I don't just spec AI — I ship it, write about it, and build with it. Twenty years of enterprise platforms, and I still open the editor myself.
Now Writing, building, and talking to teams about their next AI product. Coaching and building careers.
The Mahabharata isn't just a story of war — it's a manual for choices. The AI PM Gita translates 18 parvas into 18 lessons for building AI products that create impact, earn trust, and endure — each one tested against real production failures.
Duryodhana was offered peace three times and refused three times — each refusal a distinct sin: hype without depth, speed without safety, and features without purpose. Every failed AI product walks through at least one of these gates.
"Failed products are rarely built in ignorance. They're built by teams who knew — and shipped anyway."
Abhimanyu learned to enter the Chakravyuha but never heard how to exit — a frame for AI features brilliantly designed for entry, with no way out. Introduces the four failure modes and the Three Rings: Signal, Threshold, Exit.
"The model learned to please, not to help. Nobody files a bug report that says 'the AI was too agreeable.' That's what makes sycophancy the most dangerous failure of all."
Field notes on AI product management, GTM, agentic systems, and the business behind the hype. Many topics, roughly weekly — the AI PM Gita chapters publish here too.
I build with the same tools my teams use. A PM who can't prototype can't push back on an estimate, de-risk an agent architecture, or tell a demo from a product.
I've hired, promoted and let go of product managers for twenty years. I coach the people on the other side of that desk — career decisions, product craft, and the leap into leadership — and I help US-headquartered companies stand up product hubs in India that actually own outcomes.
The other things I make when I'm not making software.
