When You Can Build Almost Anything, What Should You Build?

I didn’t want to pay for a workout app subscription. I was, however, prepared to consider running a marketplace for personal trainers.

There were a few steps between those two positions, and each seemed reasonable at the time.

I’d been using Hevy for my weight-training sessions. I wanted to track cardio as well, and I thought I could build an app that suited my workouts. After roughly sixteen years as a software engineer, making my own version of something is an option I take seriously. AI has made it easier to act on that option.

I was also looking for a personal trainer. I hadn’t found many marketplaces that helped with that search. Perhaps I could combine the ideas: a workout tracker with a way to find a trainer. It would cover a need I had, and it might become something other people used.

I thought I could grow it slowly over time.

That last part made the idea feel manageable. I wasn’t proposing to drop everything and start a fitness company. I could get the software going, use it myself, and see where it led.

So I asked AI to build it.

The difficulty appeared when I considered what would have to happen around the application. I would need to recruit trainers. I would need to work out how to vet them. People looking for a trainer would need a reason to come to the marketplace, and trainers would need a reason to participate.

These were essential parts of making the thing useful. Someone would have to do them even if every feature worked perfectly.

And I needed one trainer.

That was when the size of the proposed solution became difficult to reconcile with the size of my problem. Even taking the idea far enough to satisfy my curiosity would mean work I wasn’t prepared to add to my existing commitments.

One of those commitments was Zaplink. I had decided to build and grow it this year. The time I was considering putting into the fitness idea was time I could have spent on that.

The fitness project never launched. I asked in groups for trainer recommendations, and I found an existing app that let me record both weights and cardio.

There wasn’t a dramatic failure or an expensive lesson. I stopped. But the sequence interests me because the question that got me started was so much easier to answer than the questions that eventually made me stop.

Could I build it? Yes, I thought so.

Did I want the work that would come with it? Once I looked more closely, no.

I’m building Estateflow alongside client engineering and fractional CTO work, and I’m trying to focus on fewer things at a time. That gets complicated when a new idea arrives with a plausible explanation for why it won’t require much attention.

The software can be built quickly. The first version can be small. Growth can happen gradually. Each of those things may be true. Together, they can still describe another business that needs someone to care about it for a long time.

Building is familiar and satisfying work for me. I can think through how something should function and start making it happen. The result is visible. There’s a degree of control.

Recruiting trainers would have required a different kind of effort, in a field where my immediate experience was being a prospective customer. I understood my frustration. To evaluate the business, I would also have needed to understand how trainers found clients, what they struggled with, and whether a marketplace could help enough to earn their participation.

I’d have needed a way to reach both sides. And somewhere in that arrangement, someone would need to be willing to pay.

My own need was a useful clue. It left most of those questions unanswered. In fact, part of the reason I had started thinking about the tracker was that I didn’t feel like paying for the app I’d been using. That isn’t evidence against a fitness business, but it’s hardly evidence for a paying customer base either.

I hadn’t worked out those answers. I had worked out enough about the commitment to decide I didn’t want to take it on.

I think that’s an important distinction when choosing among ideas. I don’t need to establish that an opportunity is bad before deciding it doesn’t belong on my list. Someone with access to trainers, a better understanding of their work, and an interest in running that business might make a different decision.

For me, the relevant comparison was Zaplink. I had already committed to giving it a chance, and I was trying to reduce the number of things competing for my attention.

That doesn’t prove Zaplink is the better business. An existing commitment isn’t a guarantee of success, and I don’t think it should become a reason to ignore new evidence. I’m still figuring out when persistence is useful and when a change of direction is justified.

But adding another project wouldn’t have resolved that uncertainty. It would have added another set of questions to answer.

The next time I consider an idea, I want to look further into the work it implies. Who would I need to understand? How would I reach them? What would they need from me after launch? What would receive less attention while I figured those things out?

AI is making the initial build a smaller part of that decision. My ability to distribute, support, maintain, and keep thinking about the result remains limited.

A workout tracker for my own use would have been a much smaller commitment. The marketplace meant attracting people, earning their trust, and taking responsibility for a service. That was the version I had to compare with the work I’d already chosen.

I wanted somewhere to record my workouts and a way to find a trainer. I could meet those needs without becoming responsible for meeting them for everyone else. For this idea, at this point in my life, that was enough.