BONUS Why a Distinguished Engineer Stopped Reading Code — Lights-Out Codebases and the End of the IC With Philip Su

In this special bonus episode of the Scrum Master Toolbox Podcast, host explores the transformative impact of AI on software development with Philip Su, a veteran engineer who has worked at Microsoft, Facebook, Amazon, and OpenAI. Su shares his unconventional career journey—from ambitious developer at Microsoft to warehouse worker at Amazon—highlighting a personal evolution from 'dirty fuel' (external validation) to 'clean fuel' (intrinsic passion). His time in Amazon's warehouse revealed critical lessons about incentives, legacy systems, and the inevitability of automation, drawing parallels to today’s resistance toward AI in coding. Su argues that the era of human code reviews is ending, with AI-generated codebases becoming 'lights-out'—operating without human inspection. He challenges the notion that human understanding of every line of code is essential, citing parallels with modern systems like cars and cloud infrastructure that we use without comprehension. As a solo founder building Superphonic, an AI-assisted podcast app, Su demonstrates how his role has shifted from coding to managing AI agents, making him more of a product manager than an individual contributor. The episode concludes with a provocative vision: the future of software teams may consist only of engineering managers, PMs, and designers, as AI handles the actual code production. The host emphasizes that the real bottleneck is no longer writing code, but strategic decision-making, and urges teams to rethink planning, coordination, and experimentation in an age of 10x–1000x productivity. Key takeaways include: (1) AI is not just a tool but a new infrastructure layer that demands rethinking team structure and workflows; (2) The focus must shift from code volume to strategic product decisions, as AI enables unprecedented speed and scale; (3) The future of engineering is not coding, but managing AI agents and resolving conflicts between their recommendations; (4) Teams must experiment with AI-driven development now—through hands-on building—not just by consuming content; (5) The real productivity gain isn’t in writing more code, but in using that surplus to run experiments, gather data, and make better decisions. Su warns against the trap of using AI solely for efficiency, urging organizations to embrace the systemic shift rather than resist it. The episode ends with a call to action to attend the Global Agile Summit on May 4th, featuring real-world tracks on AI, people, construction, and gaming.