Journal · Oct 22, 2026 · 한국어로 읽기
Ten days from the first line of code to launch
Arche, a one-person company building apps with an AI team · Part 1
It started with one clunky HTML file.
I really knew nothing. I just told the free version of Google Gemini I wanted to try making an app. It threw together a single file. Honestly, it was rough. But when I opened it, it opened. I tapped a button, and a simple feature worked.
That moment felt like magic. Then a thought followed: all those ideas I'd only ever turned over in my head, could I actually build them with this?
Half excitement, half doubt. That's how it started.
September 16, 11:21 a.m. The first line of code went in.
It was a fiction app. Today it's Niblore; back then it was called MyStory. Give it one line and the AI writes an episode; rewrite what you don't like and build your own novel. Before the day was over, I'd added a subscription button and ads. The plan from day one was an app people would pay for.
Two days later, September 18, 12:30 a.m. I started the second app.
It was called Random Explorer, a walking app that sends you somewhere you don't know yet. A day later I renamed it Wanderlure and cut features. Walking only, plus a compass. The first lesson of those ten days: cutting is faster than adding.
For several days after the 16th, I did almost nothing else. I lost track of time. I think that was the first time I really knew what flow feels like.
September 20: submission day
2:20 p.m. Final check on a real phone. Pass. I approved the submission.
Ten minutes later, the submission stopped.
A payment button that didn't work yet was still visible on one screen. Tap it and nothing happens. Small, but the kind of thing that costs days if review catches it.
The AI handling audits kept it short: hide that button and resubmit. About forty minutes lost.
At 2:56 p.m., build 49 went to Apple review. Wanderlure went in the same day, and the first entry went onto the company board. From then on, the AI agents handed work to each other through the board instead of in conversation.
Rejected, and the same day
That night Apple sent a question. The next day, the first rejection.
We fixed it and resubmitted the same day.
On the morning of September 22, MyStory passed review. A pass that came after a rejection. Honestly, that was the best moment of all. Wanderlure passed the next day. The AI agents left the launch button for me. That's when the company rule became clear: AI prepares, and a human presses the buttons you can't undo.
September 26. I pressed the launch button myself. It felt great. Both apps went live on the App Store.
Day ten.
And the next question
Ten days is enough to ship an app. That part we've shown.
But on launch day I didn't have an answer to the next question: who's going to find this and use it? That question is still open. The "no customers" from the prologue really started that day.
Next: one human and eleven AI seats. How the work gets divided.
— Arche