Descriptions:
Recorded at DeepLearning.AI’s AI Dev 26 conference in San Francisco, this panel brings together practitioners and leaders from across the AI software stack to debate what software engineering will look like in the next three to five years. Host Marina McGillo (Silicon Valley Girl podcast) moderates a conversation featuring Mika Katasta, President of Replit; Dan Maloney, CEO of Landon AI; Richmond, Director of AI Developer Experience at Oracle; and Joe Ree, author of Fundamentals of Data Engineering.
The discussion centers on a fundamental tension: coding agents are making existing developers dramatically more productive, while simultaneously opening software creation to the estimated one to two billion knowledge workers who previously depended on engineering teams. Panelists from Replit and Oracle push an optimistic view — Replit rates the future of software creators at 10/10 — while others note that junior developers now face the full complexity of systems from day one rather than being shielded by carefully scoped onboarding tasks.
Practical themes include how to manage AI agents well rather than poorly (the “Michael Scott vs. thoughtful manager” framing), the growing importance of deep domain expertise over generic coding skill, and the challenge of helping users understand agent capability ceilings to avoid frustration and backlash. A grounded, multi-perspective conversation rather than hype, making it useful for engineers and technical leaders thinking through how to position themselves or their teams in this shift.
📺 Source: DeepLearningAI · Published May 20, 2026
🏷️ Format: Interview
