Slow Down the Thinking, Speed Up the Setup

I’ve been building a lot of things recently. A lot of side projects. A lot of small projects.

And what I realized pretty quickly is that I wanted two opposite things at the same time. I wanted to speed up the clerical work of starting a new project… but actually slow down the thinking that goes into it.

What I was doing before was kind of backwards.

I’d get an idea, open my terminal, start setting things up—making a folder, naming it something random, initializing git, setting up configs, whatever. And as soon as that was done, I’d jump straight into Claude and start asking it to write code.

The problem was: I hadn’t really thought the project through yet.

So two weeks later, I’d have a bunch of generated code, a lot of files… and this vague feeling of, wait, what was I even trying to build here?

I was rushing the thinking and spending time on the boring setup.

I wanted to flip that.

I wanted the setup to be almost instant, so I could use that extra time and attention to slow myself down and actually think.

That’s why I built this little CLI called Mint.

Mint-CLI

Now when I’m starting a new project, I open my terminal and just type mint and the name of the project. Or I can just type mint and it’ll ask me what I’m trying to build.

Based on that, it suggests a few project names using DigitalOcean’s inference API — not because the name matters that much, but because it helps me get past the blank-start moment and move on.

Once I answer that, it does a few things automatically.

It creates a new project directory. Inside it, it creates a docs folder. And inside that, it creates a few files I always want at the start—like a WHY.md file and a decisions.md file.

It also initializes git, changes into the directory, and opens those files for me.

So within a few seconds, I’m not staring at an empty folder anymore. I’m staring at a file that’s asking me to think.

And this is where I intentionally slow down.

Before I write any code—before I even talk to Claude—I open up WHY.md and answer a few simple questions.

• Why am I building this?
• What am I not building?
• When is this done?

That second one—what I’m not building—is the most important for me.

It’s basically permission to ignore things. It keeps the scope small. It stops the project from quietly turning into something bigger than it needs to be.

And the nice side effect is that when I do bring Claude into the picture, it already has context.

It knows what I’m trying to do. It knows what I’m explicitly not trying to do. And it knows when to stop.

So instead of rushing straight into code, I spend maybe a few minutes thinking. The setup takes about 30 seconds. The thinking gets the time it actually deserves.

And then I’m ready to build.