Thinking on data, AI, and engineering execution.
Practical writing for CTOs, Heads of Data, and engineering leaders who are building at scale.
The Third Map
For years the answer to "where should the data live" was latency.
You Cannot Govern What You Have Not Found
Every governance program I have seen started with a policy document and a catalog license. Every governance failure I have seen started somewhere neither of them could see.
Anyone Can Build It. Almost Nobody Can Trust It.
The pipeline was green. Every job ran. Every check passed. The data was still wrong, and nobody noticed until the report had already been used to make a decision.
Why Most of My AI Never Touches the Cloud
Most of what I ask NOMAD to do never leaves this machine.
Same Prompt, Different Brains: How NOMAD Generates Images on My Own GPU
Every cover in my recent posts was made here, on my own GPU, for nothing.
I Stopped Talking to AI Models. I Built One Front Door Instead.
I run several projects at once. For a while my setup was a browser full of tabs, one per AI tool, and no real idea which one was costing me what.
Your Data Platform Was Built for Humans. The Users Are Becoming Machines.
For thirty years we built data platforms for people.
AI Is Changing Data Engineering. The Foundations Are Not.
According to Databricks' 2026 State of AI Agents report, AI agents are now creating more than 80% of databases on the platform. In October 2023, that number was 0.1%.
Companies Are Investing in AI at the Wrong Level
The budgets are enormous.
Why Most Data Platforms Fail
Most data platforms do not fail because of technology. They fail because of decisions.