Growth: Make It First, Then Make It Good

I always thought I procrastinated. There were lots of things I wanted to do and very few I actually started. An idea would go round in my head for weeks; I'd pile up notes and bookmark a pile of material, yet keep putting off the first step.

After using AI for a while, I found this got much better. Not because I suddenly became disciplined, but because the order in which I do things changed.

Underneath procrastination is perfectionism

Looking back, the things I put off had one thing in common: my expectations for them were all high.

I lean toward perfectionism. Before doing something, I always wanted to prepare fully first: get the direction clear, read all the material, pick the right method, and ideally produce something decent right from the start. Preparation itself feels like progress, but it also keeps raising the threshold for starting. The more I prepared, the more it felt like something was still missing; the more it felt like something was missing, the less I dared to begin.

Over time I came to see that often what I feared wasn't doing it badly, but that the first version would be too poor, too far from the picture in my head. As long as I didn't start, that perfect version stayed intact.

AI took apart the cost of starting

That is exactly the step AI changed.

Getting from nothing to a rough first draft used to take a great deal of time and energy; that stretch was the most painful and the easiest to give up on. Now when I have an idea, I can have AI rough something out first: the outline of an article, the skeleton of a page, the preliminary conclusions of an analysis. A few minutes later, there it is in front of me.

Of course it isn't perfect, and it's often quite rough. But once there's something concrete, my question changes from "should I start?" to "what needs changing?" The first question keeps you hesitating; the second is easy to answer: this part is wrong, that part isn't enough, this section needs rewriting. Changing something that already exists is far easier than making something from nothing.

Done matters more than perfect

This made me rethink the saying "done is better than perfect."

I used to think it was an excuse for people who didn't care enough. Now I think it's about order: it's not that perfection doesn't matter, it's that it can't come first. At the start it's a barrier; later on, it becomes a direction.

What I do now is use AI to make a version that can be looked at and used, then improve it section by section, raising it step by step. Each round only fixes the few most obvious problems, and then I look at the next round. Many things only slowly approach what I originally wanted by version five or version ten. Done the old way, most of them would have stayed at version zero, which is to say never started.

Act on it when you think of it

Another thing this change brought is that I've got used to acting on ideas as soon as I have them.

Before, when an idea came up, I'd weigh it in my head for a long time: is it worth it, has someone done it already, can I do it well. By the end of all that weighing, most ideas just drifted away. Now that testing an idea costs much less, rather than thinking it over again and again, I make a small version first and see. Once it's made, whether it's worth continuing is usually clear at a glance, which is far more reliable than reasoning it through in my head.

That's how my drive and follow-through have gradually improved. Not because I've become more persistent, but because each start has become lighter, so light it doesn't take much persistence at all.

What to watch once you speed up

Of course, this way of working also comes with things I need to remind myself of.

Easy to start also means easy to drop. Rough drafts come so fast that I sometimes start several things at once, each stalled as a half-finished piece. So I've set myself a simple rule: starting can be casual, but anything already started either gets improved to a deliverable version or is consciously set aside, not left hanging.

One more thing: the first version AI gives is only a starting point, and the judgment still has to be mine. It helps me get faster to the point of having something to improve, but what it should become and what counts as good still depend on my own understanding of the thing. The speed has gone up; the standard can't come down with it.

I used to demand perfection before I started. Now I make it first, then make it good, step by step.