Stephie’s World · late editionNotes · by Stephie · October 2026
Building without permission
When I ran Little Bun, every software idea waited on an engineer and a perfect spec. Building with AI took away the most expensive roadblock, and with it the pressure to get things right before I started.
When I ran Little Bun, software always felt like the place my ideas went to wait.
I couldn't code. So anything I wanted to build, a tool or a page or a fix to a process that was quietly costing us hours, started the same way: I needed an engineer friend to help me make it. They had their own jobs. Their time was a favor, and favors run out.
That was the real limit. It wasn't my ideas. It was who I'd have to ask.
Perfect it first, then ask
Asking for someone's time changes how you think. Before I brought anything to an engineer, I felt like it had to be finished in my head. The idea had to be clear, the design had to be settled, and I had to know exactly what I wanted, because every change after that cost them another evening.
At least, that was my experience with the engineers I talked to. A half-formed idea wasn't something they wanted to pick up, and I understood why. Building the wrong thing is expensive, and rebuilding it is worse.
Design had the same problem. You did your thinking up front, locked it, and handed it off. Trying five versions of a screen to see which one felt right wasn't really on the table.
So I perfected things on paper that I had never seen work. Which is backwards. Most of what I know about a product, I learn by using it.
The roadblock is gone
Now I build with Claude Code, and the most expensive part of that old process is just gone.
I'm still not writing the code. But I don't need to. What I need is the right questions and the right frameworks: what is this actually for, who uses it, what has to be exactly right every time, and what can be rough for now. With those, I can learn whatever the project needs as I go, and build real software I couldn't have built a year ago.
Henry went from an idea to a working product in a weekend. A few years ago, that same idea would have been a pitch to a friend and a long wait.
Sculpting, not specifying
The part I love most isn't the speed. It's that I don't have to get it right first.
I can mold it. Build a rough version, look at it, change the layout, throw out a section, try the color again, ask why something feels off and fix it in the next ten minutes. The design gets better because I'm reacting to something real, not defending a picture in my head.
There's no one's evening I'm spending. No one getting annoyed that I changed my mind again. Changing my mind is the process now, and it's free.
That's what made the difference for me. It isn't only that more people can build. It's that building feels like making something again, the way cooking or sketching does. You start with a lump of clay and you work it until it's what you meant.
What's still on me
Of course, the work didn't disappear. What's important is still important, which is:
- Asking the right questions.
- Knowing which problem is worth solving.
- Deciding what the model should never be trusted with, like the margin math in Henry.
- Taste.
- Knowing when it's done, instead of endlessly reshaping it.
But that's the part I always wanted to be doing. Back at Little Bun, every idea needed someone else's yes before it could become anything. Now I can just start, and keep my hands in it until it's right.