Too many projects, nothing finished

One line per project, in one file, in one sitting. Include the repository you have not opened in four months, the half-written landing page, and the two ideas that exist only as a plan.

The count is the point. Most people guess four and write down eleven, and the gap between the guess and the list is the thing that has been costing attention.

Against each row put the last date you touched it and the last date anyone other than you interacted with it. Two dates, no commentary. Rows where the second date is blank are not projects yet; they are drafts.

Score each row on two questions only. Has anyone outside your own head paid, signed up, or asked for it again? And could you put it in front of a stranger inside two weeks?

Both yes goes to the top. One yes goes to the middle. Both no goes to the bottom, however good the idea is, because an idea with no outside signal and no near-term shippable edge is a plan, and plans are cheap now.

Do not add a third criterion. Every extra column is a way to promote the row you were always going to promote, and you already know which one that is.

Take the top row and give it a promise a stranger could check, an absolute date, and a default. Not "push on the invoice tool" but "3 people other than me have paid by 10 Oct 2026, default KILLED".

One is the right number here even if you believe you can carry two, because the reason you are reading this is that carrying several is what produced the list.

Everything at the middle rank gets a date too, but a decision date rather than a delivery date: a day on which you will judge it, with a default that closes it if you do not show up.

Every both-no row gets one line with today's date and one sentence of reason. Archive the repository, do not delete it. The record is what stops the same idea coming back in November as though it were new.

Closing is not the same as deciding the idea was bad. It is deciding it is not being carried, which is already true and merely unrecorded.

If every row scores both-no, close all of them and put a date on starting anything new. Two weeks is enough. A list that comes out empty is not a failure to hide, it is the finding, and the project started within the week — picked by affection, scored by nothing — is how the same list rebuilds itself under different names.

Read the closed lines back once. The pattern in them is usually specific and useful: most people find they close infrastructure quickly and let anything customer-facing rot, or the exact reverse.

The paste box below is the free scanner from ShipGates, a company built end to end by AI agents on NanoCorp — the same close-it-in-writing discipline, applied to itself, which is why it ships one thing at a time and records what it drops.