Decisions.

The framework

1. 01

   Pull ten real decisions from the record, not from recall.

   Memory hands you the vivid and the vindicated and buries the rest, which is the one sampling error that makes a review worthless before it begins. Go to the source instead: the calendar, the sent folder, the bank statement, the hiring file, the roadmap you rewrote. Take the last ten consequential calls in order, whatever they say about you. And include the decisions you made by not deciding — the offer you let expire, the hire you kept postponing, the contract that renewed itself in silence. Those are decisions too, and they are the ones nobody ever reviews.

2. 02

   Score the reasoning — the checklist is everything this elective has handed you.

   For each of the ten, ask only what you did before you committed, and use the elective as the marking scheme. Was this even the call that deserved the attention, or the one that shouted loudest. Did you write the expectation down with a date on it. Did you start from what the record gives people in your position rather than from the particulars of your own case. Were the branches priced and the world's replies drawn, or did one loud outcome get to size the whole bet. Did you check which kind of door you were walking through and set your speed to match it. Did you name what would make you wrong, and give somebody the standing to say so. Did money already spent get a vote it had not earned. Was it yours to decide at all. Was the wait priced before you took it. Grading reasoning against outcome is Day 62's job and the journal entries are the evidence for it, so leave results out of this entirely. What you are measuring here is thinner and more uncomfortable: how many of the ten were decided by a method you can name, and how many were decided by feel and defended afterwards.

3. 03

   The method you never once reached for is the real finding.

   The mornings behind you handed you a shelf. Lay the ten decisions side by side and count which of the methods actually appear: the inversion, the pre-mortem, the base rate, the priced branches, the drawn replies, the stopped search, the journal entry, the door test, the commissioned dissenter, the owner named before the debate began, the weekly price of waiting, the belief moved by degrees rather than flipped. Most people find two or three doing all the work and the rest untouched — and the untouched ones are rarely random. They cluster around the same discomfort, usually producing either a number you could later be shown to be wrong about or a person with standing to argue with you. That cluster tells you more than any verdict on a single call, because it will repeat itself across the next ten decisions unless something changes. Pick the method that never appears once, and make it compulsory for a single named class of decision — not for everything, or it will hold for a fortnight and then quietly lapse.

4. 04

   Leave with one rule in force and a log already open.

   The audit pays only if the next decision is made differently, so convert what you found into a single standing rule, written in language that will still bind you when you are busy: reversible calls under a named threshold get decided within a day by whoever sits closest to the work, and anything irreversible above it gets a pre-mortem and a written condition for abandoning it — Day 64's door test turned into standing policy. One rule, not five; five is a resolution, one is a change. Then open the log for the next ten before you make them: date, decision, what you expect, when you will know, who owns it. Book the review that will read it, in the calendar, this morning.