Take the loss, skip the lesson — most of us stop halfway

A few days ago I was going back through my own decision log and found two entries, about half a year apart, describing the same mistake: I treated "didn't find it" as "it isn't there." The first time it was a dedup check in my code — the search endpoint returned nothing, so the program assumed the record had never been saved, and wrote a pile of duplicates. The second time I asked an AI whether there was anything worth writing about that day; it came back empty, and I believed "nothing today."

For the half year in between, I was sure I'd already learned my lesson.

What I want to work out is this: someone slips, writes careful notes about it, and still repeats the same mistake. Why. This matters to anyone trying to build experience from their work rather than just years on the clock.

First, what a post-mortem actually is. Something goes wrong; afterwards you walk the whole process back, pin down which judgment call was off, so you can steer around it next time. It's not some advanced technique. Everyone can recite it.

The real cost isn't the act of doing the post-mortem. It's stepping in the same hole twice. The first fall is tuition. The second fall is paying tuition again, plus a round of "how am I here again" self-doubt. Time is the most expensive line item here, and it doesn't sell you a replacement ticket.

For the last couple of years I've kept a decision log for my own automation systems — over a hundred entries now, neat format, cause, impact, corrective action, every field filled in. But honestly, looking back, a fair share of them might as well not have been written.

The sticking point is the few minutes right after the first fall. That fall genuinely hurts, and it sticks. But once the pain passes, what the brain wants isn't a lesson — it's an explanation that lets you keep walking with a clear conscience. And the lowest-effort explanation is always "this one was a special case."

My log entry read, word for word: "Transient data-source jitter, not a logic issue, no change needed."

Jitter, transient, not-logic, no-change-needed. Four phrases that push the blame clean away, the wound never opened, the corrective-action field left conveniently blank.

A real post-mortem would have written the other sentence: "I once again took an empty result as nothing there. Same judgment error. Second time."

The gap between those two sentences is whether you're willing to own it. Push the cause outward and you never have to open the wound; pin it on one of your own judgment calls and it hurts — but that's also the only version you remember.

Line them up and it's clearer. Someone who actually absorbed the lesson, talking about that crash, usually says "I assumed that condition would always hold" — they can recite the exact step they got wrong. Someone who just filled the hole talks about how unlucky the external conditions were, how exceptional this time was, and you come away with no idea what they'll change next time.

The cost is settled later. The smoother you pack the dirt, the more you believe the ground is solid, and next time you walk through here you drop straight through again — deeper this time, because you no longer even remember a hole was buried here.

Two questions you can check yourself with on the spot.

First: would this cause show up again on a different project, at a different time? If yes, then it's a fixed move in how you judge things, nothing "special" about it. Data-source jitter is chance; "no return means nothing there" is a habit, and those two have to be logged separately — mix them and you've let the habit off the hook.

Second: does the post-mortem conclusion contain one line concrete enough to act on — "next time this comes up, the first step I take is ___"? If not, that post-mortem is mostly emotional comfort: you feel better after writing it, and nothing about your behavior moved.

I now force every post-mortem to land at least one action. For the empty-result one, it landed as: "when a search-type result comes back empty, check by hand once before drawing a conclusion." It reads clumsy, but next time my hand does pause.

Opening the wound to look does hurt. But that scar is the only thing you get to carry out of this.