Independent critic B — 066 / EIGHT PLACES — V01 Verdict: bounded VISUAL PASS for the constructed substitution grammar and release from it. Viewer learning, anticipation at native speed, and full audiovisual acceptance remain INCONCLUSIVE. The ordered version offers a stable recoverable sequence that the shuffled control removes; this inspection does not measure whether an unprepared viewer learns it. No 'reference stronger' judgment is available without inspecting a comparable reference film. Method I first inspected all five one-second main sheets and then all five control sheets before reading GRAMMAR.json and GRAMMAR_CHECKS.json. Because one-second sampling aliases the 0.6-second slots, I then extracted a slot-center view across all 80 cycle slots in each actual movie, in temporal order, using 0.6-second sampling beginning at 9.9 seconds. I inspected the four resulting forty-image sheets. I also inspected a denser four-frames/second extraction of the released carton passage and its transition into the held endpoint. The first attempted OpenCV extraction was unavailable; FFmpeg performed the successful actual-frame extraction. No audio audition, native real-time playback or blind viewer-learning test occurred. I did not read a builder script or other critic report. Actual grammar visible in the export 0–9.6 seconds: eight large spare symbols are individually introduced with names: GATE, FORK, LOOP, LADDER, CROSS, BOWL, SPARK, FIELD. The symbols are easy to distinguish in still inspection. They are an invented eight-glyph vocabulary, not a displayed A–Z alphabet. Spoken naming cannot be assessed here. 9.6–14.4 seconds: all eight appear again in the same order, at 0.6 seconds each. This is the only rapid complete all-symbol cycle before replacement starts; the earlier slower introduction supplies the other exposure. Whether that is enough to teach eight positions in real time remains untested. First image replacements occur at 14.4 (carton folding), 19.8 (curling metal), 25.2 (cup painting), 30.6 (projector), 36.0 (diver), 41.4 (reclining pour), 46.8 (office), and 52.2 seconds (people with placards). Each persists in its own serial position after substitution. The actual slot-center sheets make this progression visible without relying solely on the ledger: later rows contain increasingly more pictures at their former symbol positions; no replaced position goes back to its symbol. These are advancing excerpts, not the identical six-tenths of a second repeated. The carton worker's hands change position; the metal curl rises and falls out of view; the painter reaches and returns; the projector's material moves; the diver approaches departure and becomes airborne. The office changes occupants through a source dissolve across its small number of appearances. That is a change within the source, not evidence of one person's continuous action. The placard scene has only two image appearances and the least time to develop. The film is properly a grammar of records, not eight equally complete events. The arbitrary symbol-to-record associations do not have to be semantic illustrations. 'FORK' is not a claim that the tool in the metal shot is a fork, and 'SPARK' does not describe the office record. Their meaningful constraint is persistent position and identity. Avoid describing these as eight words proved by matching objects. At 48–57.6 seconds all slots are images for two complete cycles. This is the most demanding passage for remembering the vanished symbols and the clearest separation from a generic collection of attractive symbols intercut with footage. The symbols have actually disappeared, while the image order remains. At 57.6 seconds the carton record continues beyond its customary slot. Through 67.6 seconds it has time for several operations: manipulation at the carton edge, closing/flipping the box, reopening and further arranging another box. It is the first record introduced and the one given the longest cumulative exposure before this release, so returning to it is well prepared. 'The action leaves the slot' means the fixed temporal slot is abandoned; neither the man nor box has to leave the image frame. The passage remains a repetitive demonstration, not a terminal completed task. At 67.6–70.6 the image is explicitly held and labeled 'HELD FINAL FRAME / Editorial hold after the released passage'. That endpoint is editorial, not a person observed stopping. Ordered versus shuffled actual pictures The control changes the order of identities within each cycle. Even before many substitutions, symbols shift relative to their neighbors. After replacement, the carton, metal, pottery and projector positions migrate while keeping their own record identities and advancing source action. Thus the ordered version supports the specific anticipation 'the next place should be X'; the control chiefly supports recognizing recurring content and declining symbol count. Both versions visibly perform substitution, and both share the same carton release. The control therefore tests the value of stable order, not whether images alone are sufficient or whether the ending works. I prefer the ordered version for the intended serial constraint. This is an inspection of available structure, not evidence that an audience performed better on a prediction task. Diagnostics and their limits GRAMMAR.json records 30fps, 18-frame slots, ten eight-slot cycles from 9.6–57.6, monotonic replacement and progressive source starts. Its first-replacement times agree with the actual slot sequence inspected. GRAMMAR_CHECKS.json reports passing duration/order/progression invariants and correctly says viewer_learning 'not measured'. Those structural reports do not prove real-time legibility, attention or learning. A slot-center inspection verifies each slot's visible identity, not every frame inside every 18-frame interval. Source time continuity here is supported by visibly changing records and the ledger, not an independent full source-frame checksum audit. Largest remaining gap and concrete repair condition The largest perceptual gap is whether the sequence is actually learned before replacements, particularly the rapid 9.6–14.4 all-symbol cycle. My image method cannot certify that intended experience. The last substitution at 52.2 then gives FIELD only two image returns before the release; late records have much less opportunity to become anticipated actions than early ones. A discriminating next check is a normal-speed viewing with the ledger hidden: after the initial introduction and full-symbol cycle, pause at a cycle boundary and ask for the next two positions; repeat after partial substitution and compare the shuffled control under the same conditions. Do not report those results until performed. If anticipation fails, the smallest observable repair is one additional complete fixed-order all-symbol cycle before the first replacement, funded by shortening the separately labeled final hold/credits or extending duration transparently. Preserve equal slot durations and persistent mappings; do not add persistent numbering that substitutes a visible counter for the intended memory task. This is a proposed repair conditional on an observed learning failure, not a reason to declare the present construction failed from stills. Material and lineage scope The final FIELD record visibly includes placards and a collective street action. It is socially specific, not merely neutral moving texture; the source caption 'Working Together' preserves an identity but not much context in 0.6-second bursts. Any expanded explanation should call it a recorded placard/picket scene rather than infer individual beliefs or reduce it to generic walking. The office and reclining domestic scenes are staged corporate images, and other records are industrial demonstrations. The grammar places these historical records in relation without making them simultaneous or causally connected. The closing title explicitly identifies a new eight-symbol grammar, says it is not a Frampton reconstruction, marks the requested A–Z (1974) attribution unresolved, and distinguishes related Zorns Lemma (1970). I have not inspected either reference and do not certify fidelity or reference superiority. Likewise the stated synthetic introductory voice, silent cycles and supplied carton track are documented assembly choices; this image-only review has not heard them or verified audible synchronization. Strengths to protect Large distinguishable glyphs, genuinely persistent removal, changing source excerpts, the two image-only cycles, and the prolonged carton passage give the short a concrete system and a visible way out of it. The ordered/control difference is substantive. Keep the ending's clear distinction between recorded motion and editorial hold, and keep viewer learning explicitly unmeasured until a suitable viewing can test it.