The expensive part of building alone is not the hours

The plan is always the same. Fix the outstanding bugs, then sleep at one. It is a reasonable plan and I have made it many times.

What happens is that you fix the first one and it reveals two more. You fix those two and they reveal four. None of them are hard. Each is fifteen minutes. And at some point you look up and it is half past three and the plan expired two and a half hours ago without any single moment where you decided to abandon it.

That is the part people describe when they talk about building alone, because it is the part that is easy to describe. It is not the expensive part.

#The room

You are in a closed room at night looking at a screen. Nobody else is awake, nobody else is in the project, and there is no reason for anybody to ask how it is going. This is true whether the work is going well or badly, which means the work going badly produces no external signal at all.

I want to be careful not to romanticise this. The hours are a cost and they are self-inflicted, and I would not hold them up as evidence of anything except that I have not solved my own scheduling. The sleep is the part of this I would change.

#What is actually missing is disagreement

The obvious framing is that a solo builder is short of hands, and with AI assistance that framing is close to obsolete. I can produce more code in a night than a small team could have produced in a week two years ago. Labour is not the constraint.

What is missing is somebody who will tell you that you are wrong. Not in general — specifically, about this decision, with enough context to be right about it.

• There is no review before a choice becomes permanent. The first time a bad structural decision gets examined is when it starts hurting, which is usually weeks later and expensive by then.
• There is nobody who knows the codebase. Any advice you get is from someone who has never seen it, which limits how specific that advice can be.
• There is no calibration. You cannot tell whether a thing you spent two days on was worth two days, because nobody else has an opinion about it.
• Every mistake is discovered by you, at the worst moment, and there is nobody to share the discovery with.

The scarce resource is not effort. It is being contradicted by someone who knows what they are talking about.

#The substitutes, and what each one misses

You do replace some of it. The list is short and every item on it has a specific gap.

#The half that is genuinely worth it

I would not still be doing this if it were only cost, and the honest accounting has a real other side.

You learn everything, because there is nobody to hand a problem to. Every part of the stack is a part you have had to understand — the deployment, the database, the auth flow, the thing that broke at midnight. That knowledge is not distributed across a team where you own a slice of it. It is all yours, and it compounds.

Decisions are also instant. There is no meeting, no consensus, no negotiation about scope with somebody who wants a different thing. When I decide at midnight that a section is wrong, it is rebuilt by two, and the total cost of that decision was the rebuild.

And there is a specific satisfaction in a working system that nobody else touched. It is not a better system than a team would have built. But you know exactly why every part of it is the way it is, which is not a common thing to be able to say.

#What I would tell somebody starting alone

1. Assume your structural decisions are unreviewed, because they are. Give the ones that are hard to reverse more thought than they seem to deserve.
2. When you ask an assistant for a review, ask it to argue against the approach, not to improve it. You will get a different and more useful answer.
3. Write down why you did things. You are the only person who can ever explain them later, and you will not remember.
4. Find one person who will look at it occasionally, even if they are not technical. Being asked to explain a decision out loud does most of the work that a code review does.
5. The 1am plan is a fiction. Decide what time you actually stop and stop then, with bugs outstanding.

That last one I still do not manage. But the point of writing the cost down is being able to see it, and the cost of building alone is not the nights. It is that nobody is going to tell you the thing you most need to hear, and you have to construct that for yourself, deliberately, every time.