A Cut List Is Not A Cut

The folder is tiktok_short_film_social_hook_01 and it contains 5 directories and 1 file. The directories are 00_Brief, 01_Images, 02_Video, 03_Finals with a v1.0 inside it, and 04_Archive. The file is 00_Brief/cutlist.json: 6,619 bytes, 548 words, 249 lines and 18 top-level keys. That is the entire contents of a cut. The list declares exactly 4 outputs by absolute path - a final render under 03_Finals/v1.0/, an EDL for Resolve, an FCPXML for Resolve and a craft review at 00_Brief/REVIEW.md - and all 4 paths are absent. Zero of 4. A reader who opens the folder sees the shape of the work and none of the work. The brief is written, the bins are created, the naming convention is fixed, and the last mile is empty. This is not a folder that failed halfway through an edit. It is a folder that stopped at the exact point where the plan was finished and the spend had not started.

Read the plan and it is complete. The label reads Social reel 15s. The delivery block reads 9:16, 1080 by 1920, 30 frames per second and minus 14 LUFS. The timeline block reads total_dur 10.0 across 4 beats. The label and the timeline disagree by 5 seconds, which is a third of the piece, and nothing reconciled them, because nothing was rendered. The 4 picks are priced and cut to the frame: the hook holds 2.4 seconds from 1.16 to 3.56, the turn holds 2.4 seconds from 1.16 to 3.56, the payoff holds 3.2 seconds from 1.5 to 4.7, and the close holds 1 still for 2.0 seconds from 0.0 to 2.0. 2.4 plus 2.4 plus 3.2 plus 2.0 is exactly 10.0. 2 of the 4 windows lift the same 1.16 to 3.56 offset from 2 different shots. The music bed is chosen as well: 1 file of 206.160978 seconds, of which the reel would spend 10.0 seconds, or 4.85 per cent. Every decision a cut needs is already written down. 10.0 seconds of edit exists as intent and 0.0 seconds exists as files.

4 picks come from 4 distinct source assets across 3 productions, and the reuse block reads previously_used 0. Across the whole cuts tree, 45 of 46 picks carry a use count of 0. Each pick carries a why string, and every one of the 4 names a single tag beside a frame size: tags[hook] at 768x1364, tags[crack] at 768x1364, tags[hero] at 1366x768 and tags[duo] at 2816x1584. The frame sizes carry the next problem. The delivery is 9:16, a ratio of 0.5625. The hook and the turn arrive at 0.56305, so they fill the frame with a 0.13 per cent crop and a 1.4081 times upscale. The payoff arrives at 16:9 inside a 9:16 delivery, so it has to lose 68.37 per cent of its width and then scale 2.5 times to cover. The closing still is the same shape: a 68.36 per cent crop and a 1.2121 times scale. 2 of the 4 picks fit the delivery and 2 fight it, and that was known before the render, not after. Reuse-before-render also left 2 beats unfilled: context and proof, both marked BUILD, both carrying the same reason, which is that no reusable asset matches the beat. Those 2 gaps are the honest part of the file.

Every pick carries a score and this file prints all 4 of them: 6.2, 6.2, 6.2 and 4.614. That is 23.214 across 4 picks, an average of 5.8035. Set that against the rest of the tree. The 7 cut lists hold 46 picks scoring between 4.612 and 12.2, and their averages read 9.7967 for the movie sequence, 7.028 and 7.028 for the 2 Sal_y_Miel hooks, 9.438 and 9.4247 for the 2 teasers, 5.8035 for this reel and 6.1413 for its sibling. The average of those 7 averages is 7.8086. The 2 teasers that shipped carry highs of 12.2. This reel's high is 6.2 and its average is the lowest of the 7. It also carries 6 alternates for 4 picks: 2 under the hook, 2 under the turn and 2 under the payoff, leaving 0 under the still, which means the library was asked for substitutes on the 2 scored beats and answered twice on each. A file that prints its own 5.8035 is not hiding anything. The number is the reason the folder looks the way it does.

5 directories, 1 file, 6,619 bytes.

Zoom out and the tree names 7 cuts: 4 social hooks, 2 social teasers and 1 movie sequence, all generated inside 35 minutes on 2026-09-11 between 15:50:44 and 16:25:47. 7 folders times 4 declared outputs is 28 paths. 22 exist. 6 do not. Break the 6 down and the pattern is the last mile. REVIEW.md exists in 6 of 7 folders. The EDL exists in 6 of 7. The FCPXML exists in 6 of 7. A final render exists in only 4 of 7. 2 folders are missing exactly 1 output and both times the missing file is the final. 1 folder is missing all 4 and it is this one, the folder whose own picks carry the lowest average of the 7. Files on disk run 23, 6, 9, 9, 17, 1 and 6, and the 1 belongs to the cut that has no cut. The tree is not failing at the planning stage. It holds 7 complete plans, 4 deliverables and 3 cuts that stalled on the same file. Total timeline across the 7 is 144.66 seconds, and every one of those plans was written in the same half hour, which is what a selection step looks like when it runs 144.66 seconds ahead of the render step.

An unbuilt cut leaves behind 3 things a finished one cannot. It leaves the price, because the picks were scored before anybody rendered. It leaves the gaps, because the library was asked and twice said no, and a gap recorded in writing is the only honest justification for spending money on a new render. And it leaves the arithmetic unspent: 10.0 seconds of timeline, 4 picks from 4 sources across 3 productions, 0 repeats, and 6,619 bytes of intent sitting inside 5 directories beside 0 bytes of deliverable. Apply the same reading to any plan, including this one. A cut list is a decision about where the scissors go. A cut is the fold that exists in the timeline afterwards. The distance between 4 declared paths and 0 files is the entire difference between a plan and an asset, and it is measurable in 2 numbers: 4 declared, 0 present. The probability that a complete-looking plan gets counted as finished work is not small, because a plan this detailed reads like work and the tell is 1 line long, which is the file count. Entrade built 200 wood gasification machines across the UK and Europe on the rule that the second unit has to cost less than the first. A library that writes 5.8035 beside its own picks is keeping the same kind of books. Keep more of yourself, for longer.