# Storytelling in OpenAI's recent films Expanded fieldnotes edition · September 18, 2026 16 selected films · 739 sampled still positions · 8 useful caption tracks. Sampled visual and caption analysis; no continuous playback or listening pass. [Read the illustrated chapter](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/) ## Six structures for different jobs ### Make a capability matter Human difficulty → Specific help → A continuing goal Establish whose work is difficult before introducing the feature. Show the help at the scale actually demonstrated: a plan, a decision or an answer can be a useful result without claiming a transformed life or business. **Planning question:** What does the person need to do next? **Adaptation:** Open on a concrete difficulty from the recorded account. Connect the demo to that task and name the result it actually supplies. ### Connect several capabilities A recognizable task → Related changes → Review or handoff Keep an artifact, a project identity or a question recognizable as the interface changes. Each new surface should have a job in the same piece of work. **Planning question:** What will the audience recognize in the next scene? **Adaptation:** Carry real Poe content through the capabilities that belong to one task. Show the consequence of each action before introducing another. ### Explain technical machinery An obstacle → A needed mechanism → Inspectable evidence Introduce technical detail when the task needs it. A problem gives the audience a reason to care about the mechanism and a way to judge its contribution. **Planning question:** What difficulty makes this mechanism necessary? **Adaptation:** Let the speaker's concrete problem lead the sequence. Reveal the mechanism, then hold the evidence while its meaning is explained. ### Teach a workflow Preview the destination → Build and revise → Explain the next action An early result makes the procedure worth following. One meaningful revision then demonstrates that the audience can direct the result, rather than merely receive it. **Planning question:** What should the audience be able to do afterward? **Adaptation:** Preview a real outcome, demonstrate a specific change, and end on the next action that follows from the recorded explanation. ### Tell a retrospective Starting conditions → A turning point → Present work and a next step Give the account a change in understanding, a decision or a user encounter. Artifacts make that change concrete and keep a retrospective from becoming only a chronology. **Planning question:** What happened that changed the direction of the work? **Adaptation:** Use verified team artifacts and one consequential turn. Label dates when chronology matters and distinguish present evidence from future ambition. ### Answer an adoption objection State the doubt → Show the mechanism → Retain its conditions The audience may need an explanation of trust more than another output reveal. Make the mechanism inspectable and include the human work it depends on. **Planning question:** What would stop this audience from relying on the result? **Adaptation:** Name the strongest relevant objection, show how the current product handles it, and preserve any necessary review or limitation. ## Pull-Up with ChatGPT 2025-09-28 · Human situation / short ad [Official film](https://www.youtube.com/watch?v=IqC3YrIzbQw) · [Illustrated reading](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/films/pull-up/) **Question:** What would help this person take the next step? **What carries through:** The person and the pull-up bar remain the reference point when the interface arrives. **Ending:** A usable training plan, with the larger fitness goal still ahead. The opening makes the problem understandable without a product explanation. When the request and plan arrive, the audience already knows what they are for. Returning to the person keeps the practical goal larger than the interface. The small scale of the demonstrated help is part of the structure: the ad offers a next step in a continuing effort. ### [0:00–0:12](https://www.youtube.com/watch?v=IqC3YrIzbQw) · Establish the need **Observed sequence:** A person grips a pull-up bar and strains; the physical situation precedes any product copy. **Story function (interpretation):** Physical effort supplies the reason to care. The viewer can understand the difficulty before reading a prompt or learning a feature name. ### [0:14–0:26](https://www.youtube.com/watch?v=IqC3YrIzbQw&t=14s) · Make the help specific **Observed sequence:** A request and a structured training plan appear over the same outdoor setting. **Story function (interpretation):** The plan answers the situation established in the opening. Staying in the same setting connects the digital response to the person's immediate work. ### [0:28–0:29](https://www.youtube.com/watch?v=IqC3YrIzbQw&t=28s) · Return to the person **Observed sequence:** The person stands beneath the equipment with the ChatGPT end branding. **Story function (interpretation):** The closing image leaves the goal with its owner. It supports an invitation to keep trying; it is not a before-and-after demonstration of improved strength. **Proposed adaptation:** Establish the practical frustration, then show the exact help delivered. Keep the broader aspiration distinct from the demonstrated result. - Begin with a recognizable difficulty in the team's actual work. - Show the precise help delivered, then return to the task it serves. **Keep the claim at the right scale:** An adaptation that jumps from a generated plan to a claim of achieved impact would skip the work the source film leaves ahead. **Evidence limit:** No captions listed; visual reading only. No claim about athletic improvement or the soundtrack. ## This is Sora 2 2025-09-30 · Generated-output showcase [Official film](https://www.youtube.com/watch?v=lEcg6AJ6DVY) · [Illustrated reading](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/films/sora-2/) **Question:** What kinds of scenes can this showcase put in motion? **What carries through:** Different worlds repeat the same unit: a situation, an action and a consequence. **Ending:** A sense of breadth across selected examples. The examples vary in setting and visual style, but they are not presented only as a collection of isolated images. Each gets enough sampled screen time for something to happen. That gives the montage a repeatable editorial unit even without a single protagonist or task. The argument comes from differences between examples and continuity within each example. ### [0:00–0:02](https://www.youtube.com/watch?v=lEcg6AJ6DVY) · Name the frame of reference **Observed sequence:** A short product introduction appears on black. **Story function (interpretation):** The brief product introduction identifies the showcase before the film begins changing worlds. ### [0:04–0:22](https://www.youtube.com/watch?v=lEcg6AJ6DVY&t=4s) · Establish variety **Observed sequence:** A skating sequence gives way to an animated festival exchange. **Story function (interpretation):** Live-action-style skating and an animated exchange make the change of register apparent early. They set an expectation of range. ### [0:24–0:52](https://www.youtube.com/watch?v=lEcg6AJ6DVY&t=24s) · Let events resolve **Observed sequence:** Paddleboard, space-dog and horseback sequences each retain an action and its consequence across several samples. **Story function (interpretation):** The paddleboard, space-dog and horseback examples preserve actions and consequences across the samples. Viewers get small situations to follow, rather than only new compositions. ### [0:54–0:57](https://www.youtube.com/watch?v=lEcg6AJ6DVY&t=54s) · Gather the examples under one name **Observed sequence:** Product name and brand close the montage. **Story function (interpretation):** The closing brand connects the variety back to the product. It does not add evidence about how typical or repeatable the selected outputs are. **Proposed adaptation:** For a range-of-results segment, choose distinct outcomes and let each resolve before switching examples. - For a range-of-results segment, choose examples with meaningfully different outcomes. - Give each example an observable change before moving to the next. **Keep the claim at the right scale:** A broad output montage cannot establish reliability or typical quality without separate evidence. This still study also cannot assess sound or physical accuracy. **Evidence limit:** Caption retrieval failed. This does not assess synchronized sound, generation reliability or physical accuracy. ## Introducing ChatGPT Atlas 2025-10-21 · Product launch / escalating agency [Official film](https://www.youtube.com/watch?v=Ej6hnsQgV_c) · [Illustrated reading](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/films/atlas/) **Question:** How much of the work can a browser assistant take on? **What carries through:** Each chapter gives the assistant more responsibility in a recognizable browser context. **Ending:** A prepared cart that the person can inspect before checkout. The examples are different tasks, so their connection comes from an escalation in responsibility. Contextual answers establish that the assistant can work with a page. Editing and recall then make it useful across work already underway. The shopping sequence extends that logic to action. Its permission and checkout states matter because they make the boundary of delegation visible. ### [0:00–0:16](https://www.youtube.com/watch?v=Ej6hnsQgV_c) · Establish the environment **Observed sequence:** Product introduction leads into an identifiable browser. **Story function (interpretation):** The opening gives viewers a browser to recognize before asking them to follow several types of assistance. ### [0:20–0:40](https://www.youtube.com/watch?v=Ej6hnsQgV_c&t=20s) · Start with understanding **Observed sequence:** An assistant answers questions in the context of rental, learning, financial and flight pages. **Story function (interpretation):** Questions tied to current pages make context the first source of usefulness. The assistant is introduced through what the person is already looking at. ### [0:46–1:00](https://www.youtube.com/watch?v=Ej6hnsQgV_c&t=46s) · Extend across work **Observed sequence:** Inline email assistance is followed by recovery of previously viewed shopping pages. **Story function (interpretation):** Editing and recall move from understanding a page to helping with an ongoing activity. The sequence expands scope without immediately jumping to autonomous action. ### [1:04–1:26](https://www.youtube.com/watch?v=Ej6hnsQgV_c&t=64s) · Delegate with a visible boundary **Observed sequence:** A shopping request leads through browser permission and visible activity to a filled cart with a checkout control. **Story function (interpretation):** Permission, browser activity and the filled cart turn a general agent claim into a sequence the viewer can inspect. The checkout control keeps purchase completion outside the shown result. ### [1:28–1:37](https://www.youtube.com/watch?v=Ej6hnsQgV_c&t=88s) · Give a next action **Observed sequence:** A benefit statement, icon and download destination close the film. **Story function (interpretation):** The destination follows the demonstration. The audience has seen the proposed benefit before receiving a place to try it. **Proposed adaptation:** Order related capabilities by increasing responsibility; preserve authorization and inspectable completion states. - Order connected capabilities by the responsibility they take on. - Keep the permission and completion states that define what the system actually did. **Keep the claim at the right scale:** Removing the human boundary can make a prepared result look like a completed external action. **Evidence limit:** Retrieved automatic captions contain music labels and stray words; they were not treated as narration. The cart does not prove a purchase. ## 10 years. 2025-12-11 · Retrospective / learning story [Official film](https://www.youtube.com/watch?v=8JXwrVQQ4jw) · [Illustrated reading](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/films/ten-years/) **Question:** What changed the direction of the work? **What carries through:** Uncertain experiments lead to an unexpected line of research and then to wider activity. **Ending:** Present work and a sense that the project continues. The useful organizing device is a change in understanding. The early material establishes conditions in which the eventual direction was not obvious. The text-prediction passage gives the retrospective a turn, and the later montage shows a broader present. This reading follows the captioned account and the image sequence; it does not turn the montage into a verified chronology. ### [0:00–0:22](https://www.youtube.com/watch?v=8JXwrVQQ4jw) · Make the starting conditions legible **Observed sequence:** Early workplace imagery, a mission statement and research experiments establish the starting conditions. **Story function (interpretation):** Workspace and research material establish a beginning with uncertainty and experiments. That gives the later shift something to change. ### [0:24–0:40](https://www.youtube.com/watch?v=8JXwrVQQ4jw&t=24s) · Locate the turn **Observed sequence:** Text-prediction imagery, papers and team material accompany captions about following an unexpected research lead. **Story function (interpretation):** The captioned account of following a research lead supplies a causal hinge. Papers and team imagery give the abstract change visible material. ### [0:42–1:02](https://www.youtube.com/watch?v=8JXwrVQQ4jw&t=42s) · Show the changed scale of activity **Observed sequence:** Product, event and people imagery broaden the story; captions turn toward work still ahead. **Story function (interpretation):** Products, events and people widen the story after the turn. The captions leave room for work ahead, rather than making the anniversary a final resolution. ### [1:04–1:07](https://www.youtube.com/watch?v=8JXwrVQQ4jw&t=64s) · Mark the occasion **Observed sequence:** Anniversary and company branding close the sequence. **Story function (interpretation):** The anniversary ending identifies why this retrospective is being told now. Its function differs from a task demo's output reveal. **Proposed adaptation:** Build a retrospective around one turning point and what it changed. Use dated, verifiable team evidence. - Organize a team retrospective around a decision or discovery that changed the work. - Use dated artifacts to distinguish chronology from thematic montage. **Keep the claim at the right scale:** Placing archival images in sequence can imply an exact historical order that the evidence does not support. **Evidence limit:** Narrative wording is read from automatic captions, not checked by listening. Historical and scale claims remain the film's claims. ## Introducing Prism 2026-01-27 · Expert workflow / accumulating changes [Official film](https://www.youtube.com/watch?v=xnInEsaaj9c) · [Illustrated reading](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/films/prism/) **Question:** How does this workspace help a paper move forward? **What carries through:** The paper remains the place where each capability has an observable effect. **Ending:** A return to the document after editing, diagram, literature and checking tasks. A continuing artifact reduces the orientation cost of a feature tour. Even as the type of work changes, the paper gives viewers a stable place to look for consequences. The sequence can therefore accumulate meaning: this is the same research document receiving different kinds of help. The useful adaptation is continuity of content, not a requirement to reproduce the source interface. ### [0:04–0:28](https://www.youtube.com/watch?v=xnInEsaaj9c&t=4s) · Establish an editable object **Observed sequence:** Introduction and a paper/editor pairing lead into visible proposed text changes. **Story function (interpretation):** The editor and paper pairing gives the audience a starting artifact. Proposed changes introduce assistance as something with a visible effect on that object. ### [0:32–0:56](https://www.youtube.com/watch?v=xnInEsaaj9c&t=32s) · Change the kind of input **Observed sequence:** A diagram is supplied and the paper preview becomes the result surface. **Story function (interpretation):** The diagram expands what can enter the workflow while the paper preview remains the result surface. A familiar destination holds the new capability in context. ### [1:00–1:32](https://www.youtube.com/watch?v=xnInEsaaj9c&t=60s) · Connect research back to the artifact **Observed sequence:** A literature-search chapter connects a question, references and edits back to the same document. **Story function (interpretation):** Literature and references become part of a document task. The chapter has a reason to exist beyond showing another panel. ### [1:36–1:56](https://www.youtube.com/watch?v=xnInEsaaj9c&t=96s) · Return for inspection **Observed sequence:** A content-check chapter returns to the paper before the product end card. **Story function (interpretation):** The final content-check passage keeps evaluation in the workflow. Generation is followed by another look at the object being produced. **Proposed adaptation:** Keep one real artifact through successive tasks, with a visible change after each task. - Use one real Poe artifact across the capabilities that genuinely belong to the same task. - After each capability, show the specific change to that artifact. **Keep the claim at the right scale:** Four-second retained samples can establish this broad sequence, but cannot support claims about short intermediate states or exact transitions. **Evidence limit:** Retained fieldnotes evidence, inspected again here. Current media retry failed; full originals are absent. Four-second sampling cannot resolve short intermediate states. ## A first look at the Codex app 2026-02-02 · Product tour / workflow chapters [Official film](https://www.youtube.com/watch?v=0e-Brv-gS9Q) · [Illustrated reading](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/films/codex-app/) **Question:** What does working in this app involve? **What carries through:** Enter work, organize it, inspect it and use its results. **Ending:** A website result after a tour through several work modes and projects. This tour gives viewers a vocabulary for using the app. Its unity comes from familiar stages of work rather than a single continuous build. Review is a distinct chapter, with differences and a comment made visible before the film moves toward a result. That is useful for an all-hands explanation in which judgment is part of the product story. ### [0:02–0:30](https://www.youtube.com/watch?v=0e-Brv-gS9Q&t=2s) · Make entry approachable **Observed sequence:** The app introduction moves into task entry, multitasking and dictation. **Story function (interpretation):** Task entry, multitasking and dictation establish ways to begin. Product interaction appears early because the film is explaining an app. ### [0:32–1:28](https://www.youtube.com/watch?v=0e-Brv-gS9Q&t=32s) · Expand the work vocabulary **Observed sequence:** Worktrees, a commit flow, skills and automation are shown through several projects. **Story function (interpretation):** Worktrees, commits, skills and automation show different ways to organize and repeat work. The several projects make this a tour, not a continuous task record. ### [1:32–1:46](https://www.youtube.com/watch?v=0e-Brv-gS9Q&t=92s) · Give review its own passage **Observed sequence:** A review chapter exposes code differences and an explicit comment. **Story function (interpretation):** Visible differences and a comment make the person's inspection part of the story. The sequence does not reduce the workflow to submitting a prompt. ### [1:48–2:12](https://www.youtube.com/watch?v=0e-Brv-gS9Q&t=108s) · Move toward use **Observed sequence:** A portable window leads to a website result and the brand ending. **Story function (interpretation):** A portable window and website result give the tour a destination outside its earlier setup steps. They are closing examples, not proof that every preceding chapter belongs to this one build. **Proposed adaptation:** Choose the chapters the recorded argument needs, and give review and finished output their own beats. - Select only the work stages needed by the recorded argument. - Give inspection and the usable result enough space to be understood. **Keep the claim at the right scale:** Editing examples together can imply one continuous build. Preserve changes of project when they matter to the claim. **Evidence limit:** Retained evidence re-inspected after a fresh extraction was interrupted. No uninterrupted-task or real-time-speed inference. ## Codex Super Bowl film 2026-02-08 · Brand film / history of making [Official film](https://www.youtube.com/watch?v=aCN9iCXNJqQ) · [Illustrated reading](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/films/codex-super-bowl/) **Question:** Where does this tool fit in the human activity of making? **What carries through:** Repeated gestures connect people, disciplines and tools before the product appears. **Ending:** An invitation to build. The delay before the product reveal is doing narrative work. Human gestures establish a broad category of activity, and later computing and robotics material extend it. Codex arrives near the end as another tool in that category. This is a positioning argument; the film does not promise to teach a specific workflow. ### [0:00–0:18](https://www.youtube.com/watch?v=aCN9iCXNJqQ) · Establish the human activity **Observed sequence:** Pointing, reading, drawing and assembling establish recurring human gestures. **Story function (interpretation):** Pointing, reading, drawing and assembling create a recurring subject across changing images: people trying to understand and make things. ### [0:20–0:40](https://www.youtube.com/watch?v=aCN9iCXNJqQ&t=20s) · Widen the field **Observed sequence:** Computers, mathematics, chess and robotics widen the range of making. **Story function (interpretation):** Computing, mathematics and robotics extend that activity across disciplines. The changes of subject support breadth rather than progress through one task. ### [0:42–0:50](https://www.youtube.com/watch?v=aCN9iCXNJqQ&t=42s) · Place the product in that context **Observed sequence:** The film returns to a person and a laptop; the Codex interface is visible around 50 seconds. **Story function (interpretation):** Returning to a person and laptop narrows the montage before the app appears. The product inherits the context established by the preceding images. ### [0:52–1:00](https://www.youtube.com/watch?v=aCN9iCXNJqQ&t=52s) · Offer an invitation **Observed sequence:** Large invitation copy and the company mark close the film. **Story function (interpretation):** The ending asks the audience to identify with making. It does not need the same kind of artifact proof as a feature demonstration, because it has set a different expectation. **Proposed adaptation:** Use a brief human or team context sequence when it earns the product reveal; shorten it for an all-hands demonstration. - Use team or human context when it makes the product's purpose easier to understand. - For a short all-hands demo, retain only the context the audience needs. **Keep the claim at the right scale:** Copying the long delay into a brief feature explanation can spend the time needed to show the feature's result. **Evidence limit:** No captions listed; repeated gestures and sequence are observable, but soundtrack and precise match-cut mechanics are not assessed. ## The Floral Hire 2026-04-28 · Customer story / confidence and decisions [Official film](https://www.youtube.com/watch?v=gpj9YWbr3IQ) · [Illustrated reading](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/films/floral-hire/) **Question:** What becomes easier for this business owner? **What carries through:** The founder's judgment and daily work connect the personal context to practical examples. **Ending:** Reported confidence and an everyday problem addressed. The story gives a large aspiration a human scale. Personal context explains why expressing and examining ideas matters to the founder. The expansion account then adds a useful complication: the reported advice is to strengthen the existing business before expanding. The printer episode returns from strategy to ordinary work. Together they frame usefulness as better judgment and support across the day. ### [0:00–0:22](https://www.youtube.com/watch?v=gpj9YWbr3IQ) · Place the benefit in a business **Observed sequence:** A benefit-led caption opening gives way to the flower-rental business and its physical work. **Story function (interpretation):** The opening benefit is followed by flowers and physical work. That makes the setting more specific than a generic productivity claim. ### [0:24–0:46](https://www.youtube.com/watch?v=gpj9YWbr3IQ&t=24s) · Explain the personal stakes **Observed sequence:** The founder's captions describe difficulty expressing ideas and earlier sources of encouragement. **Story function (interpretation):** The captioned account of difficulty expressing ideas helps explain why conversational assistance would matter to this person. ### [0:48–1:26](https://www.youtube.com/watch?v=gpj9YWbr3IQ&t=48s) · Make a decision consequential **Observed sequence:** The account compares expansion options and reports advice to strengthen the existing business before expanding. **Story function (interpretation):** Comparing expansion options provides a concrete use for advice. The reported decision to wait gives the passage a turn instead of treating more activity as the only form of progress. ### [1:32–1:58](https://www.youtube.com/watch?v=gpj9YWbr3IQ&t=92s) · Return to the working day **Observed sequence:** A printer problem, device use and a reported fix bring the story back to daily work and confidence. **Story function (interpretation):** The printer problem supplies a smaller counterpart to the business decision. The closing confidence remains the founder's account of her experience. **Proposed adaptation:** Connect the change to someone's work and judgment. Show the decision it enabled, including a decision not to proceed. - Show the decision a capability helped someone make. - Include a decision to wait or change direction when that is the real result. **Keep the claim at the right scale:** Customer testimony can support an account of experience. It does not independently establish revenue, saved time or other measured business impact. **Evidence limit:** Automatic captions support the account; outcomes are customer testimony, not independently measured business impact. ## It's time to fly 2026-06-03 · Brand film / visual metaphor [Official film](https://www.youtube.com/watch?v=bJcA23ckzcY) · [Illustrated reading](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/films/time-to-fly/) **Question:** What possibility does the tool represent to people at work? **What carries through:** Ordinary work settings frame a surreal departure and a return to the same people. **Ending:** A return to work after a visual metaphor for possibility. The ordinary settings make the departure legible as a change of register. Product input sits between those settings and the surreal launch, linking the metaphor to the tool. The return then closes the excursion. Its narrative result is a way of imagining work; the film does not show a completed software artifact that could serve as product proof. ### [0:00–0:18](https://www.youtube.com/watch?v=bJcA23ckzcY) · Establish an ordinary world **Observed sequence:** Several people remain at work in everyday office and home settings. **Story function (interpretation):** Several people at work create recognizable starting conditions. Their settings give the later departure something concrete to transform. ### [0:20–0:32](https://www.youtube.com/watch?v=bJcA23ckzcY&t=20s) · Connect the tool to the departure **Observed sequence:** Typing, task entry and a model-setting control connect those settings to Codex. **Story function (interpretation):** Task entry and a setting control make Codex part of the sequence before the visual world changes. ### [0:34–1:18](https://www.youtube.com/watch?v=bJcA23ckzcY&t=34s) · Develop the metaphor **Observed sequence:** Ordinary objects and rooms give way to a surreal launch and flight sequence. **Story function (interpretation):** Rooms and objects giving way to flight expand the possible scale of the experience. The sampled images support a metaphorical reading, not a literal speed claim. ### [1:20–1:30](https://www.youtube.com/watch?v=bJcA23ckzcY&t=80s) · Return to the starting world **Observed sequence:** The same people return to their work settings before the end branding. **Story function (interpretation):** The same people and settings close the loop. The return is the resolution of the visual story; it is not evidence that a particular build has been completed. **Proposed adaptation:** A short metaphor can establish motivation, but retain a separate product demonstration when the audience needs proof. - A brief metaphor can establish motivation when it fits the speaker's argument. - Follow it with a real demonstration when the audience needs to assess a capability. **Keep the claim at the right scale:** A visual metaphor for possibility is not evidence of speed, reliability or finished work. **Evidence limit:** Automatic captions mix short dialogue with launch language; no listening, sound-cue or literal speed claim is made. ## Meet ChatGPT Work 2026-07-09 · Product launch / one project across outputs [Official film](https://www.youtube.com/watch?v=yRc5HcGJ-Cs) · [Illustrated reading](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/films/chatgpt-work/) **Question:** Can this project move from an idea to work the team can use? **What carries through:** The lamp project stays recognizable through research, imagery, a page, a deck and feedback. **Ending:** Revised project material and a handoff into continuing team work. The lamp identity is the film's main continuity device. A spreadsheet, generated image, product page and deck would be a disconnected feature list without the shared project. Revision is especially useful here: changing a chart demonstrates that the result can be directed after generation. Mobile requests and feedback then extend the project into the team's continuing work. ### [0:00–0:22](https://www.youtube.com/watch?v=yRc5HcGJ-Cs) · Give the project an identity **Observed sequence:** Introduction and dictation establish a lamp-product planning request. **Story function (interpretation):** The lamp request supplies a subject the viewer can recognize again. That reduces the cost of moving across several work surfaces. ### [0:24–0:50](https://www.youtube.com/watch?v=yRc5HcGJ-Cs&t=24s) · Accumulate related outputs **Observed sequence:** Connected context leads to a supplier spreadsheet, generated imagery and a product-page composition. **Story function (interpretation):** Supplier research, imagery and a page composition each do a different job for the same project. Their order suggests progress because their content remains connected. ### [0:52–1:12](https://www.youtube.com/watch?v=yRc5HcGJ-Cs&t=52s) · Make user control visible **Observed sequence:** The same project becomes a deck; a targeted request changes a chart. **Story function (interpretation):** A deck turns the work into something presentable, and the chart revision demonstrates a specific response to direction. The first output is not treated as the end of the task. ### [1:16–1:30](https://www.youtube.com/watch?v=yRc5HcGJ-Cs&t=76s) · Continue beyond the first session **Observed sequence:** A phone request, scheduled task and feedback lead to a further deck update. **Story function (interpretation):** Phone input, scheduling and feedback place the artifact in an ongoing workflow. A later deck update answers new input instead of simply repeating the reveal. ### [1:32–1:39](https://www.youtube.com/watch?v=yRc5HcGJ-Cs&t=92s) · Close on the work proposition **Observed sequence:** The work invitation and brand close the film. **Story function (interpretation):** The invitation follows a sequence of related actions. Its coherence depends on the project content carried through the film. **Proposed adaptation:** Carry recognizable content through input, analysis, output, correction and handoff. Give each surface a reason to appear. - Carry the same real content through analysis, creation and review. - Include one meaningful correction and show what changes because of it. **Keep the claim at the right scale:** The edited progress labels do not support a real-time completion claim. Integration should be demonstrated by related content, not just transitions between apps. **Evidence limit:** Captions unavailable. Visible progress durations indicate compressed work; screen changes do not establish real-time completion. ## Voice in ChatGPT Work 2026-08-24 · Single-task feature vignette [Official film](https://www.youtube.com/watch?v=ceBruD6v5Bk) · [Illustrated reading](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/films/voice/) **Question:** Why did the recipe become too spicy? **What carries through:** The recipe comparison stays specific enough for the final explanation to answer it directly. **Ending:** An ingredient-level explanation. Voice is an input method, but the film gives it a small job with a recognizable answer. Comparing recipe versions creates a reason to retrieve context. Identifying an ingredient difference gives that retrieval a payoff. Because the question is narrow, the audience can judge whether the shown response addresses it without needing an entire new workflow explained. ### [0:00–0:06](https://www.youtube.com/watch?v=ceBruD6v5Bk) · Introduce the mode **Observed sequence:** Brief copy and a mode change introduce voice use. **Story function (interpretation):** The brief opening establishes voice use, then moves quickly into a task. The feature name does not carry the explanation by itself. ### [0:08–0:18](https://www.youtube.com/watch?v=ceBruD6v5Bk&t=8s) · Ask something answerable **Observed sequence:** A spoken request appears as text: compare recipe versions and investigate excess spice. **Story function (interpretation):** A comparison between recipe versions gives the request a concrete object and a specific discrepancy to explain. ### [0:20–0:26](https://www.youtube.com/watch?v=ceBruD6v5Bk&t=20s) · Show relevant context **Observed sequence:** Progress, email material and a recipe document provide the work context. **Story function (interpretation):** Email material and a recipe document connect the answer to information used in the task. They have a role beyond filling a progress interval. ### [0:28–0:39](https://www.youtube.com/watch?v=ceBruD6v5Bk&t=28s) · Resolve at the right scale **Observed sequence:** An ingredient-level explanation resolves the question before the close. **Story function (interpretation):** The ingredient explanation answers the original question at a useful level of detail. A broad closing productivity claim is unnecessary to understand the result. **Proposed adaptation:** Choose a small question whose answer can be recognized in one sentence or one changed state. - Choose a question that the audience can recognize as answered. - Keep the final evidence close to the question's level of detail. **Keep the claim at the right scale:** The displayed worked-time label is longer than the depicted interval. The film supports the task structure, not a speed benchmark. **Evidence limit:** Automatic captions are not a listening check. The visible worked-time label exceeds the depicted work interval; this is not a speed benchmark. ## Build a Shareable Site 2026-08-28 · Presenter tutorial / preview, build, revise, share [Official film](https://www.youtube.com/watch?v=0qK3KqvtRrg) · [Illustrated reading](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/films/shareable-site/) **Question:** How do I get to a useful site and change it when I need to? **What carries through:** The early trip-site preview provides a destination for the later request and revision. **Ending:** A next action and a second context in which the workflow can be useful. The finished-site preview gives the audience a reason to follow the construction steps. The expense-tracker revision then answers a likely next concern: whether the result can be changed after it exists. Sharing and the work-dashboard example extend the lesson from making a page to using it. The order teaches purpose, control and application. ### [0:04–0:16](https://www.youtube.com/watch?v=0qK3KqvtRrg&t=4s) · Preview the destination **Observed sequence:** The presenter promises a usable site and previews a trip itinerary. **Story function (interpretation):** Showing the itinerary early makes the promised outcome concrete. Later actions can be understood as steps toward something already recognizable. ### [0:24–0:58](https://www.youtube.com/watch?v=0qK3KqvtRrg&t=24s) · Connect context to the request **Observed sequence:** Existing trip context and a detailed request lead to a site preview. **Story function (interpretation):** The existing trip material explains what the build draws from. The request is part of a task with supplied context, not an isolated incantation. ### [1:04–1:34](https://www.youtube.com/watch?v=0qK3KqvtRrg&t=64s) · Demonstrate control and explain sharing **Observed sequence:** An expense-tracker revision is shown, followed by privacy explanation and a sharing request. **Story function (interpretation):** Changing the expense tracker makes revision an explicit part of the workflow. The privacy explanation supplies conditions for the next action. ### [1:36–2:14](https://www.youtube.com/watch?v=0qK3KqvtRrg&t=96s) · Generalize after the first example **Observed sequence:** A launch-dashboard example extends the same idea to recurring work and a sites index. **Story function (interpretation):** The launch dashboard broadens applicability once the basic workflow is established. It gives a work audience a second reason to use the same idea. ### [2:16–2:20](https://www.youtube.com/watch?v=0qK3KqvtRrg&t=136s) · Close the lesson **Observed sequence:** The presenter and brand close the lesson. **Story function (interpretation):** The presenter and brand finish a guided explanation. The tutorial's composited presenter layout is a source-film choice, separate from the all-hands playback setup. **Proposed adaptation:** Preview the useful outcome, demonstrate one meaningful revision, and end with the audience's next action. - Preview the result when the steps would otherwise be hard to motivate. - Show one meaningful revision and explain the audience's next action. **Keep the claim at the right scale:** A request to share does not establish that recipients received or accessed the site. Preserve that distinction in an adapted demonstration. **Evidence limit:** Captions supply the explanation; no listening pass. The sharing request alone does not prove recipient access, and the film is edited. ## Introducing the Agents API 2026-09-10 · Technical explainer / obstacles earn mechanisms [Official film](https://www.youtube.com/watch?v=2YHa1vhnmK0) · [Illustrated reading](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/films/agents-api/) **Question:** What makes a production investigation difficult, and what handles each difficulty? **What carries through:** An incident investigation supplies reasons for tools, execution, filtering, delegation and context management. **Ending:** A report with findings, supporting material and proposed responses. The film organizes technical detail around work that needs doing. Infrastructure comes first as a burden, then an incident supplies a concrete task. Log volume earns a filtering explanation; larger work earns delegation and context management. This order makes the mechanisms easier to evaluate because each is attached to an obstacle and an eventual report. ### [0:00–0:26](https://www.youtube.com/watch?v=2YHa1vhnmK0) · Name the burden **Observed sequence:** Infrastructure burdens precede the announcement and a simple responsibility diagram. **Story function (interpretation):** The opening explains why the architecture matters before presenting its responsibility diagram. The diagram has a problem to organize. ### [0:28–0:58](https://www.youtube.com/watch?v=2YHa1vhnmK0&t=28s) · Give the machinery a task **Observed sequence:** An incident-investigation example motivates tools, instructions and execution-environment choices. **Story function (interpretation):** The investigation makes tool and execution choices concrete. Viewers can follow what the mechanisms are for. ### [1:00–1:44](https://www.youtube.com/watch?v=2YHa1vhnmK0&t=60s) · Introduce complexity when needed **Observed sequence:** Log volume motivates filtering; larger work motivates delegation and context management. **Story function (interpretation):** Large logs, delegated work and context limits create distinct obstacles. The shown token comparison belongs to one filtering example; it is evidence inside that example, not a general benchmark. ### [1:46–1:56](https://www.youtube.com/watch?v=2YHa1vhnmK0&t=106s) · Expose the investigation's output **Observed sequence:** A report shows findings, supporting material and proposed responses. **Story function (interpretation):** The report gathers findings and support into an inspectable result. It demonstrates what the investigation produced, while proposed responses remain further actions. ### [1:58–2:17](https://www.youtube.com/watch?v=2YHa1vhnmK0&t=118s) · Broaden from the example **Observed sequence:** The presenter generalizes the example and closes. **Story function (interpretation):** The closing explanation can generalize after the audience has a specific case to refer to. The case remains the basis of the demonstration. **Proposed adaptation:** Introduce a mechanism when the current task needs it; show the evidence or decision it produces. - Introduce technical machinery at the moment the current task needs it. - Show the evidence it produces and identify any action still required. **Keep the claim at the right scale:** An investigation report is not a demonstrated production repair. A selected numerical comparison is not a universal performance result. **Evidence limit:** Shown token totals belong to this example, not a general performance result. No API behavior was independently tested. ## Detecting wildfires with ChatGPT 2026-09-14 · Customer documentary / origin, field use, continuing mission [Official film](https://www.youtube.com/watch?v=PSnJBfa183Y) · [Illustrated reading](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/films/wildfires/) **Question:** How did this project develop, and what does it need to do in the field? **What carries through:** The sensor and the landscape connect the inventor's history to present installation work. **Ending:** An ambition for a network, grounded by a return to physical work. The physical object makes an expansive story easier to follow. It appears in the opening landscape, has an origin in the archival material, and returns in field conversations and installation. Different people contribute different kinds of testimony. User questions about interpretation and location also give the technical work a practical setting, without establishing that the wider ambition has already been achieved. ### [0:00–0:26](https://www.youtube.com/watch?v=PSnJBfa183Y) · Put the object in its environment **Observed sequence:** An inventor carries a sensor into the landscape while captions establish the problem and potential value. **Story function (interpretation):** Carrying the sensor into the landscape connects the device to the problem before its development is explained. ### [0:28–1:00](https://www.youtube.com/watch?v=PSnJBfa183Y&t=28s) · Give the object a history **Observed sequence:** An early detector, archive pictures and family/partner accounts explain the project's development. **Story function (interpretation):** An early detector, archival images and family or partner accounts make the project an evolving effort. They supply personal context, not technical validation. ### [1:02–1:30](https://www.youtube.com/watch?v=PSnJBfa183Y&t=62s) · Meet the conditions of use **Observed sequence:** Field tests and conversations with intended users connect devices, readable data and location questions. **Story function (interpretation):** Field tests and user conversations move the story from invention to practical questions about readable data and location. The sensor and interpretation have distinct roles. ### [1:32–2:01](https://www.youtube.com/watch?v=PSnJBfa183Y&t=92s) · Extend the ambition, return to work **Observed sequence:** The ambition widens to a network, then returns to installation work and personal purpose. **Story function (interpretation):** The network idea broadens the potential application. Installation and personal purpose then return the ending to a concrete activity that continues. **Proposed adaptation:** Give a real project a concrete object, an origin, a user encounter and a next step. - Give a project story a recurring object the audience can recognize. - Let real user encounters explain what the next stage of work requires. **Keep the claim at the right scale:** The detector supplies sensing; ChatGPT helps interpret data. Neither potential benefit nor personal testimony establishes detection accuracy or lives saved. **Evidence limit:** The film assigns sensing to the detector and interpretation to ChatGPT. Neither the film nor this review establishes detection accuracy or lives saved. ## Data agent in ChatGPT Work 2026-09-10 · Product launch / question to distributed action [Official film](https://www.youtube.com/watch?v=MSiAd36bGeQ) · [Illustrated reading](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/films/data-agent/) **Question:** Why is traffic strong while sneaker sales are flat? **What carries through:** The same business question survives analysis, presentation, customization and communication. **Ending:** A depicted recurring-update request and team-channel handoff. The film treats the answer as an intermediate result. Analysis first explains the question, then a deck and dashboard make that explanation usable in different settings. Revision demonstrates control over those artifacts. The team-channel handoff gives the chain a social destination: the analysis is something other people can act on. The initial question keeps those surfaces connected. ### [0:00–0:14](https://www.youtube.com/watch?v=MSiAd36bGeQ) · Establish a discrepancy **Observed sequence:** Audience labels lead to a specific question about traffic and flat sneaker sales. **Story function (interpretation):** Strong traffic and flat sales create a specific tension to explain. Audience labels lead into a question with more substance than a generic request for insights. ### [0:16–0:30](https://www.youtube.com/watch?v=MSiAd36bGeQ&t=16s) · Show how the explanation is supported **Observed sequence:** Context checks, conversion analysis and a report turn the question into an explanation and recommendations. **Story function (interpretation):** Context checks, conversion analysis and a report connect the eventual recommendation to visible work. ### [0:32–0:56](https://www.youtube.com/watch?v=MSiAd36bGeQ&t=32s) · Make the answer usable **Observed sequence:** The analysis becomes a deck and a dashboard that is revised and restyled. **Story function (interpretation):** The deck and dashboard change the form of the analysis for an audience. Revision and restyling make the presentation of the answer part of the workflow. ### [0:58–1:02](https://www.youtube.com/watch?v=MSiAd36bGeQ&t=58s) · Give the work a destination **Observed sequence:** A recurring update request is followed by a team-channel message. **Story function (interpretation):** The recurring-update request and channel message place the result in team communication. The depicted handoff closes the chain without proving future scheduling reliability. ### [1:04–1:11](https://www.youtube.com/watch?v=MSiAd36bGeQ&t=64s) · Return to the original proposition **Observed sequence:** The closing statement returns to the relationship between questions and action. **Story function (interpretation):** The closing relation between questions and action has a concrete example behind it. The earlier question still explains why these artifacts were made. **Proposed adaptation:** End a data demo on the decision or communication the data enables, after showing its supporting analysis. - Keep the original question recognizable as the output changes form. - End on the decision or communication the analysis enables. **Keep the claim at the right scale:** The numbers are part of a demonstration. They do not independently establish business performance, and a depicted scheduled update does not verify future delivery. **Evidence limit:** No captions listed. Demo figures and the depicted update do not independently verify business performance or scheduling reliability. ## Grounding analysis in a semantic layer 2026-09-18 · Trust explainer / objection, mechanism, conditions [Official film](https://www.youtube.com/watch?v=bbNTY8O0GTY) · [Illustrated reading](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/films/semantic-layer/) **Question:** How will the agent use the right business definitions? **What carries through:** The trust concern leads to an inspectable definitions document and the human work needed to maintain it. **Ending:** An explanation of the mechanism and the conditions for using it. The uncertainty belongs to the audience: a convincing answer still needs the right data and definitions. The document makes the proposed mechanism inspectable. The captioned requirement for expert review is part of the answer, rather than an incidental caveat. The discussion of reuse and continuing effort keeps trust tied to a working practice. ### [0:02–0:34](https://www.youtube.com/watch?v=bbNTY8O0GTY&t=2s) · State the adoption objection **Observed sequence:** Two presenters address how an agent can choose the right data when business definitions are scattered. **Story function (interpretation):** Scattered definitions and complex business data explain why answer generation alone would not settle the audience's concern. ### [0:36–1:00](https://www.youtube.com/watch?v=bbNTY8O0GTY&t=36s) · Make the mechanism inspectable **Observed sequence:** A definitions document supplies entities, metrics and filters; captions call for a knowledgeable team member to review it. **Story function (interpretation):** Entities, metrics and filters give the explanation a concrete document. The call for a knowledgeable person to review it identifies the human judgment the mechanism depends on. ### [1:02–1:38](https://www.youtube.com/watch?v=bbNTY8O0GTY&t=62s) · Retain the operating conditions **Observed sequence:** The discussion covers reusing existing definitions and the continuing effort required. **Story function (interpretation):** Reusing definitions and maintaining them keeps the account connected to existing team work. This is a process to operate, not a single result reveal. ### [1:40–1:52](https://www.youtube.com/watch?v=bbNTY8O0GTY&t=100s) · Separate the answer from the tease **Observed sequence:** A future-product tease and brand ending close the discussion. **Story function (interpretation):** The future-product reference closes the discussion, but the demonstrated explanatory value lies in the earlier mechanism and its conditions. **Proposed adaptation:** Name the audience's strongest objection, show how it is handled, and retain the necessary human work. - Name the strongest audience objection in plain language. - Show the mechanism that addresses it and retain the human work it requires. **Keep the claim at the right scale:** An example definitions document does not prove that arbitrary business questions will receive correct answers. **Evidence limit:** Automatic captions are not listening-verified. The document is an example, not proof that arbitrary data questions are answered correctly. ## What changes when the job changes? The sample supports several structures. Choosing between them is more useful than treating one opening, duration or kind of ending as a universal formula. ### An opening sets an expectation [The Codex app tour](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/films/codex-app/) reaches task entry near the beginning. Its audience is being shown what working in the app involves. [The Super Bowl film](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/films/codex-super-bowl/) spends most of its minute on people learning and making before the app appears around 0:50. That delay places the product in a human context. For an all-hands feature segment, spend opening time on the context the explanation needs. A brand-style introduction can be appropriate when the speaker is establishing purpose; it can crowd out evidence when the speaker is demonstrating a capability. This sample does not establish one ideal reveal time. ### Continuity can come from different things [Prism](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/films/prism/) keeps a paper as its recurring result surface. [ChatGPT Work](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/films/chatgpt-work/) carries a lamp project through several kinds of artifact. [Atlas](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/films/atlas/) instead joins different tasks by increasing delegated responsibility: understand, assist, remember, act. Write down what survives the transition before selecting its animation. It might be content, a question, a project identity or a rising level of responsibility. If nothing carries over, the next scene may need an explicit new chapter in the recorded argument. ### A correction can carry the story In [Work](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/films/chatgpt-work/), a chart changes after a targeted request. In [the site tutorial](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/films/shareable-site/), the expense tracker is revised after the initial build. Both make control visible. [The Floral Hire](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/films/floral-hire/) supplies a different kind of turn: the founder reports advice to strengthen the existing business before expanding. A meaningful decision gives a segment a change the audience can understand. Use the correction or choice that actually occurred. Progress may mean improving an artifact, narrowing a proposal, or deciding to wait. ### The ending needs the right kind of evidence | What the film establishes | A fitting endpoint | What that endpoint does not establish | | --- | --- | --- | | A narrow diagnostic task · [Voice](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/films/voice/) | An ingredient-level answer to the recipe question | General accuracy or real-time speed | | Preparation for an action · [Atlas](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/films/atlas/) | A cart ready for inspection and checkout | A completed purchase | | A technical investigation · [Agents API](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/films/agents-api/) | A report with findings and supporting material | A repaired production system | | A customer experience · [Floral Hire](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/films/floral-hire/) | A reported decision and everyday help | Independently measured commercial impact | | A trust mechanism · [Semantic layer](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/films/semantic-layer/) | Definitions and the conditions for reviewing them | Correct answers to arbitrary questions | | A brand proposition · [It's time to fly](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/films/time-to-fly/) | A return to ordinary work after a metaphorical departure | A finished software artifact | The distinction is editorial as well as evidential. Choose the endpoint before arranging the intervening shots. It tells you what the audience needs to see along the way. ### Two formats with a different promise [Sora 2](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/films/sora-2/) builds breadth through several small events. Each example has an action and a consequence; the setting can change because the event is the repeatable unit. It is useful for a range-of-results segment when the results are meaningfully different. The [Codex brand films](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/?format=brand#readings) use human context or metaphor to position a tool. Their payoff is an invitation or a way of imagining the work. When an all-hands claim depends on actual functionality, pair that motivation with a separate, faithful product demonstration. ## Build around the supplied performance Start with a timed transcript of the actual talking-head recording, including pauses. For each substantial point, identify the audience's question, the starting state, the visible change and the evidence that answers the question. A review, permission or limitation needs its own beat when it changes the meaning of the result. ### Give each visual a job | In the recorded argument | The presentation can show | Check during storyboard review | | --- | --- | --- | | A problem or unmet need | The current task, an incomplete artifact or a specific discrepancy | Can the audience recognize the need before the feature arrives? | | An explanation of how something works | The real mechanism in the current Poe UI, or a clearly labeled explanatory diagram | Does this behavior match the actual product? | | A change in the work | A result that preserves recognizable input or project content | Is the consequence visible, and is the task still identifiable? | | A correction or decision | The relevant revision, comparison, review or permission state | Can the audience see what human judgment changes? | | A conclusion | The artifact, answer or handoff that supports the spoken claim | Does the evidence answer the question at the right scale? | These are proposed planning choices, not an edit list. Choose the few references that fit the recorded argument and inspect the current Poe product before planning or capturing its demos. Use real UI and original demo material; the source films supply references, not production assets. ### Keep the two playback assets separate The talking head plays through OBS. The companion presentation plays through Google Slides and remains silent by default, so the spoken narration is not duplicated. A tutorial's presenter inset is a reference-film composition, not a reason to combine the all-hands assets. The presentation opens first. A recognizable content transition cues the manual switch to the OBS scene, which starts the talking-head recording. Measure the actual playback offset in rehearsal: the cue time and the time the recording begins can differ. The presentation should finish while the talking head's recorded wave and pause are still playing. Compare both files on the shared playback timeline. Opening lead-in and ending buffer are separate requirements, and exact lengths come from the supplied footage and rehearsal. ### Use one consolidated storyboard review Bring the timed storyboard, demo list, selected references, start cue, measured offset and ending buffer into the planned review. Resolve missing footage, material limitations and export requirements there before production. The worksheet below helps collect those decisions; it does not imply that footage has already been supplied or that a production storyboard has been approved. ## What this study covers This chapter is an interpretive reading of **16 selected official OpenAI films published from September 18, 2025 through September 18, 2026**. They were chosen for contrasts useful to this all-hands project: human stories, product workflows, technical explanation, a tutorial, brand films and an output showcase. The analysis inspected **739 timecoded still positions across 44 contact sheets**, spanning films with a combined duration of about 25 minutes. Useful English captions were read for **eight films**. This is sampled visual and caption analysis. It does not claim continuous playback, a listening pass or frame-accurate cut detection. Fourteen films have fresh samples at approximately two-second intervals, with a near-end sample where available. Prism uses retained four-second samples after a failed fresh acquisition. The Codex app tour uses retained two-second samples after an interrupted fresh extraction. Those retained sheets were inspected again for this study. A short action or intermediate state can fall between samples. ### How to read the evidence **Observed sequence** describes what appears in the sampled images or the available caption account. **Story function** is the author's interpretation of the role that passage plays. **Adaptation** proposes a use in this project. These are distinct kinds of statement; the adaptations are not official OpenAI production guidance. Beat ranges are approximate narrative intervals, not exact cut boundaries. The three illustrated samples on each film page locate specific pieces of visual evidence; they do not represent every reviewed position. The compact stills are cropped from the retained contact sheets without adding imagery. Their captions are analysis, not quotations from the films. Each film page links to its official YouTube source and to the starting time of every discussed interval. Watch the original to assess delivery, sound, motion and the intervening states. Automatic captions were not checked by listening. Atlas's retrieved automatic track contained music labels and stray words and was not counted as useful narration. ### What “last year” means here The channel's Videos, Shorts and Live listings were collected on September 18, 2026. Their union contained **805 distinct IDs**, including **461 with publication dates in the rolling year**. Forty-two newly found dates were retrieved that day; other inventory dates came from identified September 5 watch metadata. All 16 selected films had their publication dates and official channel identities checked again. That inventory is a discovery snapshot. It is not a historical archive of deleted or unlisted videos, an independently audited pagination endpoint, or a count of films watched for this chapter. The 16 readings do not estimate channel-wide prevalence, audience response or which approach “performs best.” Twelve subjects overlap the existing 100-film fieldnotes; four are newer September uploads. ### Storage and publication Source acquisitions were processed one at a time in disposable directories. Media, partial downloads, downloader metadata and loose extraction frames were removed after each completed or failed attempt. One early interrupted directory needed explicit cleanup; the collector was then extended to clean its current directory on termination. The retained research bundle is about 7.3 MB and contains no source videos. The largest recorded completed-video scratch directory was 13.2 MB. The public chapter adds only text, code and selected small stills from that research; it does not upload complete films or caption transcripts. Existing project assets were preserved. ### Use the references with their limits The source films and marks belong to their respective owners. The stills accompany commentary on particular sequences and are not a cleared asset library. Use the current Poe product and original material for the all-hands presentation. The [public reading ledger](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/data/readings.json) includes the observations, interpretations, adaptations and illustrated samples. A [Markdown edition](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/data/storytelling-study.md) makes the expanded readings easy to carry into planning. The [storyboard worksheet](https://openai-film-fieldnotes-work.pages.dev/videos/storytelling/worksheet/) connects the study to a real recording and its separate playback assets.