A small, working library of AI products, tools, and experiments — shipped fast, learned from faster. Each product begins as a hypothesis. The lab keeps running.
Most companies plan a five-year march to a single answer. A lab does the opposite — it runs many small, honest experiments and lets the market decide which ones deserve compounding attention.
After eighteen years of working across strategy, business, and technology — inside boardrooms, product roadmaps, and go-to-market plans — the same lesson keeps returning: the best products feel obvious in hindsight. They emerge from someone who understood the workflow before the software was possible — and then, when the tools finally caught up, moved quickly.
That is what this lab is for. AI has collapsed the cost of building software to something a single motivated person can carry. So the question is no longer "can I build it" — it is "which hypothesis is worth testing next."
An interactive deal room for Series B+ fundraising — replacing static decks and spreadsheets with live, queryable investor materials. Ask questions of the model, not of the founder at 11pm.
Research and targeting for B2B sales teams — mapping market volumes, decision-makers, and competitive sources into a live intelligence worksheet. Prospecting that does the reading for you.
A new hypothesis is on the bench. Follow along on LinkedIn to be the first to know when the notebook opens.
Four phases — run once, then again, then again. The best products only look inevitable in retrospect; behind them is a stubborn loop.
Every deep hypothesis starts as a private irritation. Investor updates in Google Docs. Sales research on ten browser tabs. Someone's Tuesday afternoon — a task that has always been painful, and that a language model could now do differently.
Speed is the whole game. The prototype exists to be shown to five real users; the second version exists because those five people told you what was wrong. The lab does not build in stealth. It writes in public and edits under pressure.
Traction has one honest form: someone gives you money for the thing. Everything else — waitlists, upvotes, retweets — is noise dressed as signal. When a hypothesis clears this bar, the lab doubles the resource. When it doesn't, the lab closes the notebook.
A lab is a portfolio, not a monolith. Most experiments will be quiet. One or two will not. The role of the lab is to keep the option open long enough for that asymmetric outcome to arrive — and to have learned enough by then to recognise it.
The tools have arrived. In 2026 a single person with judgment, a laptop, and a coding assistant can ship what used to take a small team — and the wider market has quietly caught up to that reality. Small applied-AI software is now a category, not a hobby.
India is now the world's third-largest home for AI startups, and applied labs like this one are learning to ship for both local and global users at the same time. The economics have quietly become one of the most attractive setups in software.
Occasional writing on building, business, and the odd habits of working out loud. Longer pieces live on LinkedIn.
The case for shipping side projects alongside a full working life — and the specific things you learn that no single role can teach you.
After running AI workshops across multiple teams: the bottleneck is never the model. It is almost always the manager, the incentive, or the meeting nobody wants to change.
Pattern recognition, asymmetric bets, and why the best AI products feel obvious in hindsight — field notes from a working builder.
New products, mid-build notes, and the occasional wrong turn — shared on LinkedIn as they happen. No newsletter. No noise. If a product below fits a problem you have, the inbox is open.