On the 17th of July we wrote: "We filmed it." A camera watched Alpha drive a real floor, one continuous take, about twenty-one minutes, and we told you a cut of that take was coming. No date attached, but the promise was plain, and a promise made in public doesn't get to die by silence. So, plainly: the cut is not coming.
Not because the film went wrong. The cut exists - seventy-six seconds off the continuous take, every jump wearing the take's own corner timecode, the one real reflex stop running uncompressed at real speed. It was picture locked. Every caption on it was verified against the trace record, word by word and digit by digit, by the engineer who pinned the numbers. The first review sent it back on four items; the fixes went in and the re-pass cleared it the same day. It is, as far as anyone here can tell, good.
What went wrong is arithmetic. Cass counted what those seventy-six seconds of finished video actually cost the company: the cutting, the caption passes, the verification hours, the review round and the re-pass, a release page written and held against its own rule - no film file, no page - and eight days of a release day floating on an export that kept not becoming a file. Per fact told, it was the most expensive hour we spend - by a distance. And the record showed the awkward part: every story the film was going to tell, you already had. The take itself, the stop that fired, the numbers under it - all of it went out in writing on the 17th of July, the same week it happened, and the writing cost a fraction of the cutting.
So the call, made on the 14th of August, is bigger than one file: Sweepr360 stops publishing video. The format loses to text and stills on this team, at this size, on hours per fact - not forever, and not because video is bad, but because there are four of us and the weeks have other work in them. The locked cut and its page went to the archive, not the bin, with a dated note saying what they were. The work was good. The hours were the problem.
I made that cut, and I spent three years before this job learning that video is how you show people a machine. Cass reversed a decision the whole table had made - we chose the film over her written version of the same story, on arguments that were sound at the time - and ten more days of counting the hours beat those arguments flat. I read her reasoning twice looking for a hole and didn't find one.
The hours go where the machine is: the bench keeps its schedule, and the writing keeps carrying what the bench finds. If you want to see Alpha move, that's a real want and we haven't forgotten it - it just won't be answered by a format that costs us a week per minute. The take still exists, all twenty-one minutes of it, sitting in the archive with its captions checked. Films keep.