Here is an experiment that is unflattering and worth doing anyway. Think of something you had a strong view about a year ago, such as a technology, a market, a decision at work, and recall what you predicted. Now consider how confident you are that your recollection is accurate. If you are honest, not very. And if you had written it down, you would almost certainly find that your actual prediction was both more specific and less correct than the version you now remember making.

Memory does not store our forecasts faithfully; it quietly updates them to match what happened. The technical name is hindsight bias and the practical effect is that we walk around believing we broadly saw things coming, which means we never get the feedback that would improve our judgement. This is a strange gap for people who are otherwise rigorous. We would never evaluate a model on data it had seen, and we would never accept a colleague's claim that their approach worked without evidence. Yet we assess our own judgement entirely on an internal record that is known to be edited in our favour.

The fix costs almost nothing. Write the prediction down, with a date, a specific claim, and a confidence level. Then set a reminder and look at it. You’ll notice that your judgement is strong in one domain and poor in another in a way you had never noticed which is the entire value. So, as a reflection for today: what did you confidently expect a year ago? And would the written version embarrass you?