The Joe Reis Show From ODBC to ADBC: Modernizing the Data Stack for AI and Analytics w/ Ian Cook
20 snips
Feb 17, 2026 Ian Cook, co-founder of Columnar and longtime Apache Arrow contributor, modernizes database connectivity with ADBC. He explains why row-based protocols persist, the costs of converting columns to rows, and how keeping data columnar speeds analytics and AI. Ian also explores tabular data’s role in LLMs and the balance between rapid coding and stable open-source infrastructure.
AI Snips
Chapters
Transcript
Episode notes
Arrow Standardizes In-Memory Columnar Data
- Apache Arrow standardizes an in-memory, columnar format that lets different languages and systems share tabular data efficiently.
- Ian Cook says Arrow avoids decoding and enables fast network transfer for analytics workloads.
ADBC Keeps Data Columnar End-To-End
- ADBC (Arrow Database Connectivity) is a modern, columnar replacement for ODBC/JDBC that keeps data columnar end-to-end.
- Ian Cook claims ADBC can materially displace legacy row-based protocols for analytics and AI.
Serialization Tax Slows Analytics Transfers
- ODBC/JDBC impose a costly serialization tax by transposing columnar results into rows and back again.
- Ian Cook attributes slow query transfers (e.g., Snowflake to Polars) to that conversion overhead, not just network limits.