Part 2 of the PotatoLab series

I didn't realize I had a broken definition of "done" until I tried to write one down.

A few months ago I sat with a blank doc and asked myself: what would it actually take for this project to be finished? I couldn't answer it. Not because the project was complicated. I just never defined the finish line. I had just been running.

That's the thing about "done" when you're a perfectionist engineer. It doesn't really exist. There's always a refactor that would make it cleaner, a feature that would make it more useful, an edge case that would make it more correct. "Done" becomes a horizon. You move, it moves.

So in practice, my projects were never done. They were just abandoned. Quietly. No announcement. Just a private repo that I stopped pushing to.

What changed wasn't motivation. It was the contract.

When I started PotatoLab, I forced myself to write down what the project needed to do to be real before writing a single line of code. Not a spec. Not a roadmap. Just a small, honest list.

This needs to do X. When it does X, it ships.

That's it. That's the whole contract.

The first time I did this it felt almost too simple. But simple is what made it work. "Done" was now a specific thing I had already agreed to. I couldn't quietly move the goalpost. I signed the contract with myself before I got attached to the thing I was building.

Day-to-day, here's what actually shifted

Before, I'd open a project and ask what should I build next? Now I open it and ask am I done yet? That's a different question. It points toward an exit instead of deeper in.

When I catch myself wanting to add something that wasn't in the original list, I write it down separately. Not as a backlog, more like a "maybe next experiment" note. It scratches the itch without bloating the current thing.

I also stopped calling unshipped work "in progress." It's not in progress. It's in limbo. Progress means moving toward something. If I can't tell you what done looks like, I'm not progressing. I'm just busy.

The graveyard didn't close. But it stopped getting new residents.

I still have the old projects. I'm not going to pretend I finished them or that any of this makes them okay. They're still there.

But since I started writing the contract first, I've shipped three experiments. None of them are big in scope. But all three of them exist outside my laptop, which means all three of them kept their promise.

In the next posts I'll go through each one. What it does, how it works, and what I learned building it small on purpose.

That's what done means now. Not perfect. Not scalable.

Just the small promise, kept.

In Part 3 I'll close out the why before getting into the what. Each experiment, a name and a reason. Then the real posts start.

PotatoLab is at potatolab.cc. Lumpy things, shipped.