I build tools that make data and AI accessible to everyone. Creator of Seeknal, co-founder of ClaimMind. I write about data engineering, harness engineering, and the journey from engineer to founder and CTO.
Loop engineering is not just running agents longer. It is deciding which unknowns minute-, hour-, and day-scale feedback loops are allowed to close.
Loop engineering moves AI work beyond better prompts into repeatable systems that ask, check, retry, remember, and know when authority requires verification.
Agentic engineering moves engineers from chatting with AI tools to designing harnesses, software factories, verification gates, access controls, and token-to-value systems that make agents reliable in production.
Jevons' paradox applied to analytics: when AI collapses the cost of querying data, value moves from access to judgment, workflow, and systems of intelligence that turn raw data into trusted decisions.
Token-maxing looks like AI burn from the outside, but for builders it can be a disciplined R&D budget for buying clarity, reusable artifacts, and a larger ambition frontier.
Why the final 10% of work — judgment, verification, integration, and accountability — is becoming the job in the AI era. AI is not just an efficiency engine; it's an ambition engine.
The best AI agents are not the most general ones. They are the ones designed around a specific objective — with the right tools, system prompt, skills, memory model, and operating environment.
After a decade building data infrastructure — from data scientist to CTO — I got tired of stitching together 5+ tools that don't talk to each other. So I built Seeknal — one CLI for pipelines, features, metrics, and AI analysis.