By Xiangan He.
While running my dev studio, I was curious to see what building my own product would look like. Up to that point, everything had been a service contract. In my first two months, I shipped Slopless, a security auditing tool that I bootstrapped to $20k in annual recurring revenue. I joined the Spring 2026 AI for Education Fellowship to start another product sprint and explore opportunities in education.
My first instinct was to build for younger students. I made a math tutor for middle school students and got my first paying customer, at a few dollars a month, within a week. It was fun to build and demo, and I moved faster than I had in my first product sprint.
But I ran into two challenges: the token costs didn’t make sense for my use case, and I was lukewarm about middle school math. While finishing my master’s at Boston University, I had experienced a much more personal frustration: studying for the CCSP exam. I had to pay thousands of dollars for PDFs and teach myself. That was a problem I understood, with a bigger pain point and more room to charge for solving it.
I started with a “Wizard of Oz” pilot: I did the teaching myself before building software to do it. I mentored BU students through a national competition. With a team one-third the size of the previous year’s, we placed in the top three. I then took that preparation process and built it into a console-like environment for real-time testing.
Once that started working, I wanted to see whether the approach could transfer to another exam. Anthropic’s CCA certification had just come out. Could I turn professional certification self-study into an immersive experience for developers too? That experiment led to my first pilot for the agent exam. The project became Whetstone.
What I learned from building Whetstone
The journey changed how I decide what to build and how I test it. These are the lessons I’d take into my next product sprint.
Start with a problem you know well enough to care about
I could build a math tutor and find someone willing to pay for it. But the costs, the outreach, and my own interest made it hard to see how I would keep going. My experience studying for a cybersecurity exam gave me a much clearer sense of what was frustrating and what better preparation might look like.
My dev studio also put me close to people who needed that help. I subcontracted for larger agencies, worked with engineers, and helped teach developers to use AI. When one agency needed its engineers to prepare for the CCA exam, I could recognize the opportunity and act on it. The first goal was concrete: help ten engineers prepare to pass.
Teach it by hand before turning it into software
A workshop gave me a way to learn before committing to an app. Working directly with learners helped me understand the preparation process and decide which parts to bring into the product.
It also made the next step smaller. I didn’t have to figure out how to serve ten thousand users immediately. I could help one group, see where they struggled, and test whether software could make that experience useful for more people.
Give early users a goal they can test against
The early pilot had bugs. Users told me where it fell short, and I kept improving it. I remember feedback averaging around 7.5 to 8 out of 10, with some users saying the tool had helped them pass their exams.
Having an exam to prepare for gave those conversations a focus. Was studying easier? Did people feel more confident? Were they learning more, or spending less time switching between materials? Those questions helped me understand what people valued and what still needed work.
Make finding demand part of the build
I kept demoing Whetstone at community calls, to agencies I’d worked with, and to colleagues. Each conversation helped me find out who needed it. I was lucky that the certification opportunity appeared when it did, but working in that community and sharing what I was building helped me respond.
I also learned to keep investigating opportunities I initially doubted. I hadn’t expected certification prep to be a promising direction. Earlier, I hadn’t expected businesses to buy Slopless either. Both experiences gave me a reason to test my assumptions before ruling something out.
For my next sprint, I’ll ask: what’s the smallest amount of work that could get momentum going? It might be a conversation, a demo, or a workshop. Demand is king, and learning to generate it is a skill I want to keep building.