All posts
Writing for data teams
Data engineering, the medallion architecture, and lessons from building OptimaFlo.
Connecting an LLM to Your Data Is the 21% Solution.
The connector pitch is everywhere. The public numbers on it: 21% at Anthropic, 10 to 17% on Spider 2.0. What closes the gap, and what a connector never builds.
Anthropic Proved the Agentic Data Stack Works.
Anthropic's self-service analytics post nails the hard part: the agentic stack around the SQL. It also assumes a data team most mid-market companies don't have.
Everyone Just Agreed on the Foundation. Nobody's Answering Who Gets to Run It.
Fivetran and dbt bet that trust in AI is built at the infrastructure layer, on open standards and governed data. I agree, except for one line: no human in the loop.
Apache Iceberg in Your Own Cloud, Without the Overhead
You do not need a data team to run Apache Iceberg anymore. The real choice is convenience versus control. Here is how OptimaFlo gives you open Iceberg in your own cloud without the operational overhead.
The Medallion Architecture, Explained: Raw, Clean, Ready
A plain-English guide to the medallion architecture: what the Raw, Clean, and Ready layers do, when to use each, and how to build one without a data team.