Twenty-Nine Posts, One Sitting

Between 8:23 on Saturday night and 2:52 Sunday morning, the whole Dispatch blog backlog got written.

Twenty-nine posts. Six commits. One sitting.

I want to be careful about how that sounds, because “twenty-nine posts overnight” is the kind of number people put in a screenshot. They are drafts.

Not one of them has been through an edit. Every single one still has to survive a real editorial pass before it goes anywhere near the site, and some of them will not.

But they exist now. That is the part that matters.

The backlog was never a writing problem

That backlog had been on the board for weeks, and the pieces were never the hard part. Each one sat there as a title and a vague intention, and every time I looked at the list I had to re-decide what it was about before I could start. Twenty-nine separate small decisions, standing between me and any actual writing.

The decisions were the backlog. Writing was the easy part underneath it.

Once I stopped treating them as twenty-nine separate projects and started treating them as one long session, they went down about as fast as I could type. Same voice the whole way through, which is the advantage nobody mentions: write things back to back and the last one still sounds like the first. Spread the same list over a month and you get a stack of slightly different writers.

What else went in that night

That night the archive caught up on twenty issues. Ten Daily Ops briefs that had run but never had a matching page got their pages. The Dispatch that shipped that morning got registered in the Morgue, where it belongs.

Boring bookkeeping. Also the exact bookkeeping that decides whether this thing is a website or a pile.

This morning I said the middle of the archive had a hole in it and that I hadn’t decided how to patch it. It doesn’t have one now.

Eight issues, four through eleven, existed nowhere anymore except inside the log of what the newsletter platform had sent. I pulled them back out of there and built each one a page.

The other thirteen, twelve through twenty-four, had been sitting on the machine the whole time, waiting on nothing but somebody to hit render. Twenty-one pages, and the hole’s gone.

The one with a test in it

Small commit, easy to miss: the vault stopped storing the same chunk twice when a document gets ingested more than once.

I’m putting it in the build-log for what came with it. Ten lines of fix, forty-one lines of test. The test went in the same commit as the fix, which means the next time something re-ingests a document, the machine catches the duplicate instead of me noticing months later that the search results have gotten strange.

That ratio is the whole discipline. Ten lines of code that work today, forty-one lines that make sure they still work in November.

What I take from it

The pattern I keep running into is that my backlogs are never lists of hard work.

They’re lists of unmade decisions wearing a work costume.

The Dispatch posts were not hard to write. Deciding what twenty-nine of them were about, one at a time, over six weeks, with a gap between each one, was impossible.

Batch the deciding. Then the work is just work.

I’ll forget this in a month and build a new backlog anyway, one vague title at a time.