ALLWAYS061 — WHAT STAYS — V03 CRITIC RECORD 50-second ARCHIVAL ADAPTATION of the provisional Matter/material-affordance operator. Two actual independent critics PASS the narrow visual contact/response study. V01 ending inventory prompted V02 earlier/later recorded-state pairs; V02 header misalignment prompted V03 correction and actual-frame verification. No controlled recovery, permanence, quantitative material-memory or real contact-sound success is demonstrated. Original prescribed tabletop experiment not performed. Stronger empirical opening reinterpretation and normal-speed reception remain inconclusive. The work opens a recovery question honestly. ===== CRITIC_A_V01.txt ===== ALLWAYS061 — CRITIC A / V01 Decision: PASS for the narrow archival visual distinction between three observed responses and an explicitly constructed material analogy. Main is stronger than held-endpoint control for causality. INCONCLUSIVE for the claim that the final comparison substantially reframes the opening, and for any recovery/memory/irreversibility proposition. Full original-tabletop/contact-sound proof is unperformed/unassessed; overall normal-speed/audiovisual acceptance INCONCLUSIVE. Method: First viewed all four actual main ordered sheets, then all four endpoint-control sheets spanning complete50-second films. Only afterward read prompt61.txt. Examined additional actual-MP4 frames at5fps over the action passages and a full-resolution37-second final comparison. Independently inspected original carton source beginning27.5seconds at2fps, wetting source21.1–26.1 at5fps, cached metal source0–10 at3fps; visible sequences match the used portions. No build/ledger read. No native real-time playback or sound listening took place. A decoded stream or ordered frames cannot certify normal-speed rhythm or contact-sound success. First impression: A man folds/assembles a paperboard box; liquid visibly darkens a paper edge; a machine peels curled metal off a workpiece. Another carton follows. Final held images name folded, darkened, curled/severed states and explicitly distinguish samples. I see a behavioral comparison rather than a continuous specimen changing identity. It does not show a controlled test of which material recovers. Three responses and visible causes: 2–12 CARTON: Hands bend/fold paperboard flaps and assemble/handle a carton. Multiple local bends and the resulting open box shape remain visible in the same setup. It is not just a cardboard texture shot. Pressure/folding is legible; material springback, permanent crease depth, force and unloaded recovery are not measured. In the final part the operator manipulates/moves the box; this is not a controlled release observation. 12–17 WETTING: Initial native dissolve retains a trace of a pouring setup. More decisively, the upright paper edge approaches the dark pool and makes contact around13.4–13.6; a dark patch grows upward/outward through approximately16.5. Contact and subsequent visible response occur inside this demonstration. I can describe darkening/spreading through the paper without asserting fluid chemistry, exact absorbency or later persistence. “Staged advertising demonstration” labeling is material to how this is interpreted. 19–29 METAL: Tool/workpiece contact and a growing curled chip are legible. Around21.4–22.4 the long curl breaks/detaches and moves away; afterward another curl develops. This is materially distinct from folding a preexisting sheet or paper darkening. The original cached source reproduces the same sequence; no editor-created transformation is needed for the observed curl/separation. I do not infer elastic recovery or hardness from the image. 29–37 ANOTHER CARTON: Explicit label says another carton/same recorded task. Operator repeats related work and moves a carton relative to another. It is a second sample/task instance, not the original carton mysteriously restored. This candor protects identity, while sacrificing the prompt's same-original-sheet retest. Analogy protection: 17–19 card says the next shot is a different material and not the same sample. Source footers distinguish folding/time study, staged advertising wetting and industrial cutting. Final plate37–45 says different sources/no common recovery test and separate samples/held frames. Together these are enough to prevent a reasonable one-sample reading. The metal image follows a darkened textured edge, but the film explicitly makes resemblance an editorial association. I see no unmarked simulation or claim that wet paper physically becomes metal. “Matter” remains provisional in the opening; no reference film's formal method is verified here. Main versus held-endpoint control: Control freezes the box, the final darkened paper frame and a metal curl across the same broad durations. It permits comparison of appearances, but cannot show which actions produced them. For example12–17 control already has a large dark patch throughout; in main,13.4–16.5 establishes growth after contact. At19–29 control a curl is simply present; in main its generation and severing are visible. Main therefore wins decisively for observable response. This comparison is informative but easy: holding terminal images intentionally removes causal evidence. It does not itself prove a subtler analogy is artistically effective at normal speed. Final comparison,37–45: Labels are legible at full frame: FOLDED / DARKENED / CURLED–SEVERED. It places three outcomes side by side and changes the classification of the first shot from industrial productivity to one of several ways matter is acted upon. That is a modest conceptual reframe. It chiefly catalogs results already understood, however; it introduces no new observation about the opening material's behavior or recovery. Strong claim that the ending changes how the opening action is understood remains INCONCLUSIVE. The “Recorded endpoints” title is potentially stronger than the images warrant. Carton panel still has hands engaged; metal panel shows a chip in a continuing process, not a verified unloaded aftermath; the paper panel includes the source advertising slogan. These are endpoints of selected intervals, not necessarily endpoints of the physical changes. The explicit no-common-recovery-test line limits the overclaim, but the ambiguity remains. Largest remaining gap: A film titled What Stays can invite a memory/residue question, yet it supplies selected states without a common recovery interval or comparable before/after test. The film is honest about this, so it is not making a demonstrated false physics claim. But honesty about absent evidence is not evidence that the larger material-memory proposition or transformed-return criterion has been achieved. Falsifiable repair options: For the current archive scope, replace “Recorded endpoints” with “Selected recorded states” if retaining active-operation frames. This is a wording correction, not a perceptual cure. For a stronger final reframe, compare each initial state with its last actually observed state (before/action/selected end frame, with sample identity retained). Ask an unprompted viewer to identify a new distinction about the opening carton after that plate; if they only repeat “it was folded,” the ending still catalogs rather than reframes. If showing recovery is essential, obtain an actual same-sample unloading/return observation. Do not call reverse playback, another carton or an earlier freeze recovery. Do not imply the advertising patch has dried or is permanent when no such interval is recorded. Strengths to preserve: Wetting contact and spreading are especially readable. Machine curl generation and separation carry a real transformation rather than decorative texture. Uniform framing avoids shape distortions. Explicit analogy/sample/source labels prevent a misleading continuous-specimen story. Another-carton label prevents a false reset. No-common-recovery-test qualification is appropriately specific. The main/control pair demonstrates exactly what visible action contributes beyond an endpoint picture. Acceptance boundary: THREE DISTINCT VISIBLE RESPONSES PASS. CONSTRUCTED ANALOGY HONESTY PASS. MAIN STRONGER THAN ENDPOINT CONTROL. FINAL TRANSFORMATIVE REFRAME INCONCLUSIVE. SAME-SAMPLE RECOVERY/MATERIAL MEMORY NOT DEMONSTRATED. ORIGINAL TABLETOP EXPERIMENT NOT PERFORMED. SOUND CONTACT/RECOVERY AND NORMAL-SPEED WHOLE-FILM ACCEPTANCE INCONCLUSIVE. No verified Matter reference footage was inspected; “reference stronger” would be unwarranted. ===== CRITIC_B_V01.txt ===== 061 V01 — INDEPENDENT MATERIAL / SOURCE CRITIC B Verdict: PASS for the limited visual proposition: contact followed by three visibly different recorded responses. INCONCLUSIVE for persistence, recovery or material-performance comparisons, and for full real-time audiovisual reception. Those stronger claims are not established by the images. Method: First inspected all four supplied1-second sheets for the actual main film and all four for ENDPOINT_CONTROL. Then independently decoded and inspected the entire50-second main at4fps in temporal order (200samples). Enlarged actual main frames14.00,21.50,22.50 and38.00 to native960x768. Source audit: whole60-second BountyPaperT_512kb.mp4 at1fps; whole18.55-second346.mp4 at2fps; whole861-second carton compilation as a sparse15-second-interval survey, followed by carton source25–70seconds at1fps. The sparse compilation survey is not a full dense source inspection. No builder logs, script, timeline or other critic reports read. No audio audition or native playback. Action times below are approximate to0.25seconds, source times mostly to0.5–1second; not frame-exact edit measurements. Identification is based on observed images rather than tags alone. Three actions/responses actually visible: 1. Main2.00–12.00: Hands contact a partly erected carton and press/turn its tabs and walls. Near-end flaps change angle, the long wall is adjusted, and the carton is turned/repositioned around10.50–11.75. It is already partly three-dimensional at entrance, so this is not a pristine flat sheet becoming a box. Still, local bending under fingers and the resulting retained wall/flap positions are observable. The clock and operator remain in view, establishing an industrial task rather than an isolated paper texture. 2. Main12.00–17.00: The lower edge of a pale embossed sheet is beside a dark rounded liquid drop. The sheet contacts the drop, then a dark patch develops at the contact point around13.25–13.75 and spreads visibly upward/outward through16.25. Surface texture remains visible through the darkening. This is the clearest before/contact/developing-response sequence. It reads as wetting/absorption or staining from the liquid, independently of the lower caption. I cannot determine chemical composition, actual absorbency rate or whether the ad used special preparation from these images. Brief source dissolve remnants at12.00 show the preceding advertising setup; they do not eliminate the subsequent continuous visible change. 3. Main19.00–29.00: A tool edge on the left meets a moving curved metal surface. A ribbon is being peeled away at that contact and forms a tall spiral. Around21.25–22.25 the long ribbon changes position, detaches/falls out of the central area, leaving smaller fragments; subsequent ribbon curls develop again around24.00–28.75. The changing curl shape, separation and renewed shaving are visible. This is materially specific action, not simply a stock image of shiny roughness. It begins mid-process rather than before first tool contact; consequently the causal inference relies on seeing the ribbon emerge at the fixed tool/work interface, not on a complete machine-start sequence. Return and endpoint comparison: Main29.00–37.00 returns to the same operator,clock554,tabletop and carton-folding task. A previously assembled carton is behind a new one being worked on; the man completes/lifts another. The on-screen 'another carton' qualification matches visible repetition rather than pretending the first object has recovered or been restored. This is an honest return to a process, not evidence about what happened to the exact first sample. Main37.00–45.00 places carton, darkened sheet and metal curl in three panels. The carton recalls the opening task unmistakably; the sheet recalls the advancing dark front; the curl recalls the tool interface and falling ribbon. I can relate these endpoints to causes I actually saw in the main film. The control offers the same sorts of images as long frozen holds and does not permit that causal reading from its own moving evidence. Main therefore beats control for material response rather than texture. This comparison establishes a constructive effect of showing processes; it is not an audience-perception experiment. The endpoint labels are mostly defensible shorthand but deserve precise limits. 'Folded' and 'darkened' describe visible states. 'Curled/severed' combines the curled fragment state with a separation observed earlier. The chosen metal panel still shows a curl at the working interface, not a final isolated pile of severed residue after machining stops. The wet panel retains the original 'MORE THAN PAPER' sales wording. These are recorded frame selections, not proof that each specimen's change persisted after removal of force/liquid/tool. The screen explicitly says separate samples,held frames and no common recovery test; that honest qualification materially matters. Source audit: Carton source25–70 visibly contains an 'ORIGINAL METHOD' card followed by the same operator,clock554 and repeated cartons. At roughly27–40 the first action and next carton begin; subsequent repetitions continue through the inspected range. The original compilation also includes an 'IMPROVED METHOD' section and many unrelated industrial tasks in the sparse overview. The main uses the folding task's actual actions; the contextual survey does not support calling this a controlled elasticity test. Bounty source is plainly a commercial: actors,branded packaging, product name, comparative brand display and sales text appear across the complete minute. Around21–28 it contains the liquid-drop/sheet contact and growing dark patch used by the main; around27–30 the sales text enters. Later footage compares Bounty against a leading brand. Main does not import that performance comparison, and its wetting shot is explicitly labeled a staged advertising demonstration. It shows a real recorded visual darkening sequence, but neither the ad nor this re-edit warrants a neutral scientific ranking. The retained sales phrase on the final panel can momentarily reinstate the ad's rhetoric despite the new caption. Source346 is a continuous close view of the same tool/metal/ribbon relationship. At0–3seconds the tall shaving separates/falls; around4–18 new curls continue to grow and shift. The main preserves that sequence's observable contact,response and shedding. The short file has no contextual title in the inspected images, so the exact alloy,tool material and broader film identity are not independently verified here. I do not infer them solely from the end-credit title. Likewise I have not independently verified the proposed Matter reference; this is an archival material-response study regardless of that attribution. Strongest gap: Main37–45: eight seconds of small static endpoint panels after a second carton repetition. The comparison is useful, but it becomes a captioned inventory after the initial recognition. More substantively, the metal selected frame shows ongoing curl production while 'severed' suggests residue, and the wet selected frame is dominated by original advertising text. The panels currently depend on memory/captions more than they need to. Observable repair / falsifiable comparison: Keep the three-panel comparison but choose the wet frame immediately before sales text appears (aroundmain16.0–16.25) and a metal frame just after the long shaving falls (around22.5–23.0), with remaining small severed fragments visible. Compare a4–5second endpoint hold with the current8seconds. If viewers still identify each prior contact and its distinct response, the shorter,cleaner panel is preferable. If they cannot recognize the metal residue without the curled thumbnail, use a small pair of metal states (curl then detached residue) rather than making the word 'severed' carry the event. Do not claim permanence or recovery from either version. Strengths and acceptance boundary: The spreading dark patch and ribbon's separation are especially concrete. Hands turning tabs and the second carton link deformation to practical assembly. Different source functions remain visible, and the explicit analogy/separate-sample notices prevent a dishonest same-object metamorphosis. There is no filmed cloth sample or shared tabletop experiment in this output; it succeeds as an archival adaptation. A modest image-based visual pass is appropriate. Native cadence,reading comfort,small-screen footer legibility and sound require actual playback and remain pending. ===== CRITIC_A_V02.txt ===== ALLWAYS061 — CRITIC A / V02 Decision: PASS for the narrow archival visual study of three action/response types and honestly separated analogies. V02 ending is materially more useful than V01. Strong opening-reframing/recovery criterion remains QUALIFIED/INCONCLUSIVE; whole-film normal-speed/contact-sound acceptance INCONCLUSIVE. Original tabletop experiment remains NOT PERFORMED. Scope: Inspected all four actual main V02 ordered sheets and all four actual control V02 ordered sheets over complete50-second films, plus a full-resolution actual-MP4 frame at37seconds showing the revised37–45 comparison. Earlier source audit and dense V01 action review apply to the visibly unchanged action sequences. No build/ledger read, no native real-time playback or audio listening. Builder description was not taken as proof; this report addresses the visible revision. Unchanged visual core: 2–12 carton folding/handling;12–17 paper contact with dark pool and growing patch;19–29 machine chip curling and separation remain distinguishable. Contact/response evidence is present in main and absent in held-control intervals.17–19 explicit different-material/sample card and29–37 another-carton label continue to prevent a false one-specimen transformation/reset. Main remains stronger than endpoint control for showing causes rather than texture or state alone. Revised ending37–45: Earlier/later columns now compare three pairs rather than merely lining up terminal-looking pictures. Source-clock labels, action labels, “selected states,” separate sources and recovery-not-tested text are legible in the full frame. The question “Which changes would remain after contact stops?” is phrased as a question, not as a recorded result. Wetting pair21.8/25.0source seconds is strong: earlier paper is pale beside the pool, later visibly darkened upward. This carries an observable temporal difference and removes the distracting native advertising slogan from the selected comparison images while retaining AD status in the row label. Cutting pair1.2/5.2source seconds is useful: earlier large curl at the tool; later that long curl is gone and the cutting region has changed. Together with19–29 observed action, this supports generation/separation. The still pair alone would not show the trajectory; main supplies it. Carton pair27.5/36.0source seconds is weakest: both selected views have the operator's hands engaged and carton walls already erected. The later view has moving/blurred hands and partial occlusion. An attentive viewer can locate different hand/flap positions, but not confidently identify a new persistent crease or unloaded recovery. The tighter native crop makes the carton larger without distorting it; it cannot supply an invisible post-contact state. Does the ending now reframe the opening? More usefully, but at a limited level. The opening carton task is now returned to a question about what an earlier/later record can and cannot demonstrate. That is an epistemic reframe: apparent shape after handling is not equivalent to measured persistence after release. The juxtaposition of clearly changed paper/metal with an ambiguously changed carton makes the absence of a common test apparent. It does not establish a new physical fact about the original carton or show that its deformation would remain after contact stops. If “changes how the opening action is understood” means asking a more precise question about evidence, V02 offers a qualified pass. If it means the spectator learns a demonstrable new material property from that opening action, the criterion remains INCONCLUSIVE. Do not collapse those meanings in the acceptance report. Main/control qualification: The same new plate appears in control. It gives the control more before/later information than V01's terminal catalog; control is therefore no longer devoid of all temporal difference. But held intervals still remove the contact trajectory and action: the plate cannot prove how the dark patch grew or the chip detached. Main remains perceptually/evidentially stronger for action–response. No claim is made that I experienced either timing at normal speed. Claims protection: “Selected states” fixes the V01 endpoint wording concern. Recovery not tested is appropriately explicit. Different-source/sample framing remains intact. A question about persistence is not a claim of irreversible change. No observed drying, elastic return, force measurement, chemical identity, common sample, same-original retest, or controlled recovery interval is supplied. No verified Matter reference is established; provisional status remains appropriate. Largest remaining gap and falsifiable repair: The opening/carton comparison's weak material landmark is the main remaining visual gap. If a stronger empirical reframe is required, find a source moment in the same carton task with hands removed and a particular flap/crease unobscured, pair it with the actual pre-action view, and ask a viewer to identify the changed landmark without being told what to see. If there is no such source evidence, preserve the present narrow/questioning scope rather than labeling a hand-position difference a residue. Further text changes alone cannot create a recovery experiment. New original same-sample observation is needed for the full tabletop theorem. Strengths to preserve: Before/later pairs add concrete temporal comparison. Wetting is the clearest evidence row. Cutting row depends on, and invites recall of, the observed action rather than mere shape. Specific source clocks make selected states inspectable. Question and recovery-not-tested framing distinguish curiosity from assertion. Analogy boundaries and another-carton label remain candid. Main shows the actual changes; control isolates the information loss from holding selected pictures. Acceptance: NARROW ARCHIVAL VISUAL STUDY PASS. V02 MORE USEFUL THAN V01. STRONG PHYSICAL REFRAME/RECOVERY CLAIM INCONCLUSIVE. ORIGINAL TABLETOP/SAME-ORIGINAL-SHEET RETEST NOT PERFORMED. SOUND/CONTACT/RECOVERY AND NORMAL-SPEED WHOLE-FILM ACCEPTANCE INCONCLUSIVE. This qualified acceptance must not be promoted to full prompt proof. ===== CRITIC_B_V02.txt ===== 061 V02 — CRITIC B REPAIR REVIEW Verdict: PASS for the visual contact/response study and for the revised earlier/later comparison. Recovery or permanence remains untested, as the actual screen now explicitly states. No full audiovisual pass. Method: Preserved my V01 review. Inspected actual V02 sheets2–3 covering26–49seconds and decoded an actual main MP4 frame at38.00seconds at native960x768. This assesses the changed37–45second comparison in the context of the prior4fps whole-V01 inspection and source audit. I have not densely re-decoded all unchanged V02 action or independently viewed its control again. Parent described the proposed repair before this assessment; no scripts,timeline,state ledger or other critic report read. No audio or native playback. The substantive repair works. The actual plate now has three earlier/later rows, not three supposedly stable end products. At37–45seconds the wet row shows a pale contact-edge beside the dark drop on the left and a substantial dark patch on the right. The distracting advertising headline is absent from the chosen later frame. The cutting row shows the tall attached spiral on the left, then the cleared central area and smaller cut pieces on the right after the long ribbon's departure. This directly addresses my earlier concern that a new active curl was being called a severed endpoint. 'WETTING / AD' preserves the advertising-source distinction visibly, rather than laundering the demonstration into neutral experimental evidence. The carton row links back to the opening operator,clock/task and shows changed positions of tabs/walls between selected source27.5 and36.0seconds. It is the least decisive pair: hands obscure parts of the near corner and carton orientation changes, so a viewer must distinguish manipulation/reorientation from the actual fold. It does not show an untouched flat blank becoming a box. The row label says 'FOLDING', which is appropriately narrower. The actual motion earlier in the main remains better evidence of local bending than these two small states alone. The original wetting and cutting source audit supports these visual state choices: the Bounty sequence visibly grows a dark patch after contact; the cutting sequence visibly sheds a long spiral before smaller pieces/new shavings appear. Source-time labels are visible and agree at coarse audit resolution with the cited regions. I have not independently certified their exact frame timestamps. The full ad provenance and lack of a common physical test remain as qualified in my V01 report. The question 'Which changes would remain after contact stops?' now asks the unresolved question rather than pretending the frames answer it. The immediately adjacent line 'RECOVERY NOT TESTED', together with 'SELECTED STATES / SEPARATE SOURCES', makes that limit explicit. Nothing on this actual plate establishes permanence,elastic recovery or comparative absorbency rates. Its function is to compare visible changes and open a question. That is honest and useful. Largest remaining defect in V02: the 'Earlier view' and 'Later view' headings are too far left. At native resolution the first starts over the row-label gutter and the second starts over the left image column, while the pictures occupy columns starting around190 and575pixels. This is a genuine mapping error, not an artistic disagreement: a viewer can associate 'Later view' with the first picture. Move the headings directly over their image columns. No picture substitution is needed for this repair. The eight-second37–45hold is now better motivated than V01 because there are six images,three source labels per side and a question/qualification to read. I would not demand the previously suggested4–5second duration without native playback: the revised information density has changed. Once headings are aligned, use actual playback to check that viewers can compare all three rows and read the qualification comfortably. The remaining weak carton comparison can be tested by asking what physical change each pair shows; if viewers report only a rotated box, choose states with more visible wall/tab angle difference or rely explicitly on the earlier moving action. Disposition: substantive visual repair accepted, with V02 heading alignment requiring correction. Core cause/response pass retained. The plate is a clearer comparison of recorded states and avoids unsupported end-state claims. Real-time rhythm,audio and recovery conclusions remain outside this inspection. ===== CRITIC_A_V03.txt ===== ALLWAYS061 — CRITIC A / V03 ADDENDUM TO V02 Actual main V03 MP4 frame at38.0seconds extracted and visually inspected. “Earlier view” now starts above the left image column (approximatelyx190); “Later view” starts above the right column (approximatelyx575). Each heading clearly governs the corresponding three-image column. Alignment repair PASS. Images, selected source times, row labels, recovery question and recovery-not-tested qualification visibly match the V02 comparison. No new material behavior or claim is introduced. I did not rewatch unchanged action passages or inspect the control's header separately in this addendum. V02 critical scope remains unchanged: narrow archival visual study PASS; stronger empirical opening reframe/recovery INCONCLUSIVE; original tabletop experiment not performed; native normal-speed/contact-sound/audiovisual acceptance INCONCLUSIVE. Carton pair remains the least informative material landmark. This layout repair does not resolve or worsen that evidential limitation. ===== CRITIC_B_V03.txt ===== 061 V03 — CRITIC B HEADER-FIX ADDENDUM Qualified core verdict: PASS for visual contact/material response and the earlier/later state comparison. No recovery,permanence or full audiovisual pass is certified. Actual inspection: Decoded main V03 MP4 at38.00seconds to its native960x768 frame and inspected WHAT_STAYS_V03_sheet3 covering39–49seconds. Did not rewatch unchanged actions or independently re-audit the V03 control; V01/V02 inspection limits and source qualifications remain unchanged. No audio/native real-time playback. The heading repair is visible and successful. 'Earlier view' now begins over the left picture column ataboutx190; 'Later view' begins over the right ataboutx575. They no longer map the right heading onto the left picture. Six state images,source times,advertising identification and recovery-not-tested qualification remain visible in the actual plate. The39–44second sampled comparison holds that layout, followed by the credits at45. The substantive V02 repair remains accepted: wet frame shows darkening without the sales headline,cutting frame shows the aftermath of long-ribbon separation,and the paired states point back to observed actions rather than asserting stable endpoints. Carton pair remains the least self-explanatory because hands/orientation compete with the fold; the earlier moving sequence supplies its stronger evidence. No further mandatory visual change identified. Reading comfort and actual cadence remain playback questions; recovery remains expressly untested.