High Leverage Ep. #9, The AI Coding Paradigm Shift with Simon Willison
70 snips
May 5, 2026 
Simon Willison, veteran web developer and Django co-creator who now covers AI and LLM tooling. He discusses the 2025 tipping point when models mastered code, the shift from vibe coding to agentic engineering, how coding agents compress delivery and reshape workflows, and the trust, security, and hiring tradeoffs that follow.
AI Snips
Chapters
Transcript
Episode notes
Use Agents Like Trusted Team Services
- Treat coding agents like semi-black-box team services: rely on docs and tests first, inspect code only when problems arise.
- Simon compares trusting an agent to trusting another engineering team that has proven past quality.
Usage Trumps Surface Signals Of Quality
- Public indicators of quality like tidy repos, tests, and docs are now cheap to fake with agents, so usage history matters more.
- Simon values projects that have been used daily for weeks over freshly generated-looking repos.
Always Audit Security Adjacent Code
- Review security-adjacent code yourself and don't outsource it entirely to agents.
- Simon emphasizes that knowing what's security-adjacent is a skill developed through years of experience and must be guarded.