My Most Expensive Mistake Had Nothing to Do With Code

The most expensive mistake I ever made had nothing to do with code. It had to do with a promise I made before I understood what the promise required.

I built my first company shooting weddings. For over a decade, my team and I documented hundreds of events across four states. Wedding work has a property that most software work pretends it does not have. You get one take. The first kiss happens once. The father-daughter dance happens once. If your card fails, if your exposure is wrong, if you were changing a lens at the wrong second, there is no reshoot. The client does not get married again so you can try harder.

That constraint taught me the lesson that everything since has been built on. Ship it right the first time, because there is no second time.

The mistake I want to talk about came later, on the government side, and it looked nothing like a missed shot. It looked like a commitment I made on a timeline I had not stress tested. I told a customer we could deliver by a date. I believed it when I said it. What I had not done was walk the whole path from the commitment back to the constraint. I had counted the work. I had not counted the handoffs, the dependencies, the single person who had to touch three of the steps, or the reviews that could not start until the step before them finished.

The date was real to the customer the moment I said it. It was a hope to me. That gap is where the expense lives.

The cost was not the delay

The delay was recoverable. We caught it, we escalated, we did the unglamorous work, and we closed it out. The expensive part was what the near miss revealed about how I was operating. I had been treating my own confidence as a plan. Confidence is not a plan. Confidence is a feeling that shows up whether or not the underlying system can carry the load.

In wedding work I never confused the two. I would never have told a couple that I had the ceremony covered without knowing exactly which camera was on the aisle, which one was on their faces, what the backup was, and what happened if the room went dark. The discipline was total because the consequence was total. Somewhere in the move to a business with more steps and more people, I let confidence stand in for that discipline. The result was a promise the system had not earned yet.

What I changed

I stopped making commitments from the top down and started making them from the constraint up.

Now, before a date leaves my mouth, I trace the actual path. Not the happy path. The real one. Who has to do the work. What has to be true before they can start. Where the single points of failure are. What the review adds. What happens on the day someone is out. I look for the one step that everything else waits on, because there is almost always one step that everything else waits on, and a plan that ignores it is not a plan. It is a wish with a calendar attached.

This is the same reason our AI systems are built the way they are. When an agent at Trunnion takes a consequential action, it does not get to run on confidence. It clears policy. It clears authorization. When the move is high stakes, it pauses for a named human being who signs off. Every step lands in an audit trail that cannot be quietly rewritten later. That is not bureaucracy. That is the same lesson from the wedding, encoded into software. Do the work up front so the thing holds when it matters, because when it matters is exactly when you do not get a second take.

Why I tell this one

Founders are told to sell the vision, and the vision is usually a set of confident promises about the future. I am not against vision. I am against the moment where confidence quietly replaces the walk from the promise back to the constraint.

The teams that deliver are not the ones that feel the most sure. They are the ones that have looked hardest at the exact place their plan could break, and built for that place before anyone asked them to. The near miss taught me to go find that place on purpose, every time, before I commit. It is the cheapest expensive lesson I have ever paid for.

Systems that hold up under scrutiny are not built by people who are certain. They are built by people who went looking for the crack while there was still time to fix it.