Facebook, and Parse
Open source contributions and the limits of individual metrics
The 2025 AI inflection point and the rise of tooling
AI coding capabilities and the shift to immutable infrastructure
Code generation, evals, and the role of QA and sysadmins
Rethinking code reviews and testing in production
AI-validated PRs and behavioral testing strategies
Deterministic vs non-deterministic systems and AI norms
Sponsors: WorkOS authorization and BuildKite CI/CD
Respect, attention economy, and the rising quality bar
SEVZeros, on-call stress, and bridging the AI divide
AI as a tool, code vs prose, and future hiring trends
DevOps, separation of concerns, and fast feedback loops
Telemetry, observability, and AI agent insights
Observability book updates and vendor partnerships
Leadership skills, business acumen, and team dynamics
Career advice: Embracing AI and navigating job markets
AI fatigue, social media burnout, and bottom-up adoption
Closing thoughts: Ethics, limits of growth, and trust
Listen ad-free on Castria