Finishing is a skill
By Flavio
Tagged: Essays
How I get things done: the three states every project is in, why everything I ship is version one, the seven-day rule, why I ship lots of small experiments, and why deleting or automating something counts as finishing it.
Between November 18 and November 21 I launched 32 Mac apps, eight a day.
This site got 3,921 commits so far this year. In all of 2025 it got 434.
People ask me how. They expect a tool, a framework or a secret schedule. The real answer is less exciting: at some point I decided that finishing is a skill, separate from having ideas and separate from being good at the work, and I started practicing it on purpose.
This is how I think about it.
Every project is in one of three places
At any moment, anything I’m working on sits in one of three states.
- Figuring it out. I don’t know yet what it is or how to make it.
- Building it. I know enough to make progress, and I’m making it.
- Wrapping it up. It works. Now it needs to leave my hands.
That’s all. When I feel stuck, the first thing I do is ask which of the three a project is in. The answer usually tells me what’s wrong.
Most stuck projects are in the first state and have been for weeks. Some are in the third state and pretending to be in the second, because wrapping up is boring and building feels productive.
Notice the last page of that chart. “Done” has more than one door, and that’s the most useful idea in this whole post. I’ll get there.
Everything I ship is version one
I treat everything as a draft. Every app, every post, every course.
This sounds like lowering the bar, but it’s the opposite. Calling something a draft takes away the pressure that stops me from finishing it. A draft can go out. A masterpiece can’t, because it’s never ready.
Software makes this easy because it has version numbers. Flashcard PDF Printer, an app for teachers, shipped as 1.0.0. Then came 1.0.1, and now it’s at 1.2.0. Every buyer got each version through the app’s updater, so shipping 1.0.0 early cost them nothing.
Posts work the same way. Many posts on this site have an updated: date in their frontmatter, because I went back and made them better once they were already out and useful. A published post I improve later is worth more than a perfect post sitting in a folder.
There’s no polishing phase
I don’t write first and edit later. I edit while I write, then I stop.
A separate “polish” phase is where projects go to die. It has no clear end. There’s always one more detail to fix, one more screenshot to retake, one more paragraph to tighten. And it comes right at the moment the project is almost done, which is when my motivation is lowest.
So polish happens inside the work, a little at a time. When the work is done, it ships.
Perfection is also boring. The thing I’d remove from a project to make it perfect is usually the part that made it mine. Rough edges are fine. Unfinished things aren’t.
Act like you know what you’re doing
You don’t need to know how to do something to start doing it. Acting like you know is close enough, and you’ll know soon.
I don’t have a Windows PC. Footage Ferry for Windows was built entirely in a Windows virtual machine on my Mac, and it shipped 1.0.0 on November 13. If I had waited until I felt like a Windows developer, it wouldn’t exist.
The confidence I mean is about yourself, never about fooling the people you sell to. Stop asking yourself for permission, start, and let the work teach you what you need.
The seven-day rule
If an idea hasn’t moved in a week, I let it go.
It leaves my list, with no “I’ll get back to it”. If it was a real idea, it’ll come back on its own with more energy.
This rule exists because an idea that sits there doesn’t cost nothing. It costs attention. Every time I see it I feel a little guilty, and that guilt leaks into the things I’m actually doing.
I’m not perfect at this. Right now there are 75 draft posts in this site’s repository, and 35 of them were started before November. Most of them should go.
Deleting something is finishing it
This is the idea that changed how I work the most.
“Done” doesn’t only mean shipped. Throwing something away is also done. So is deciding a thing should stop existing.
Some of what I finished this month by removing it:
- On November 10 I retired six AI courses and turned their lessons into posts.
- On November 11 I deleted the Paddle purchase webhook, because none of the courses it handled were sold anymore.
- In August the old Netlify deployment of this site was decommissioned, along with its functions.
- I deleted 50 work notes for tasks that were already finished.
Every one of those freed me from something I had to keep in my head. That’s the same relief I get from shipping.
Automating something is finishing it
There’s one more way I finish things: I make them stop needing me.
Most of the work around a launch here is automated. One release command bumps the version, builds the app, signs it, has Apple notarize it, publishes the release and checks that the in-app update finds it. The social card for this post was rendered from code. Every app screenshot gets its gradient backdrop from one script. Launches are planned in a queue file, eight a day, and each one goes live at its scheduled time without me touching it.
Every one of those used to be a task I did by hand, every single time. Writing the automation is a project like any other, and when it’s done the task is done too, permanently. It stays done no matter how many times it runs.
That’s the best kind of done I know. A shipped project is finished once. An automated task is finished every time it runs.
Ship lots of small experiments
I don’t bet everything on one big project. I ship many small ones and watch what sticks.
The 32 apps of this week are a good example. Each one does one job, like taking the audio tracks out of a batch of videos, or making ZIP files without the Mac housekeeping files inside. There’s no way to know which of them people will want until they’re out.
Small is the point. A small experiment gets to done in days, and done is the only state where it can tell me anything. Most of them won’t stick, and that’s fine, because by every rule in this post they’re still finished. The ones that work get more of my time.
Automation is what makes this possible. Eight launches a day would be impossible if each one needed a day of manual release work. Because the boring part is already finished, I can spend my time on the part that’s different every time: the app itself.
Mistakes count as done
A launch nobody buys is a finished launch. A feature I built and then removed is a finished feature. I know more than before, and I can move on.
One of my apps launched as “Chats”. Halfway through the release I realized the name was too generic for anyone to find it, so I renamed it Answer Keeper before it went out. That was a mistake. It was also done, in the same afternoon, and the app shipped.
The only outcome I count as a failure is the one that never ends, the project that stays “in progress” forever because ending it would mean admitting something.
Publishing an idea is a down payment
When I can’t build something yet, I write about it.
A post is a smaller thing I can finish today. It gets the idea out of my head, it tells me if anyone cares, and it gives me something to point to when I come back to it. It’s not the whole thing, but it’s a real step toward done, and that’s a lot better than an idea in a notes file.
This blog is, in a sense, more than 2,000 finished thoughts. Each one was small enough to finish in a day.
The people building are the ones who are right
I weigh opinions by whether the person has made the thing.
Anyone can say an app should have more features, or that a post should cover more cases. The opinions that change my mind come from people who have shipped something similar and know where it hurts.
This applies to me too. My opinion about a project I haven’t started is worth very little. Ten minutes of building teaches me more than an hour of thinking about it.
Done is fuel
I don’t finish things to have finished them. I finish them so I can start the next thing.
Every project I close gives me back the attention it was using, and that attention goes into the next one. It’s the main reason my output went up so much this year. Yes, coding agents made the building faster, and I wrote about what I learned using them. But faster building with no finishing just gives you more open projects.
The 2026 bar is mostly finished things. Apps launched, courses retired, posts merged, old code removed, eight launches a day that each ended so the next eight could start.
There’s a risk in this, and I wrote about it in when more people build software: finishing a lot of things isn’t the same as finishing the right things. Done tells you which ones work. It can’t pick them for you.
Try it today
Pick the oldest open thing you have. A side project, a draft, a half-built feature, a course you started making two years ago.
Then finish it, one of three ways. Ship it as version one. Write a post about it. Or delete it.
All three count.
Want me to talk about your product? You can sponsor this site.