
Vibe Coding Insights and Pitfalls

CTO Coach: Why Tech Companies are Really Laying Off Developers (It’s Not Just AI) - Stephan Schmidt
Topic 10 of 16
Published Jan 12, 2026 · Clip 27:41–32:08 (4:27)
Stephan says he no longer looks at source code and vibe codes a 35,000-line application happily, but stresses you need to be a good coder to be a good vibe coder because prompting anticipates where AI derails. He explains LLMs are non-deterministic probability machines that generate the most likely code fulfilling your prompt and existing codebase — so a bad codebase produces bad code. Starting from a great template and refactoring periodically are essential; without these, vibe coding goes bad and it's easy to shoot yourself in the foot.