Latest Articles

Mapping the Variant Type in Iceberg v3: Standardizing Semi-Structured AI JSON Payloads

Mapping the Variant Type in Iceberg v3: Standardizing Semi-Structured AI JSON Payloads

How Iceberg v3 variant-style typing stores evolving AI payloads like prompts and tool calls, and when to promote JSON fields into governed, typed columns.

Read more
Designing Idempotent Pipelines in the Agentic Lakehouse: Eliminating Double-Write Anomalies

Designing Idempotent Pipelines in the Agentic Lakehouse: Eliminating Double-Write Anomalies

How to stop agent retries and reruns from writing duplicates in an Iceberg lakehouse, using operation IDs, intent logs, deterministic writes, and validation.

Read more
File Encryption for the Lakehouse: The Terminology, the Machinery, and the Hard Problem of Interoperable Encrypted Tables

File Encryption for the Lakehouse: The Terminology, the Machinery, and the Hard Problem of Interoperable Encrypted Tables

Encryption terms, Parquet modular encryption, and Iceberg 1.11 table encryption explained, plus the problems of sharing encrypted tables across engines.

Read more
MCP 2026-07-28 Spec: Stateless MCP for Data Platforms

MCP 2026-07-28 Spec: Stateless MCP for Data Platforms

MCP 2026-07-28 shipped July 28, 2026: no initialize handshake, no session IDs, a new server/discover call. What stateless MCP means for data platform tools.

Read more

Work with Alex