Blog

One unchanged pipeline — load, profile, test, model, explain — run across ten open datasets. Each section pairs a plain-language business read with the full technical report.

The reason serious data analysis stayed out of reach for small businesses was cost, not need. AI agents change the arithmetic — and open the door to the millions of companies that were priced out.

Research, write, send, check, recover — an AI agent ran my whole cold-email pipeline using Claude, Resend, and a domain I already owned. Cost: zero.

Running a Recency-Frequency-Monetary breakdown on an open orders dataset — and what it says about where revenue really sits.

The unglamorous corner of India's job market that the big portals forgot — and why AI finally made it worth solving.

Build in public: share the learning, earn the trust, and the rest follows.

Synchronous, blocking code review is the biggest tax on shipping. Make review asynchronous — and trust the tests.

Why 'one skill per use case' is going to collapse under its own weight — and the three-tier model to build instead.

A room full of people thinking out loud, all at once, is the worst environment for thinking. Write first. Read first. Then meet.