Data Tech 2026

A year ago I set a specific goal: build a production system without writing a single line of application code. I reached it. And then it kept going. I also stopped reading the generated code, stopped touching configuration files, stopped issuing CLI commands, and eventually the running system began monitoring itself and correcting drift before I was notified. This talk is the story of how each of those steps happened, what structural discipline made them possible, and where the cascade ends when you keep asking what comes next.

The discipline is called Generative Specification. Attendees will leave with a clear understanding of why AI sessions drift and what architecture prevents it, what a complete specification looks like for a real data or software system, and a concrete starting point for their own work. We will also look at what happens when this approach transfers beyond one practitioner: in April 2026, 60 developers worked on two projects under two conditions: standard AI assistance versus specification-governed AI assistance. The results have direct implications for data and analytics teams building on AI tooling.