Writing

Worked out here first,
then built.

Essays on enterprise data, agentic AI, and the judgment layer that decides what a model's output should be trusted to do. Seven pieces, oldest going back to an Oracle optimizer question from 2013.

Governance26 May 202612 min
Storage Has an Owner. Meaning Has an Accident.
We assign storage ownership with precision and interpretation authority by accident. Four moves to control the handoffs where a wrong meaning feeds a decision that matters.
Read
Agentic AI17 March 202611 min
The Human in the Loop Has to Know What Right Looks Like
Human in the loop without domain expertise is not a safety net. It is a spectator.
Read
Architecture17 June 202510 min
The Architect's Gambit: AI-Powered Courage for a Future of Possibility
Architecting for probability is essential. Architecting for possibility is where courage wins.
Read
AI Strategy26 February 202518 min
It Is Time for a UMI Moment for India
19,500 dialects, $100M compute economics, and 0.1% of global training data. Why distillation is India's viable path.
Read
AI Strategy4 February 202512 min
DeepSeek R1: Hype, Reality, and the Future of AI Deployment
Value delivery was never about compute size. What that means for enterprise AI strategy.
Read
Thought Experiment20 January 20257 min
Agentic AI: Monetizing Your Own Digital Twin
A deliberately fictional piece. What if agentic AI let individuals monetize their own data instead of platforms doing it for them?
Read
Oracle25 April 20138 min
Indexes and IS NULL
When the Oracle optimizer will and will not use an index for an IS NULL predicate, tested rather than assumed.
Read