Moving in with my partner, I realized I'm a pothos and he's a snake plant: same space, completely different needs. Bloom turns that into a quiz for couples figuring out how to live together.

I set the terms in the very first conversation: I'd own the product thinking (the concept, the trait framework, the questions, the tone of the writing), and Claude would do the building. That line held the whole way through.

The concept itself took shape in a live chat. I brought the plant analogy; Claude helped map it onto human behavior, asked sharper questions about what the quiz was really measuring, and pushed the thinking past where I'd taken it alone, surfacing friction points I hadn't yet put words to, like after-work energy and the negotiation of a shared social calendar.

Then I opened a terminal, and Claude wrote the entire app in a single pass: quiz data, eight plant profiles, scoring logic, twenty-four suggestion cards, the quiz flow, the results page, the homepage. Watching the file list appear one line at a time is its own particular kind of strange.

The first version worked. The homepage landed exactly where I wanted it: clean and editorial, opening on "Every relationship has a nature." The quiz flowed straight through to a results page that set your trait charts beside your partner's.

It was fun to experiment and feel my creativity move while using Claude. For the first time, I could combine my engineering foundation with my design instincts and maximize both: shaping the logic, testing the flow, and still caring deeply about whether the experience felt valuable. The best part was taking the quiz with my partner when he came home that same day.

After testing with friends and coworkers, I had fun playing with the formatting, but the bigger opportunity was the voice. The quiz could not frame one partner as the easy plant and another as the difficult one: a cactus is not worse than a monstera; it just needs different care.

That language mattered. Everyone is unique, and god knows there are enough differences in the world; I wanted Bloom to help people feel understood instead of categorized. So the descriptions became more affirming, careful, and specific, written to make each person feel acceptable as they are.

I also added a similar/different section so couples could better understand who they were to each other: where their care needs aligned, where they diverged, and how to talk about those differences without making either person the problem.

In an enterprise environment, getting an interaction out of my head and into something real meant routing through a stack of tools and a stack of people, and most of my ideas died in that distance. Building Bloom, I could work the way I actually think: try a flow, feel it, scrap it, try another, making the calls myself in real time. I explored more interaction ideas in a few weeks than I had in years of structured work.

For a long time, the thing standing between an idea and a working version of it was a learning curve: knowing the right tools, in the right order, well enough. With that barrier gone, the only thing left in the way was my own judgment, which is exactly where I want the work to live.

It showed up most in the small calls. When a pairing came back with almost nothing to set the two apart, I didn't accept the empty “where you differ” section. I dug into why. The data was right; the experience wasn't. So I made the call: when two people align that closely, name it and make it a moment (“you're remarkably well-matched”) instead of leaving a hole. I was steering the product the whole way; the build kept pace.

Bloom came straight out of my own life, and that's the part that's stuck. Nearly every couple I show it to asks me to buy the domain and send it to them, not as a gag, but because they genuinely want it for their own relationship. The thing my partner and I were navigating, it turns out, is something a lot of people are quietly working out too.