From the Blog
Notes on building with AI
Practical lessons on AI, automation, and shipping software from inside the venture studio.
The Difference Between Building Fast and Building Right
Fast and right aren't opposites. A real bug that shipped silently, and a feature deliberately left unbuilt until it was needed, show what actually decides which one a piece of the system needs.
What International Students Actually Need From an AI Tool (Not What Founders Assume)
The founder instinct is a slicker application tracker. The actual fear looks different. What building ScholarPilot revealed about the gap between assumption and signal.
Why Most AI SaaS Products Fail Before They Ship
Operator never shipped. It's parked. Here's the honest reason why, and what it revealed about why most AI SaaS products fail before they ever reach a live URL.
Building a Company on a Wyoming LLC From Outside the US
An LLC filed from outside the US in days. Being ready to take a customer's payment through it — still not finished. The honest account of what this setup actually involves.
AI Agents vs Automation: What's Actually Different
One task in the portfolio runs the same steps every day without variation. Another reasons through a different path each time. Here's the real difference between automation and an agent, and how to tell which one a task actually needs.
How to Validate a SaaS Idea Before Writing a Line of Code
ScholarPilot had real signups before it was built. The other eight products in the portfolio didn't. Here's what that difference actually proved, and the honest limit of what it proved.
How to Price a SaaS Product When You Have Zero Customers
Pricing research without a real customer to react to isn't research — it's avoidance wearing a spreadsheet. Here's the actual reasoning behind ScholarPilot's four pricing tiers, with zero customers to validate any of it.
Claude Code vs Hiring a Developer: A Real Comparison
Claude Code has been the entire engineering team behind nine live products. Here's the honest version — including the failures — of what that actually replaces and where it doesn't.
What I Learned Shipping Seven Products and Getting Zero Customers
Building was never the hard part. Here's the honest breakdown of what seven shipped products and zero paying customers actually taught me about avoidance disguised as progress.
AI Venture Studio vs Accelerator vs Agency: What's Actually Different
Three labels, one question people keep asking. The honest breakdown of what separates a venture studio from an accelerator and an agency — and which one actually describes running seven AI products alone.
How I Use Claude Code to Run Seven AI Products Solo
Seven live AI products, one person. The honest version of how that works: Claude Code used like an engineering hire, and a hard line about which decisions never get delegated.
Building an AI Venture Studio from Bangladesh
How FrictionLab turns a one-person operation into a portfolio of AI-first companies — and the systems that make it possible.
Agents That Actually Ship Work, Not Demos
The difference between a flashy AI demo and an agent you can trust in production comes down to a few hard design choices.
Prompt Libraries as Product Infrastructure
Why treating prompts like reusable, versioned assets is the unlock for shipping AI features fast and consistently.
Automation That Removes Real Friction
Not all automation is created equal. The highest-leverage automations remove the work nobody should be doing by hand.
Shipping a SaaS MVP in Weeks, Not Months
A practical playbook for taking a SaaS idea from blank repo to paying users without drowning in scope.
AI for Study Abroad and Admissions
How AI is reshaping test prep, university matching, and the admissions journey for students worldwide.