The Pragmatic Engineer From IDEs to AI Agents with Steve Yegge
1363 snips
Mar 11, 2026 
Steve Yegge, veteran engineer and former Amazon and Google developer, dives into AI’s reshaping of software work. He talks about why hand coding may fade. He maps levels of AI adoption, from avoiding tools to orchestrating parallel agents. They explore Gastown, monoliths blocking agentic workflows, prototype driven product building, and the new technical debt and productivity pressure AI may create.
AI Snips
Chapters
Books
Transcript
Episode notes
Big Companies May Miss The AI Innovation Wave
- Steve Yegge thinks visible AI gains lag in large businesses because they are risk-averse and structurally unable to absorb faster engineering throughput.
- He argues innovation will surface first from small AI-native startups and API-like building blocks, not entrenched giants.
Agentic Codebases Accumulate Heresies Not Just Debt
- Vibe-coded systems accumulate a new debt Steve Yegge calls heresies: wrong architectural ideas that agents keep reintroducing after partial cleanup.
- His Gastown examples include duplicate live databases and agents repeatedly creating pull requests despite explicit intent not to.
The Bitter Lesson Favors Scale Over Cleverness
- The Bitter Lesson, as Steve Yegge applies it, says hand-crafted cleverness loses to bigger models and more compute over time.
- He argues current S-curve skeptics may be technically right eventually, but at least two more scaling cycles still remain.
