Vibes won't cut it
summary
Argues that AI can generate lots of code but 30% generated code is not a solution for production systems because that code still runs in complex, failure-prone environments and requires human judgment. Recommends concrete practices: documented standards, reproducible developer environments, fast tests, clear task boundaries, and a create refine loop where plans are authored as files, executed, and iterated. Emphasizes that code review is the critical skill going forward and warns LLMs only generate text not definitive behavior, so humans must validate and own changes.
AI-written summary · about
chapters (16)
- 0:00Intro
- 0:59Big code bases
- 1:57AI still writing code
- 3:13Vibes
- 3:45Vibe coding
- 4:09Code is not the job
- 4:46Jeff Atwood
- 5:19All codes have tradeoffs
- 5:31Microservices vs monoliths
- 6:36Software engineering
- 7:36Production
- 8:28AI coding
- 9:13AI coding tips
- 10:30Software engineering vs AI
- 11:20Code review
- 12:08Bonus tips
related talks
38mNobody Pages the LLM: Engineering Rigour for Vibe Coding
43mCode Review in the Age of Vibe Coding
32mFrom Vibes to Outages: Riding the AI Code Wave
36mEngineering Intent: Solving the AI Productivity Paradox
50mPanel: Beyond The Vibe: Code Quality First
26mMore. Better.
32mCan We Trust AI-Generated Code?
21mAI's impact on software architecture