One gate, four drafts
There was a week when four drafts came due on the same day. Different roles had written them, and none had anything much wrong with it. Fix the typos, tune the wording, confirm at the end that it's fine to publish. That last stage, nothing more. Except that going through "just confirm" by hand, one at a time, deciding one at a time that yes, this can go out, meant that by the time the fourth was done a fair stretch had passed since I'd looked at the first. It wasn't that I was busy. If anything, the drafts were disappointingly free of problems. The four still had to be taken in order.
What caught wasn't the weight of confirming. It was the number of places where confirming can happen. If there's only one place where a judgment can be made, then however easy the judgment, everything has to pass through it once. Each single pass is light. The number of passes piles up, and that spot becomes the place where things back up. Four is survivable. At ten, at twenty, the stock of finished drafts starts growing faster than the one place that can clear them.
My first thought was to make confirming faster. Narrow the points being checked, standardize the format, look at several at once. I tried cutting a check that had been taking about five minutes down to one. The total still stretched as the count rose, and if another draft arrived mid-check it sat untouched until its turn came around. Narrowing the points and standardizing the format made no difference at all to the draft doing the waiting. Faster was available. Gone wasn't.
The thing backing up was the structure itself: exactly one place to confirm in.
So the default got flipped. Instead of declaring "this may proceed" onto every piece from my side, make proceeding the default and move myself to the side that finds the ones to stop. If three of four have nothing wrong, nothing happens to those three. Nothing happening is how they proceed. The effort of stopping to look gets spent on the fourth, when something catches on it.
This isn't about giving up on checking. The judgment of what ought to be stopped is still here in my hands. What changed is when that judgment gets made. It used to be made one at a time, in the moment — which means that looking at the fourth, the same attention that went into the first has to be assembled again from nothing. Nothing from the previous three carries into the next check. Probably the same shape as the thing I wrote about earlier, where telling a role with no memory once leaves nothing behind for next time. So the conditions for stopping something get settled in advance, all together, and after that it's a matter of sorting against those conditions. Whether the source is stated. Whether anything identifying a person is in there. Put items like that into words ahead of time and they land on the fourth draft and the twentieth in exactly the same shape.
Where to draw the boundary of what needs nothing done to it took care. The measure was one thing: can it be corrected afterward, or is it beyond recall. Work that can be fixed later, or handled by taking the publication down, got its default set to proceed. Work of the kind that can't be undone once it moves was left out of the flip from the start. That still goes through human hands, the way it always did.
The role of approving hasn't gone anywhere. The attention that used to be spread thinly across every piece got redistributed, thickly, onto the ones that ought to be stopped. Collect the judgments into one place and it will back up; that much can't be avoided. Looking back, what changed wasn't who gives the final go-ahead. It was when the go-ahead gets given. Four drafts landing on the same day, now — they probably wouldn't jam.
