{ "review_date": "2026-09-04", "studies": [ { "id": "chapter-reset", "title": "A chapter card is a cut, not a slide deck", "slug": "atlas-intro", "start": 16.6, "end": 20.2, "poster": 18, "observation": "The browser leaves at approximately 16.900 seconds and returns at 19.900. The intervening white statement card lasts about 3.0 seconds. Dense samples distinguish the title interval from the surrounding product views.", "adaptation": "Use an original title-to-content transition as the recognizable OBS start cue. The required lead-in is set by the presenter's playback setup, not by this film's three-second card.", "sample_step": "Selected 0.1–0.4 second samples around the two boundaries; source about 30 fps.", "shots": [ 5 ] }, { "id": "hard-cut-crop", "title": "Sometimes the right zoom is a hard cut", "slug": "atlas-intro", "start": 22.5, "end": 24.8, "poster": 24, "observation": "The full property-page view changes directly to the sidebar crop at approximately 23.033 seconds. Frames either side show a discontinuity, rather than a long traveling camera move.", "adaptation": "Establish the full your product window first, then cut directly to the control the speaker is discussing. Retain a little source context at the edge of the crop.", "sample_step": "0.1-second samples around the boundary; source about 30 fps.", "shots": [ 6, 7 ] }, { "id": "push-settle", "title": "Move quickly; let the interaction breathe", "slug": "codex-first-look", "start": 3.4, "end": 5.2, "poster": 4.3, "observation": "The opening camera push is concentrated around 3.9–4.3 seconds. It settles on the project area before the interaction continues. This is an approximate sampled window, not an official animation token.", "adaptation": "Spend movement on arriving at a useful crop. Stop the camera while viewers read the control and watch the click. Avoid continuous decorative zooming.", "sample_step": "0.2-second samples across the push; source about 23.976 fps.", "shots": [ 2, 3 ] }, { "id": "gallery-to-device", "title": "Reveal where the results live", "slug": "chatgpt-images", "start": 47, "end": 51.5, "poster": 50, "observation": "The image montage resolves into a grid at 47.5–48.0 seconds. By the 48.5-second sample, the grid sits inside a phone interface; the Images home surface remains visible through 51.5. The preceding gallery is output artwork, not app chrome.", "adaptation": "After a short set of real results, return to the actual product surface that contains them. Do not make a device mockup imply functionality that was never demonstrated.", "sample_step": "0.5-second samples from 46 to 52 seconds.", "shots": [ 8, 9 ] }, { "id": "report-to-detail", "title": "Open the artifact before explaining its controls", "slug": "research-update", "start": 58.5, "end": 63, "poster": 62, "observation": "The result starts in the conversation, becomes a report view by 60.0 seconds, then fills more of the frame. The table of contents is open by 61.5 seconds. Product state changes and editorial reframing are separate events here.", "adaptation": "Show the finished artifact first, then its useful navigation or source controls. A whole-window view should precede a detail crop.", "sample_step": "0.5-second samples from 58 to 63 seconds.", "shots": [ 10, 11 ] }, { "id": "persistent-anchor", "title": "Keep one object through the context change", "slug": "work-voice", "start": 18, "end": 24, "poster": 22, "observation": "Between the 18.0 and 19.0-second samples, the voice orb shrinks and moves to the lower right. It stays recognizable as the desktop appears around 23.5 seconds. Changing on-screen work durations demonstrate editorial time compression; this is not a latency measurement.", "adaptation": "Keep a stable visual anchor when a real workflow changes surface. Preserve state continuity, but use your product's actual control and a clean capture desktop.", "sample_step": "0.5-second samples around the relocation plus samples through 24 seconds.", "shots": [ 4, 5 ] }, { "id": "targeted-revision", "title": "Show the target, the request, and the changed result", "slug": "chatgpt-work", "start": 64, "end": 72, "poster": 66, "observation": "A full slide editor at 64.0 seconds becomes a selected-chart close-up at 64.5. The wider editor returns by 67.5; a revised chart is visible by 70.5. The excerpt includes the completed visual change, not just the edit prompt.", "adaptation": "Keep the same object through before, action, and after. Frame the final change long enough for the spoken explanation; do not substitute an invented result.", "sample_step": "0.5-second samples from 64 to 72 seconds.", "shots": [ 15, 16, 17 ] }, { "id": "anchored-task-card", "title": "Keep the task frame; change the evidence", "slug": "agent-look", "start": 7, "end": 12, "poster": 10.5, "observation": "The homepage remains through the 8.5-second sample; results are visible by 9.0. The floating task card keeps its position while a pointer-adjacent explanation grows between 10.0 and 11.5. The surrounding composition remains stable across the content change.", "adaptation": "Keep the viewer oriented with a stable container, then attach a short callout to the relevant action without covering the result. Use a real your product state, not the source film's illustrated agent container.", "sample_step": "0.5-second samples from 7 to 12 seconds.", "shots": [ 3, 4 ] }, { "id": "code-selection-growth", "title": "Let the selected code carry the emphasis", "slug": "agent-plugins", "start": 70, "end": 76, "poster": 73, "observation": "The configuration is unselected at 70.0 seconds. Selection is visible by 71.5 and grows through the 74.0-second sample. The file tree and editor frame remain in place while the highlighted block changes.", "adaptation": "Use the actual editor selection to identify a readable block of relevant code. Keep enough surrounding structure to explain its location; do not add a decorative highlight that pretends to be a product feature.", "sample_step": "0.5-second samples from 70 to 76 seconds.", "shots": [ 14, 15 ] }, { "id": "recording-to-artifact", "title": "End the process on an inspectable artifact", "slug": "codex-record-replay", "start": 44, "end": 51, "poster": 49, "observation": "The completion instruction is being entered at 44.0 seconds and is submitted by 45.0. Processing states give way to skill cards and edited files by 49.0. Changing on-screen work durations reveal editorial compression; this is not a seven-second performance measurement.", "adaptation": "Finish a process demonstration on the saved object the audience can inspect. Show real status and preserve any validation caveats instead of making a fast edit imply instant or fully verified execution.", "sample_step": "0.5-second samples from 44 to 51 seconds.", "shots": [ 9, 10 ] }, { "id": "simulator-selection", "title": "Select the target before entering the instruction", "slug": "codex-ios", "start": 10, "end": 15, "poster": 13.5, "observation": "The phone view is unhighlighted at 10.0 seconds. An inspection overlay appears by 11.0, a tooltip is visible by 12.0, and an instruction input appears by 13.0. The phone stays in the same frame throughout the sequence.", "adaptation": "Show the native inspection target before the request so the audience understands what the instruction addresses. Use only selection and inspection behavior that the demonstrated product actually supports.", "sample_step": "0.5-second samples from 10 to 15 seconds.", "shots": [ 3, 4 ] }, { "id": "split-preview-reveal", "title": "Give the result a pane of its own", "slug": "codex-prototypes", "start": 51, "end": 55, "poster": 54, "observation": "The conversation fills the view at 51.0 seconds. By 52.0 it occupies a narrow left pane beside a blank right pane. The running calendar prototype is visible on the right by 53.5. The change reveals a generated prototype, not a shipped calendar feature of the host app.", "adaptation": "Let a real split view make room for the result while retaining the originating request. Label generated demo artifacts clearly and do not imply that their functionality belongs to your product itself.", "sample_step": "0.5-second samples from 51 to 55 seconds.", "shots": [ 9, 10 ] }, { "id": "permission-scope", "title": "Slow down for the access decision", "slug": "codex-apps", "start": 25, "end": 29.5, "poster": 28, "observation": "A title is present at 25.0 seconds. The dashboard sharing surface appears by 26.0, with a tighter crop by 27.0. The access menu is open by 28.0 and the selected scope has changed by 29.0.", "adaptation": "Keep permission decisions in the demonstration rather than cutting around them. Use safe recipients and authentic controls, and hold the chosen scope long enough to read.", "sample_step": "0.5-second samples from 25 to 29.5 seconds.", "shots": [ 8, 9 ] }, { "id": "chart-representation-change", "title": "Show a control changing the result", "slug": "codex-data-science", "start": 38, "end": 44, "poster": 42, "observation": "A line chart is visible at 38.0 seconds. Its representation menu is open by 40.0, the pointer reaches the bar option by 41.5, and bars with a tooltip appear by 42.0. The title and neighboring controls preserve continuity.", "adaptation": "Demonstrate a genuine state change from control to result. Retain the same chart context and use readable data; do not fabricate a transition between unrelated screenshots.", "sample_step": "0.5-second samples from 38 to 44 seconds.", "shots": [ 7, 8 ] }, { "id": "image-revision-result", "title": "Pair a local request with the changed image", "slug": "codex-campaign", "start": 23, "end": 29, "poster": 28, "observation": "The original image and its edit request remain visible through the 26.0-second sample. The input is gone by 26.5, and a wider changed composition appears by 27.0 and holds. Composition changes as well as the requested detail, so the excerpt does not establish a strictly local pixel edit.", "adaptation": "Keep the same creative subject through the request and result, then give the changed image a clean hold. Describe the actual extent of the revision rather than claiming that everything outside a small region stayed identical.", "sample_step": "0.5-second samples from 23 to 29 seconds.", "shots": [ 5, 6 ] }, { "id": "chart-to-chart-pan", "title": "Move between related details, then stop", "slug": "codex-earnings", "start": 65, "end": 69, "poster": 68, "observation": "The dashboard is wide at 65.0 seconds and cropped to a bar chart by 65.5. Motion toward the adjacent line chart is visible at 67.5; the new detail is settled by 68.0. These samples show the move and its endpoints, not an official easing curve.", "adaptation": "Move only between related details the narration is comparing, then stop so the audience can read. Preserve axes, units, and caveats; a polished chart is not independent validation of its claims.", "sample_step": "0.5-second samples from 65 to 69 seconds.", "shots": [ 12 ] }, { "id": "page-annotation-followup", "title": "Attach the follow-up to its source", "slug": "computer-use", "start": 30, "end": 35, "poster": 34, "observation": "The page heading is unselected at 30.0 seconds and selected by 31.0. A contextual input appears by 31.5 and its request grows through 34.5. The browser page remains visible. This excerpt shows the follow-up request, not its eventual result.", "adaptation": "Show the native contextual target with the instruction so the audience can connect them. Pair it with a separate verified result later if the narration promises an outcome.", "sample_step": "0.5-second samples from 30 to 35 seconds.", "shots": [ 7, 8 ] }, { "id": "detached-working-preview", "title": "Keep background work visible through a context change", "slug": "computer-pip", "start": 25, "end": 29, "poster": 27.5, "observation": "A small slide preview sits inside the conversation at 25.0 seconds. By 27.0 it is outside the main window near the companion control; its contents change by 28.0 while the main window recedes.", "adaptation": "Use a persistent preview only where your product actually supports it. Treat the small view as a continuity cue, not the sole evidence of a readable or completed result.", "sample_step": "0.5-second samples from 25 to 29 seconds.", "shots": [ 7 ] }, { "id": "table-selection-reframe", "title": "Crop to the selected object, then restore context", "slug": "work-documents", "start": 23, "end": 27.5, "poster": 25.5, "observation": "The document is wide at 23.0 seconds. A selected table fills the close-up by 23.5 and the contextual prompt is visible by 24.5. Pullback motion appears at 26.5, restoring the wider view by 27.0. The excerpt alone does not prove that every requested table change completed.", "adaptation": "Use selection to explain the object being edited, crop only as far as readability requires, and restore the surrounding document. Show a verified result separately when the claim depends on it.", "sample_step": "0.5-second samples from 23 to 27.5 seconds.", "shots": [ 6, 7 ] }, { "id": "phone-desktop-pair", "title": "Prove continuity with both devices in view", "slug": "work-phone", "start": 91, "end": 95, "poster": 94, "observation": "The phone opens the task around the 91.5–92.0-second samples, and its response is visible by 92.5. A desktop view begins fading in by 93.0; both are visible by 93.5, with the phone on the left. This is an edited comparison, not a synchronization benchmark.", "adaptation": "Use matching task content to establish cross-device continuity, and hold both views long enough to compare. Claim only the continuity the real product demonstrates, not an inferred sync speed.", "sample_step": "0.5-second samples from 91 to 95 seconds.", "shots": [ 14, 15 ] }, { "id": "list-to-calendar", "title": "Change the representation to answer a different question", "slug": "information-visual", "start": 44, "end": 49, "poster": 48, "observation": "A task list and decisions are visible at 44.0 seconds, followed by a request for a calendar view. A month grid enters below and fills the view around 47.5–48.0 as the page scrolls. This demonstrates a generated visual representation, not an update to an external calendar.", "adaptation": "Change representation when it makes the spoken point clearer, such as replacing a list with chronology. Label generated views accurately and keep the underlying tasks consistent.", "sample_step": "0.5-second samples from 44 to 49 seconds.", "shots": [ 4, 5 ] }, { "id": "physical-screen-bridge", "title": "Move from the physical action into screen detail", "slug": "codex-appshots", "start": 2, "end": 8, "poster": 7.5, "shots": [ 1, 2 ], "sample_step": "0.5-second samples from 2 to 8 seconds.", "observation": "The title remains at 3.0 seconds; a close screen is visible by 3.5 and the whole laptop by 4.0. The framing drops to the keyboard at 5.0, returns to the laptop by 6.5, and tightens toward the two app windows by 7.5. Motion-blurred samples mark the camera moves.", "adaptation": "Use one physical interaction to establish the human action, then settle on a readable screen. Do not keep moving while the audience needs to read." }, { "id": "deck-population-hold", "title": "Keep the workspace fixed while the artifact changes", "slug": "powerpoint-editing", "start": 70, "end": 77, "poster": 74, "shots": [ 4 ], "sample_step": "0.5-second samples from 70 to 77 seconds.", "observation": "A blank main slide remains through 73.0 seconds while the thumbnail rail and assistant text change. By 73.5, a formatted slide occupies the same main canvas and then holds. The stable window makes the artifact change easy to compare.", "adaptation": "Show the same workspace before and after generation. Give the finished slide a reading hold; do not turn the edited interval into a claim about generation speed." }, { "id": "plugin-access-menu", "title": "Keep the sharing decision in the edit", "slug": "team-plugins", "start": 18, "end": 25, "poster": 22, "shots": [ 2 ], "sample_step": "0.5-second samples from 18 to 25 seconds.", "observation": "The plugin page is unobstructed at 18.0 seconds. A sharing dialog is open by 18.5, the crop is tighter by 19.0, and access choices are expanded by 20.5. The dialog remains the focal point rather than being skipped.", "adaptation": "Hold the actual access scope and recipient controls long enough to read. Use safe demo identities and do not imply that a share action is harmless or automatic." }, { "id": "spreadsheet-summary-crop", "title": "Keep the data beside the completion summary", "slug": "spreadsheet-launch", "start": 18, "end": 24, "poster": 22, "shots": [ 3 ], "sample_step": "0.5-second samples from 18 to 24 seconds.", "observation": "The whole spreadsheet window is visible through 20.0 seconds. By 20.5, the crop retains the data grid and assistant column at larger scale. A completed change summary is visible by 22.0 and stays beside the affected sheet.", "adaptation": "Crop toward both the changed artifact and its explanation, not the explanation alone. Retain units and caveats; a completion message is not an independent data audit." }, { "id": "sidebar-editorial-callouts", "title": "Explain navigation with one callout at a time", "slug": "codex-import", "start": 44, "end": 52, "poster": 50, "shots": [ 4 ], "sample_step": "0.5-second samples from 44 to 52 seconds.", "observation": "The view moves closer to the sidebar by 44.5 seconds. A black rectangular outline surrounds the upper navigation group by 46.5; by 49.5, the outline instead surrounds the project group. The app layout itself remains stable.", "adaptation": "Use an unmistakably editorial outline to explain two regions in sequence. Do not make an added callout look like a native product control." }, { "id": "test-chapter-reset", "title": "Give verification its own chapter", "slug": "workspace-risk", "start": 68, "end": 76, "poster": 71, "shots": [ 3, 4, 5 ], "sample_step": "0.5-second samples from 68 to 76 seconds.", "observation": "The two-pane setup is visible through 69.5 seconds. A white title begins at the 70.0 sample and is complete by 70.5; the setup returns by 73.0. The title separates configuration from a test stage without redesigning the workspace.", "adaptation": "Use a short chapter reset before a genuine test or review. Show the test result later; the title itself is not proof that validation succeeded." }, { "id": "photo-prompt-to-phone", "title": "Carry the same input into the device reveal", "slug": "shopping-room", "start": 8, "end": 14, "poster": 12, "shots": [ 1, 2 ], "sample_step": "0.5-second samples from 8 to 14 seconds.", "observation": "A floating composer contains a chair photo at 8.0 seconds; its request is complete by 11.0. At 12.0, the same photo and request appear inside a phone on a colored field. A search state and then response text follow.", "adaptation": "Repeat the input content across the transition so the new framing feels continuous. Treat the colored field as presentation design, not a native app background." }, { "id": "comparison-axis-change", "title": "Change scroll direction only when the content does", "slug": "shopping-compare", "start": 24, "end": 32, "poster": 29, "shots": [ 3, 4 ], "sample_step": "0.5-second samples from 24 to 32 seconds.", "observation": "Vertical movement passes product summaries from 24.0 seconds. A comparison table is visible by 26.5. From 28.0 to 29.5, different product columns move horizontally through the same phone frame; the earlier columns return by 30.5.", "adaptation": "Establish the table before moving across columns. Hold comparable rows and labels, and avoid presenting the source's historical prices as current shopping advice." }, { "id": "prism-native-diff", "title": "Let native change marks do the emphasis", "slug": "prism-intro", "start": 16, "end": 24, "poster": 20, "shots": [ 2 ], "sample_step": "0.5-second samples from 16 to 24 seconds.", "observation": "A proofreading request sits below the dark document source at 16.0 seconds. A working state is visible by 16.5. The view shifts at 18.5, and red/green marked revisions with per-change controls are legible by 19.0 and remain visible.", "adaptation": "Use the real review interface to explain changed text. Keep acceptance controls visible and avoid treating proposed wording as scientifically verified." }, { "id": "embedded-map-detail", "title": "Expand the result without losing the next action", "slug": "chatgpt-apps", "start": 28, "end": 36, "poster": 33, "shots": [ 3 ], "sample_step": "0.5-second samples from 28 to 36 seconds.", "observation": "A phone conversation is wide at 28.0 seconds and contains a map by 29.0. By 31.0, the map fills a tighter crop with a follow-up input below it. A request is typed from 32.0 while the map remains visible; another request begins by 35.0.", "adaptation": "Keep the actionable input adjacent to the content it refines. Preserve the result's context rather than zooming to an isolated button." }, { "id": "payment-state-sequence", "title": "Show action, processing, and confirmation separately", "slug": "instant-checkout", "start": 18, "end": 26, "poster": 24, "shots": [ 3, 4 ], "sample_step": "0.5-second samples from 18 to 26 seconds.", "observation": "A product card is visible at 18.0 seconds and a payment sheet by 19.0. The payment action is isolated at 20.5, becomes a working state by 21.0, and gives way to a green confirmation mark by 23.0. A whole-phone confirmation view returns at 24.0.", "adaptation": "Keep consequential action and status changes distinct. Use a safe demo transaction and label simulation; the source edit does not verify a real purchase or its processing time." }, { "id": "deck-card-to-viewer", "title": "Open the deliverable, not just its download card", "slug": "agent-slideshows", "start": 66, "end": 74, "poster": 72, "shots": [ 4 ], "sample_step": "0.5-second samples from 66 to 74 seconds.", "observation": "The presenters are visible at 66.0 seconds. An attachment card appears at 67.0, its slide preview by 68.0, and a larger viewer by 71.5. The viewer includes a thumbnail rail by 72.0 before a human insert at 73.5.", "adaptation": "Let the audience inspect a delivered artifact after the completion message. Preserve the same cover image across card and viewer to maintain continuity." }, { "id": "ide-patch-review", "title": "Reveal the proposed patch before returning to the app", "slug": "shipping-ide", "start": 92, "end": 100, "poster": 98, "shots": [ 5 ], "sample_step": "0.5-second samples from 92 to 100 seconds.", "observation": "The linked-editor card shows an editing progress state through 95.5 seconds. A red/green diff and review/apply controls are visible by 96.0. The panel shifts upward as an explanation appears below by 99.0.", "adaptation": "Make the actual proposed change inspectable. Keep the later rerun separate; visible diff colors and a confident explanation are not test evidence." }, { "id": "canvas-code-to-preview", "title": "Keep the originating request beside the running preview", "slug": "canvas-prototyping", "start": 90, "end": 98, "poster": 94, "shots": [ 5, 6 ], "sample_step": "0.5-second samples from 90 to 98 seconds.", "observation": "The code view and preview control are close at 90.0 seconds. A wider split workspace shows a loading state by 91.0. The dashboard is visible by 93.0, and its chart scrolls upward by 96.5 while the conversation stays on the left.", "adaptation": "Use a real preview transition and keep the original request in view. Label the result as a generated prototype rather than a shipped feature of the host product." }, { "id": "selected-text-local-request", "title": "Anchor the request to the selected passage", "slug": "canvas-writing", "start": 82, "end": 90, "poster": 88, "shots": [ 4, 5 ], "sample_step": "0.5-second samples from 82 to 90 seconds.", "observation": "The text selection expands by 83.0 seconds and a contextual toolbar is present by 83.5. An input replaces that toolbar by 86.5, with a request typed directly under the selected passage. At 90.0, the view returns to the whole split workspace.", "adaptation": "Show selection, instruction, then context. Keep the target visible while entering the request and do not imply that the subsequent answer has been fact-checked." }, { "id": "blend-control-push", "title": "Move toward the parameter while retaining its sources", "slug": "sora-blend", "start": 46, "end": 54, "poster": 48, "shots": [ 3, 4 ], "sample_step": "0.5-second samples from 46 to 54 seconds.", "observation": "Both source previews and the blend panel are visible at 46.0 seconds. The panel is larger by 46.5 and settles by 47.0. The source filmstrips and influence curve remain visible through 54.0, with the pointer approaching the curve around 50.5.", "adaptation": "Enlarge a meaningful control while retaining enough source context to explain it. Do not infer an easing formula or deterministic output from the camera move." } ] }