Notes from the intersection.

AI should assist decisions, not make them

The most useful AI products I've built don't replace human judgment — they sharpen it. When I started designing DecisionSphere, the temptation was to let the model output "the answer". Users didn't want that. They wanted to see the decision: the options, the risks, the trade-offs, laid out visually.

That's the pattern I now design for everywhere: AI as an analyst that never sleeps, presenting evidence — while the person stays the decision-maker. It builds trust, and trust is what makes people actually use AI tools twice.

Micro-SaaS is the new startup school

A small, focused product like Scaler Profile forces you to learn the entire stack of a company — product, engineering, pricing, onboarding, support — on a scale where mistakes are cheap and lessons are fast.

For a student founder, that's a better education than any accelerator pitch deck. Build small, charge early, listen hard. The skills compound into whatever you build next — for me, that's QIXARC.

Automate the boring, ship the important

Every hour a business spends manually answering the same WhatsApp question is an hour not spent on growth. That gap — repetitive work that software can absorb — is exactly where automation products win.

The rule I follow: automate the predictable, never the personal. Customers should feel the speed of automation and the warmth of a human, in the same conversation. Products that get that balance right keep their users.