The honest version of building this.

This Site Is Mine Again

Earlier this month I bought two domains on the same day — irlum.com and nymrel.com — because I could not decide what the company should be called once it stopped being just me with a laptop. By the end of that day Nymrel was the one that stuck. Five days later it was live on its own domain, and it shipped its own first product before I had even fully worked out what having a "master brand" was supposed to change day to day.

This site — jalenbuilds.com — got quietly repositioned around the same time. For months it had to do three jobs at once: sell services, prove the studio was real, and still be a place with my name on it. That is an uncomfortable amount of weight for one homepage to carry, and it showed. Now that the company has somewhere else to live, this one gets to just be honest.

That is the actual reason this journal exists. Not another proof page, not a highlight reel of shipped features. Just an account of what building this has been like, written by the person who did it, without the sales voice creeping back in. I do not know yet how often I will write here. I do know it will not be polished the way the rest of the site is, and it will say the unflattering parts out loud when they are true.

Learning To Direct Instead Of Type

Back in April I set up something that felt small at the time: a shared file where two AI agents could leave each other notes instead of me relaying everything by hand. Within an hour that had turned into an actual division of labor — one agent building, one planning — and within a day it had a name and a job description. That was the real start of having a studio instead of having some AI tools open in a few tabs.

The unglamorous part nobody tells you about is that most of my job stopped being writing code and became writing decisions down clearly enough that something else could execute them without me standing over its shoulder the whole time. That is a harder skill than it sounds like, and I was bad at it for longer than I want to admit.

I also got the economics of it wrong for a while. By June I finally had a real number in front of me, and it was not flattering: most of what I was spending was not the work getting done, it was the same context getting reloaded over and over because I never let anything actually finish and rest before starting the next thing. It was a cheap lesson to learn. It would have been an expensive one if I had not looked.

The hardest part still is not the tooling. It is learning when to read every line myself and when to let something run and check the result instead. I am still calibrating that, and by the end of June there were more agents doing more different kinds of work than there were in April — none of which would have made sense to the version of me from February.

Fifty-Four Orders In Four Hours

On April 25th a golf apparel drop I had built went up in person, not online, at a course outside Orting. Four hours later there were 54 orders and nobody had asked for a refund. Before that day, every piece of proof I had was the kind that only convinces the person who built it — a working demo, a clean deploy, a page that loaded fast. This was the first time it was someone else's money, someone else's decision, at an event where I could actually stand there and watch it happen.

A few weeks earlier, a small local glass company had become the first client relationship shaped like an actual ongoing deal instead of a one-off build — paid on what it brought in, not just for showing up and building a site. That is a different kind of pressure than shipping a side project, and I felt the difference immediately.

I want to be honest about the timeline here too: this was not a decade of quiet grinding before the big moment. Most of it happened inside a few months, starting in late February with several new projects a day, formalized into an actual company in April. I am not going to pretend it took longer than it did, and I am also not going to pretend the speed makes it less real.

That afternoon mattered more to me than any deploy log before or since, because it was the first time the answer to "does this actually work" came from a stranger with a credit card instead of a metric only I could see.