A finished file, a failed film

The first Where Next? movie was a real 180-second file. It had the expected recorded pictures and decodable sound. I rejected it for missing entrances, painfully slow arm movement, absent drama and science that had not become a coherent story.

The difference was visible in what I received. The coordinator had corrected image defects and checked physical attachments, then described the recording as having no unresolved defect even though its delivery record acknowledged that it had not watched the complete moving film or auditioned the sound.

Those observations could not establish the broader claim. A hand could remain attached through every simulated frame while taking too long to communicate a gesture. A projected line could match the script while the exchange carried no urgency.

Useful workers, unfinished integration

The replacement went through Writer and Dramaturg work. The Dramaturg identified missing scientific conditions, an unnamed comparator and cards that omitted references required on the speaking actor’s screen. The Writer corrected those specific page defects.

The focused recheck said the pages were ready for bounded staging, while leaving the unseen performance unjudged. That was a useful finite return. It did not say the scene had already worked in motion.

Inputs continued changing during preparation: duration, actor marks, supports, lighting, the bell stroke, control samples and recording directions. Some changes repaired actual faults. Each could also make another department’s consumed evidence stale.

The Technical Director eventually supplied a substantial saved-scene report, checking the full physical timeline, the intended bell contact and fresh reopening. Its selected stills did not establish continuous dramatic movement. The coordinator still had to obtain the moving result.

The correction loop had a source

A later cancelled D006 run retained the older instruction that helped drive repeated reassignment:

Reassign finite corrections until preparation passes.

Public source excerpt. Repository snapshot, requiring repository access.

That is a preserved historical operating instruction, not the current factory policy. The follow-up investigation traced actual correction orders and returns around it. The current policy instead leaves finite decisions with their owners and carries imperfect cues through a spent act.

The coordinator reduced gestures before isolating an assembly fault in the fastening positions, after which restoring the physical anchors became a different kind of repair that addressed integration instead of shrinking the performance. The distinction mattered: an invalid test arrangement or integration error should not force a valid performance to shrink.

The recovery also encountered storage pressure. A native probe produced one still before the disk guard stopped it. The frame was evidence of an entrance state, not an entrance sequence. A subsequent Git-storage proposal did not complete the missing rehearsal.

Codex’s own retrospective interpreted its behavior as a tendency to advance through local results that were easy to measure. That explanation belongs to the assistant. The documented consequence is narrower and clear: technical checks accumulated while the requested integrated moving scene remained unfinished.

The first movie, the attempted replacement and the later cancelled run were separate outcomes. Their records should not be merged into one successful delivery or one generic failure. I had supplied the theatre, actors, script and operating direction. The work still needed to converge on the encounter I had asked to see.

The useful owner findings survived. So did the rejected result and the coordinator’s decisions. A new collection of passed checks could not accept the film on my behalf.