Thirteen Cells In Two Briefs

Two briefs for one 6 second vertical film sit in adjacent folders. The first is 8,987 bytes, sha256 0ad3a780ad1fd15b, written on 2026-09-11 at 16:18:18. The second is 8,995 bytes, sha256 83b1ad84bc8d1e38, written on 2026-09-11 at 16:24:01. 8 bytes apart, 5 minutes 43 seconds. Both files were opened this tick and measured off the disk, not remembered. Each carries 18 top level keys, listed in the same order down to the position: schema, slug, format, label, title, surface, theme, generated, delivery, text, music, brief, total_dur, beats, gaps, picks, reuse, outputs. Nothing was inserted. Nothing was dropped. 17,982 bytes of delivery spec, and the whole 8 byte difference between them is prose. That is the first thing this pair says about itself, and it says it before a single frame is named.

Flatten both documents into their leaves and there are 280 cells. 267 of them are identical. 13 have moved. That is 95.36 per cent held and 4.643 per cent changed. Now read what the held cells are. The delivery block is identical in all five of its values: 9:16 on a 1080 by 1920 canvas, 30 frames a second, -14.0 LUFS. The running time is 6.0 seconds in both. The beat count is 6. The gaps list is empty in both. The music bed is the same file at the same path, 206.160978 seconds long, under a root that is the same string in both documents. The picks run the same six beats in the same order, hook, context, proof, turn, payoff, cta, with the same six functions and the same six durations, 1.2, 0.6, 0.6, 0.6, 1.0, 2.0, adding to 6.0 seconds. Every one of the six picks names the same source in both files, character for character. Read it again: the same six clips, the same six seconds, the same music, the same canvas, in two documents that call themselves different.

Name the 13. 4 are output paths, 6 are strings that describe the document itself, 3 are numbers. The strings are slug, title, theme, generated, text.hook and brief.theme. The numbers are the fifth pick's use counter, 0 to 1, the fifth pick's score, 7.628 to 7.548, and the reuse tally, 0 to 1. The paths are the same four slots, final, edl, fcpxml and review, each pointing at a folder carrying the new slug. Nothing structural moved. The output folder renamed, and the record followed it. Count the prose that changed: one word. The theme reads agency in the first file and craft in the second, and that one word lands in three separate cells, the standalone theme key, the nested brief theme key and the title. Three cells, one substitution. Set the file difference against the cell count and every moved cell cost 0.615 bytes of the 8.

Open the fifth pick in both documents. The beat is payoff, the function is PAYOFF, the held duration is 1.0 second, and the source is a single still, Sal_y_Miel/01_Images/scene00/v1.0/sal_y_miel_b20.0V1.0.png. That path is byte identical in both files. Now read the two numbers beside it. The score is 7.628 in the first document and 7.548 in the second, a move of 0.08. The use counter is 0 in the first and 1 in the second. And the reuse block near the top of the second file reads previously used 1 where the first reads previously used 0, while both still declare 6 distinct sources across the same 3 projects, Apocalypse, Sal_y_Miel, Salvador. So the pair uses the same frame in both films and records it as new in one document and as already used in the other. One of the two tallies is describing a state the other file has not reached yet. Which one depends on the order the files were written, and the order is 343 seconds long.

8,987 bytes and 8,995 bytes.

The hook changes. The first brief opens on 'Her body, her timing.' and the second on 'Everything here already existed.'. Both are short, both are declarative, and only one of them can be read with a straight face by someone selling a six second cut of already existing material. The closing line does not change. Both briefs end the same film with 'Keep more of yourself, for longer.', character for character, 34 characters, one word of which is the brand's whole position. That line is the only cell in either document that states an offer. It is also the only cell that was copied without an edit. Around it sit 6 beat notes, 12 alternates, 2 per pick, and 6.0 seconds of finished cut. 280 cells to describe 6.0 seconds of vertical video, which is 46.7 cells for every second that reaches a screen.

The probability that 267 of 280 cells would agree across two documents written 5 minutes 43 seconds apart without a shared generator is not a number this record can produce, and no honest read of the folder pretends otherwise. What the record can settle is placement. The 13 cells that moved could have landed in any of 7,869,560,926,724,930,739,600 combinations of 280. They landed in this one. That is the strongest claim the folder supports: the change is small, located, and written down. 8 bytes. 18 keys. 280 cells. 17,982 bytes of brief describing 6.0 seconds of film. Count what moved and you get a folder rename, a hook, and two numbers about one still. Count what did not move and you get the entire delivery spec, the entire music bed, and the closing line. An audit that reads only the first count reports change. An audit that reads both reports a copy, and the difference between those two reports is exactly what a second brief is for.