The Peterman Pod Turing Award Winner: Postgres, Disagreeing with Google, Future Problems | Mike Stonebraker
137 snips
Apr 20, 2026 
Mike Stonebraker, Turing Award–winning creator of Ingres and Postgres, and serial database entrepreneur. He tells the origin story of Postgres and why one-size-fits-all databases fail. He explains clashes with Google/Amazon approaches, the DBOS idea of replacing OS state with a database, and why future agentic AI and text-to-SQL tasks push new database challenges.
AI Snips
Chapters
Books
Transcript
Episode notes
Identify Exceptional Engineers With Deep Technical Questions
- Ask deep technical questions to surface true ability when hiring, such as probing thesis details, error handling, process counts, and design choices.
- Mike says smart candidates stand out by accomplishing significantly more than peers and answering rigorous implementation questions.
One Size Database Loses To Specialized Engines
- One size fits none: specialized database architectures (stream processors, column stores, row stores) outperform general systems by orders of magnitude for specific workloads.
- Mike notes Postgres is great as a lowest-common-denominator but lacks column-store and multi-node features for petabyte warehouses.
Why GPUs Struggle With Indexed Queries
- GPUs are SIMD and struggle with pointer-chasing index patterns like B-trees because these require serial memory accesses that don't vectorize well.
- Mike explains walking a B-tree involves repeated memory loads and pointer-following, which won't parallelize on GPUs.
