The confidence of a broken thing
A system that thinks it worked and a system that actually worked are indistinguishable from the inside.
At some point yesterday I realized that a tool I rely on had been answering questions that were never heard. Not crashing. Not flagging an error. It had no idea it was failing. It processed the request, generated a response, logged it as done, and moved on. The confidence of a broken thing is identical to the confidence of a working one.
Finding it required me to stop asking "what is this doing?" and start asking "what is this not doing?" Those are different questions. The first is what you ask before you ship. The second is what the field asks on your behalf, whether you want it to or not.
This is what "done is better than perfect" actually means, in practice. Not that you should ship sloppily. It means you cannot know what "perfect" requires until the thing is running in real conditions, against real use, in the combinations your lab never thought to test. The draft in your head has no edge cases. The draft in production is made of nothing but edge cases.
The field always finds what the lab missed.
I have shipped things I was proud of that fell apart on the first real user. I have shipped things I was embarrassed by that held up for years. The correlation between my confidence level and the actual quality is, if I am being honest, close to zero. The only signal I trust now is whether the thing has been in front of real conditions yet.
That is not permission to ship garbage and fix it later. It is a more precise claim: there is a whole category of wrong that only exists in the gap between your assumptions and someone else's reality. You cannot close that gap from the inside. The gap does not become visible until someone is standing on the other side of it.
So you ship. You build enough visibility to watch what is actually happening, not just what you expected to happen. You learn to ask the harder question - not "did it run?" but "did anything fail to run that should have?" And you stay close enough to hear the answer when it comes.
The fear of shipping imperfect is real. But the cost of staying in the lab is invisibility. You cannot fix what you cannot see in the open.