
AI's impact on software engineering

High growth startups: Uber and CloudKitchens with Charles-Axel Dein
Topic 16 of 17
Published Sep 24, 2025 · Clip 1:32:46–1:42:27 (9:41)
Charles shares Cloud Kitchens' internal study finding AI coding tools moderately useful. He finds AI best for well-defined problems, navigating code in the monorepo, refactoring, and well-specified APIs — design remains a human task. Agents let him code more by multitasking between meetings. He discusses AI in engineering management (using it as a coach for document reviews), the principle of never copy-pasting AI text because writing is thinking, and the absurdity of AI-generating then AI-summarizing text. He emphasizes reading AI-generated code to make it your own, warns that vibe-coded codebases become unmaintainable, predicts security engineering will boom due to AI agents at scale and supply chain attacks, and highlights migrations as Cloud Kitchens' best AI use case with copy-paste prompts plus code review. The segment closes with his favorite languages: Python and TypeScript, with Clojure and Rust as languages he learned but rarely used.