89 commits for a chili plant.

All I wanted was a reminder not to let my chilis die over winter. One automated spreadsheet. Check the plants, log the watering, get a nudge if I had not done it in a while. That is it. That is all I needed.

Then someone said "you should use Supabase." So I did. Then "it needs to be on Vercel." Fine. Then "you will want a signup page." For a one person app. For my chilis. I should have stopped there.

I sent the link to the Mrs's brother. He is a graphic designer and app developer. This was a mistake. The feedback came back immediately - make it mobile responsive, add edit buttons, and it needs shared gardens so two people can look after one garden. The Mrs said she was not going to use it. I built the sharing feature anyway because the developer said it needed to be there.

Then obviously you need photos. How else do you track the progress of a plant. Then one of the Mrs's EMBA cohort asked about indoor plants. So I added indoor plants. Then seedlings came up. Then sowing schedules. What is in season. Watering reminders. Fertilising reminders. Weather lookups so the app knows if it rained and can adjust.

The loop looked like this. Saturday morning, coffee outside, open the app, water some plants, notice something missing or something that could be better. Laptop on the patio, work the fix, push it, watch it auto-deploy. Sip coffee. Repeat.

I watched every single step of this happen. I knew exactly what was going on. I kept going anyway. 89 commits later the chilis are fine and I have a full stack web app with shared gardens, indoor and outdoor zoning, seasonal planting guides, and a cohort who uses it more than I do.

Here is the thing though. This is not a cautionary tale.

The app is good. I had the time, I had the context, and I could see where it was going at every turn. I made the calls. Some were deliberate, some were just curiosity getting the better of me on a Sunday morning.

I did not write every line of this from scratch. I do not need to. I needed to know what I wanted, what was missing, and whether what came back was right. I needed to know how to drive. Not how to assemble the car.

A year ago this app does not exist. Or it exists half-finished, abandoned somewhere around commit 12 when the complexity outpaced my patience. Now it is market ready. Shared gardens, indoor and outdoor zoning, seasonal planting guides, weather integration, the lot. Built on a patio with a coffee. The only question now is what to do with it.

Scope creep is real. Looking back it is either funny or terrifying depending on which side of a deadline you are on. But the thing that crept into scope? It's good. It works. And it would not have happened without AI closing the gap between what I could imagine and what I could actually build.

That is the shift. Not faster. Not cheaper. Different ceiling entirely.