The AI Skill Flip

Sheamus McGovern is the founder of ODSC, the Open Data Science Conference, and a venture partner and Head of AI at Cortical Ventures. He is the author of The AI Skill Flip, a book about how AI is changing the skills professionals need. Dan and Sheamus discuss the shift from producing work directly to evaluating what AI produces. Sheamus argues that domain expertise and human judgment become more important as output gets cheaper. Prompt engineering, context engineering, and skills remain related ways of giving an agent the information it needs, even as the terminology changes. Their discussion also confronts the costs of faster building. Sheamus describes accumulating technical debt with every new feature and cognitive debt when people stop practicing the work they delegate. He argues that junior engineers still need programming fundamentals alongside an understanding of AI. Companies need to invest in training and evaluation, rather than assuming that access to a coding tool makes people effective. Chapters: 00:00:00 Introduction 00:01:10 The AI Skill Flip and why doom and utopia are both wrong 00:06:20 Passion, creativity, and judgment 00:12:37 Can judgment be taught? Evals and domain expertise 00:17:33 Not magic, just math: seeing AI as pattern recognition 00:19:41 Prompt engineering, context engineering, and the harness 00:24:40 Why knowledge work is open-ended and unstructured 00:28:02 How Sheamus's own AI use changed while writing the book 00:31:05 An AI assistant for venture deal flow 00:34:58 Building a speaker CRM with planning mode 00:36:39 The AI credit card and AI technical debt 00:39:09 Build vs. maintain, spreadsheets, and generative UI 00:44:49 The unglamorous data problems AI doesn't solve 00:46:28 Junior engineers and cognitive debt 00:53:23 What AI still gets wrong: latency, tokens, and query plans 00:55:54 What's next: computer use, brownfield coding, and real training 01:00:31 How companies should actually teach AI skills