Writing / Blog

Thinking in public.

AI, software, entrepreneurship, career, education — long-form thinking with excellent typography. Built to grow for years.

What building AI agents taught me about thinking

Agents fail like interns, not like code — and that changes how you design them.

AI • Jan 2026 • 6 min read

Exploring widely looked slow. It turned out to be the shortcut.

Flutter, marketing, chatbots, CS theory — why exploring widely made me a sharper, more pragmatic builder.

AI leverage plus tiny teams: notes from building as a solo founder in Bangladesh.

AI automation + SaaS + distribution: notes from building small, useful products as a solo founder in Bangladesh.

Pointers, problem sets, and the week I almost quit.

The hardest course I've finished — and the one that changed my relationship with difficult things.

Lessons from shipping SupportIQ for a real nonprofit team.

Grounding, source links, and graceful handoffs — what made a nonprofit chatbot actually trustworthy.

Uncertainty is the whole game in AI. Math is how you see it.

Stanford's Probability for AI convinced me: the math isn't a detour, it's the map.

Shorter, rougher thinking lives in the Research & Notes archive and the AI Lab — updated whenever an experiment teaches me something.