Decisions that outlast the deciding
The choices that turned out to be permanent were not the ones that felt important while we were making them.
In the opening weeks of anything there is a burst of decisions. Dozens of them, most made quickly, because at that stage there is nothing to slow you down. No customers, no data, nobody’s expectations attached to any of it. It feels like the cheapest moment you will ever have.
I have watched that from close range twice now, on two very different things both in their first year: an Ontario corporation registered in February 2026, and a marketplace co-founded five months after it.
A few of those decisions are still load bearing and will be load bearing for as long as the thing exists. The rest were forgotten within a month. I could not have told you which was which at the time, and I do not believe anyone can.
Importance and permanence are different axes
The decisions that felt important were the ones with a lot of discussion around them. What we were going to do, who it was for, what it was worth. Those got hours. Almost all of them have since been revised, some more than once, and revising them cost very little, because a position you hold in a conversation can be replaced with another position.
The decisions that turned out to be permanent were structural and mostly boring. What counts as one unit of the thing. Which side of a transaction the fee attaches to. Whether two records are the same record. Whether a thing that can happen twice is modelled as one row or two.
On the marketplace, contractors spend credits to bid on work that clients post for nothing. That is a shape, not a price, and the shape is early. A price can be changed on a Tuesday. What the credit is a credit for is baked into how every screen counts, how the balance reads, what a refund even means. Nobody sat down and decided that with an appropriate sense of occasion. It arrived as an obvious answer to a small question and then everything else was built while standing on it.
So irreversibility does not track stakes. It tracks how many other things ended up pointing at the decision, and that is a fact about the future, which is exactly the fact you do not have.
The overcorrection
Once I understood this I did the obvious wrong thing, which was to slow down and treat small choices as though they might be the permanent ones.
That is unlivable, and it is also self defeating. Deliberating over a hundred small decisions costs more than getting a few of them wrong, and it produces a thing that is late, which is its own permanent decision. The first year of anything is mostly a race to find out whether the idea is real. Spending it carefully is not a virtue.
The test I ended up with is not about how important a choice feels. It is: after this, will other things be built on top of it, and will they be built by people who never see this conversation. If yes, spend the extra hour. If no, choose in a minute and move.
That test is not reliable either. It just fails less often than my instincts do.
The part nobody tells you
The choices you are still free to change are the ones nobody used.
That sounds obvious written down and it is genuinely disorienting in practice, because it means flexibility is a symptom. Any part of a young thing that is still soft after a year is soft because nothing grew into it. Rigidity shows up exactly where the thing worked.
I have caught myself feeling relieved that something is still easy to change, and then realising what that relief was actually telling me.