Data Renegades Ep. #11, Contrarian Bets and AI Skepticism with Michael Stonebraker
Apr 28, 2026
Michael Stonebraker, legendary database pioneer behind Ingres and PostgreSQL, shares his long arc in data systems. He discusses why specialized database engines beat one-size-fits-all platforms. He critiques AI hype around applying LLMs to messy enterprise warehouses. He recounts startup lessons, research pivots, and what keeps him curious after five decades.
AI Snips
Chapters
Transcript
Episode notes
Use Stored Procedures But Build Debugging Tooling
- Do put compute next to data via stored procedures and UDFs when performance matters to avoid round trips.
- Invest in developer tooling and debuggers for stored procedures because lack of ergonomics blocks adoption.
Dboss Evolved From A Linux Replacement Prototype
- Dboss began as a research OS-replacement prototype after noticing Linux scheduler limits at Databricks and moved to a commercial TypeScript/Postgres runtime.
- VCs forced focus away from replacing Linux toward durable programming environments.
NoSQL Pushed SQL To Evolve Not Replace It
- NoSQL’s UX innovations (schema flexibility, JSON) forced SQL systems to adopt similar features, narrowing differences.
- Stonebraker argues transactions and optimizers were always crucial and NoSQL systems moved back toward SQL semantics.