The Experiment I Was Hoping Would Fail

For about two months, over the holidays and into January, I did the same small thing every few days without really noticing I was doing it.

I'd open Twitter or LinkedIn, scroll through the people I follow in tech, and watch someone rave about Claude Code. How they'd built a whole product over a weekend. How much better it had gotten in just a few months. How it had quietly changed the way they worked.

And every time, I'd nod, feel a little flicker of something, and keep scrolling.

I told myself I was being disciplined. I'd picked my lane. I had a product, a stack, a plan. The last thing a stretched solo founder needs is to go chasing a new shiny thing every time it trends.

That was the story I told myself, anyway. Looking back, the truth was a little less flattering.

The lane I'd already picked

Here's the part I wasn't admitting.

When I started Coachstack a couple of years ago, I wasn't an engineer. I'd done a couple of coding bootcamps, I could put together a basic website, but I had no computer science background and no business pretending otherwise. So when it came time to build the first version, I looked at what was out there and bet on Bubble, a no-code platform.

I went all in. I took the best-rated beginner course I could find. I built the wireframes in Figma. I got a real way into building it myself, then hit the usual wall somewhere around the Stripe and Calendly integrations, and realized I needed help.

The agency quotes I was getting ran from $20,000 to $50,000 and up, three to six months each. For a bootstrapping founder, that wasn't a number I could say yes to. Then I found a developer online who could build my MVP for $8,000 in three weeks. I had a Honda Accord sitting in the driveway that we barely drove. I sold it, and that car became the first version of Coachstack.

That version ran a pilot with real coaches last year. It got good feedback. I built on top of it for a year and a half.

So you can see the problem. By the time everyone started shouting about Claude Code, I wasn't a neutral observer. I had a sold car, eighteen months, and thousands of hours riding on the bet I'd already made.

When you're that deep into a decision, you don't actually want to find out there was a better way. You want to be right. And the easiest way to stay right is to never run the test.

I think that's what the scrolling really was. Not discipline. Self-protection.

The day I finally tried it

President's Day this year was a Monday, and the office was empty. I'd been carrying this low-grade itch for weeks, and I finally decided to scratch it. One day. One experiment. I half-expected, and if I'm honest half-hoped, that I'd spend a few hours with Claude Code, find it overhyped, and go back to my lane with a clear conscience.

I set it up as a fair fight. I had Claude help me write a proper product requirements document for a simple marketing website for an executive coach, the same kind of spec I used to hand to my agency. Then I gave that exact document to two tools, side by side: Claude Code, and Codex, the equivalent from OpenAI. I connected my GitHub, set up hosting, and let them both run.

Five minutes

Both of them built the website in about five minutes.

Not a template I'd spent an afternoon bending to fit me. A fully custom site. Custom brand, custom copy, custom images, all of it generated from the document I'd written.

I sat there and stared at the screen, and I had the feeling I think a lot of people have had this past year. The quiet "oh no." Oh no, this changes things. Oh no, the thing I was hoping would let me off the hook just did, in five minutes, what I'd been paying thousands of dollars and waiting months for.

The experiment I ran to stay in my lane was the thing that pulled me out of it.

What happened next

I've used Claude Code every single day since.

It started small. Simple one and two page sites, built from a document and a conversation. Then I found a tool called Paper.design that let me generate visual wireframes in real time, so I went from staring at a terminal to actually watching the thing take shape, the way I used to in my product manager days. Then I built the full website for Solo Founder Coach. Then a newsletter system to go with it. Then a tool to help write the newsletter itself.

None of it was a master plan. I just kept following the thing that was working. And somewhere in there, without ever deciding to, I'd built the foundation of the new Coachstack.

You already know how that part ends. A couple of weeks ago I wrote about sunsetting the Bubble app, the eighteen month build, the sold car, all of it. What I didn't say in that issue is where the nerve to do it actually came from. It came from a holiday Monday I almost spent doing nothing, running an experiment I was quietly hoping would fail.

The takeaway, if you want one

We talk a lot about sunk costs in terms of what we've built. The money, the hours, the thing we can't bring ourselves to walk away from. I wrote a whole issue about that one.

But there's a sneakier version, and I think it's the one that gets most of us. It's the sunk cost in being right. Once you've made a big, expensive, public decision, admitting there might be a better way doesn't just cost you money. It costs you the story you've been telling about yourself. So you quietly avoid the test. You scroll past the thing. You call it discipline.

If you catch yourself dismissing the same idea over and over without ever actually trying it, it's worth asking which one you're really protecting: your conclusion, or your ego.

The good news is that the test is usually cheap. Mine was one quiet Monday. Block out the day. Run the experiment. Let reality cast the deciding vote, even when, especially when, part of you is hoping it won't.

One last note. This whole issue was about blocking out a single day to try something new. Next Saturday, I'm blocking out a very different kind of day, the Fourth of July, and there won't be a laptop anywhere near it. So no edition next week.

I'll be back on Saturday, July 11th. If you're stateside, enjoy the long weekend.

See you on the 11th.

Peter