{"id":"22765aef934fc06c7610ff0c9cba9038defc201d3541a6497b934a41dc685333","title":"Tim Brooks and coauthors argue in OpenAI that video models may become world simulators","argument":"The report was published on February 15, 2024. Tim Brooks and coauthors describe Sora, a video generation model trained on videos and images represented as spacetime patches. They argue that scaling video models is a promising path towards building simulators of the physical and digital world. Their case rests on examples in which Sora keeps people and objects consistent as scenes change, simulates simple interactions, and renders a Minecraft world while controlling its player. The authors also report limitations: Sora does not accurately model the physics of many basic interactions, and changes to objects after actions such as eating food are not always correct. They conclude that continued scaling is a promising path towards capable simulators, while noting failures such as incoherence in long samples and objects appearing spontaneously. The report does not include model or implementation details, and its examples do not establish how reliably Sora would simulate the physical world beyond the demonstrated cases.","topicIds":["artificial-general-intelligence"],"domain":"other","claimType":"attributed reading","limits":"Text as retrieved on November 10, 2026; originally published February 15, 2024. This is an archived copy. The authors say model and implementation details are not included in the report; the reading therefore cannot assess those details.","speaker":{"name":"OpenAI","roleAtSource":"Publisher"},"source":{"title":"Video generation models as world simulators","publisher":"OpenAI","format":"Article","url":"https://openai.com/index/video-generation-models-as-world-simulators","publishedAt":"2024-02-15T08:00:00.000Z","dateBasis":"Date supplied by the publisher’s feed."},"evidence":[{"text":"Tim Brooks and coauthors argue that scaling video generation models could be a path towards simulators of the physical world.","excerpt":"Our results suggest that scaling video generation models is a promising path towards building general purpose simulators of the physical world.","url":"https://openai.com/index/video-generation-models-as-world-simulators","relationship":"Source-stated"},{"text":"The authors say Sora can show some consistency across scene changes and simulate simple interactions, but these capabilities have limits.","excerpt":"These capabilities enable Sora to simulate some aspects of people, animals and environments from the physical world.","url":"https://openai.com/index/video-generation-models-as-world-simulators","relationship":"Source-stated"},{"text":"The authors report that Sora does not accurately model the physics of many basic interactions.","excerpt":"it does not accurately model the physics of many basic interactions, like glass shattering.","url":"https://openai.com/index/video-generation-models-as-world-simulators","relationship":"Source-stated"},{"text":"The report does not provide model or implementation details.","excerpt":"Model and implementation details are not included in this report.","url":"https://openai.com/index/video-generation-models-as-world-simulators","relationship":"Source-stated"},{"text":"Escape Meridian reads the examples as evidence of capabilities demonstrated in selected samples, not of reliable physical simulation beyond them.","excerpt":"Sora currently exhibits numerous limitations as a simulator.","url":"https://openai.com/index/video-generation-models-as-world-simulators","relationship":"Escape Meridian interpretation"}],"sourceMedia":{"status":"unchecked","sourceUrl":"https://openai.com/index/video-generation-models-as-world-simulators","checkedAt":null,"items":[]},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How reliably would Sora simulate physical interactions beyond the examples in the report, given the failures the authors describe?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"Tim Brooks and coauthors argue in OpenAI that video models may become world simulators","summary":"The report was published on February 15, 2024. Tim Brooks and coauthors describe Sora, a video generation model trained on videos and images represented as spacetime patches. They argue that scaling video models is a promising path towards building simulators of the physical and digital world. Their case rests on examples in which Sora keeps people and objects consistent as scenes change, simulates simple interactions, and renders a Minecraft world while controlling its player. The authors also report limitations: Sora does not accurately model the physics of many basic interactions, and changes to objects after actions such as eating food are not always correct. They conclude that continued scaling is a promising path towards capable simulators, while noting failures such as incoherence in long samples and objects appearing spontaneously. The report does not include model or implementation details, and its examples do not establish how reliably Sora would simulate the physical world beyond the demonstrated cases.","question":"How reliably would Sora simulate physical interactions beyond the examples in the report, given the failures the authors describe?","claims":[{"text":"Tim Brooks and coauthors argue that scaling video generation models could be a path towards simulators of the physical world.","excerpt":"Our results suggest that scaling video generation models is a promising path towards building general purpose simulators of the physical world.","support":"source-stated"},{"text":"The authors say Sora can show some consistency across scene changes and simulate simple interactions, but these capabilities have limits.","excerpt":"These capabilities enable Sora to simulate some aspects of people, animals and environments from the physical world.","support":"source-stated"},{"text":"The authors report that Sora does not accurately model the physics of many basic interactions.","excerpt":"it does not accurately model the physics of many basic interactions, like glass shattering.","support":"source-stated"},{"text":"The report does not provide model or implementation details.","excerpt":"Model and implementation details are not included in this report.","support":"source-stated"},{"text":"Escape Meridian reads the examples as evidence of capabilities demonstrated in selected samples, not of reliable physical simulation beyond them.","excerpt":"Sora currently exhibits numerous limitations as a simulator.","support":"interpretation"}],"limitations":["Text as retrieved on November 10, 2026; originally published February 15, 2024. This is an archived copy.","The authors say model and implementation details are not included in the report; the reading therefore cannot assess those details."],"mediaId":null,"correctionReason":null,"id":"22765aef934fc06c7610ff0c9cba9038defc201d3541a6497b934a41dc685333","publicationId":"188567d7-40d4-86c0-9dda-0b7222397f57","version":1,"sourceUrl":"https://openai.com/index/video-generation-models-as-world-simulators","sourceTitle":"Video generation models as world simulators","publisher":"OpenAI","publishedAt":"2024-02-15T08:00:00.000Z","originalPublished":{"value":"2024-02-15T08:00:00.000Z","precision":"instant","basis":"backfill:feed_published"},"firstObservedAt":"2026-11-10T16:44:51.214Z","createdAt":"2026-11-21T15:28:10.187Z","historical":true,"entityIds":["openai"],"topicIds":["artificial-general-intelligence"],"media":[],"sourceMedia":{"status":"unchecked","sourceUrl":"https://openai.com/index/video-generation-models-as-world-simulators","checkedAt":null,"items":[]},"editorialDecision":{"outcome":"accept","mediaId":null,"inputHash":"096e3681bdbe8d2ff9b9e4111a2923c9eda3af8238642da3b98e6a04b1b50994","rubricVersion":"unrecorded","provider":"typesafe","model":"jev-1.13.0","reasonCodes":["bridge_admitted_by_model"],"confidence":0.56,"decidedAt":"2026-11-10T16:44:50.731Z"},"provenance":{"textHash":"56e16a4b000f9f6bd90b619f0a2b0e2ab154fb0b9c959456833838c3d3857c5e","observationId":"733cf281c4158fab1dcd26b10a72deb62da0dbfb8257fad4520081a1a2be0287","retrievedAt":"2026-11-10T17:31:52.887Z","kind":"archive-copy","preparer":{"model":"openai/gpt-6-luna","runId":"c47bf9ff-96fa-d054-2fc4-572ea0cb6641"},"reviewer":{"model":"anthropic/claude-opus-5.5","runId":"975eb1f9-858a-2d52-e226-ecd59240536c"}},"audit":{"status":"passed","reviewedAt":"2026-11-21T15:23:39.096Z","model":"anthropic/claude-opus-5.5"}},"canonicalUrl":"https://escapemeridian.com/readings/tim-brooks-and-coauthors-argue-in-openai-that-video-models-may-22765aef","formats":{"html":"/readings/tim-brooks-and-coauthors-argue-in-openai-that-video-models-may-22765aef","markdown":"/readings/tim-brooks-and-coauthors-argue-in-openai-that-video-models-may-22765aef.md","json":"/readings/tim-brooks-and-coauthors-argue-in-openai-that-video-models-may-22765aef.json"},"attribution":"Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.","maintainedGuideLinks":[],"conditions":[],"readerContext":{"version":"reader-context-v1","attribution":"Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.","terms":[],"entities":[{"id":"openai","name":"OpenAI","kind":"organization","context":"OpenAI develops AI models and products and publishes research, release announcements and evaluations.","referenceUrls":["https://openai.com/news/"],"path":"/entities/openai","json":"/entities/openai.json","markdown":"/entities/openai.md"}]},"researchQuestions":[],"withdrawn":false,"references":{"schemaVersion":1,"target":{"type":"reading","id":"22765aef934fc06c7610ff0c9cba9038defc201d3541a6497b934a41dc685333","revision":"188567d7-40d4-86c0-9dda-0b7222397f57","revisionKind":"publication-id","title":"Tim Brooks and coauthors argue in OpenAI that video models may become world simulators","url":"https://escapemeridian.com/readings/tim-brooks-and-coauthors-argue-in-openai-that-video-models-may-22765aef"},"source":{"type":"source","id":"https://openai.com/index/video-generation-models-as-world-simulators","revision":"56e16a4b000f9f6bd90b619f0a2b0e2ab154fb0b9c959456833838c3d3857c5e","revisionKind":"sha256-source-text","title":"Video generation models as world simulators","url":"https://openai.com/index/video-generation-models-as-world-simulators"},"status":"available","items":[],"next":null,"scope":"Explicit citations in public readings to this reading. Shared topics are not citations. Other record types need an explicit reference adapter."},"revisions":[{"id":"22765aef934fc06c7610ff0c9cba9038defc201d3541a6497b934a41dc685333","title":"Tim Brooks and coauthors argue in OpenAI that video models may become world simulators","argument":"The report was published on February 15, 2024. Tim Brooks and coauthors describe Sora, a video generation model trained on videos and images represented as spacetime patches. They argue that scaling video models is a promising path towards building simulators of the physical and digital world. Their case rests on examples in which Sora keeps people and objects consistent as scenes change, simulates simple interactions, and renders a Minecraft world while controlling its player. The authors also report limitations: Sora does not accurately model the physics of many basic interactions, and changes to objects after actions such as eating food are not always correct. They conclude that continued scaling is a promising path towards capable simulators, while noting failures such as incoherence in long samples and objects appearing spontaneously. The report does not include model or implementation details, and its examples do not establish how reliably Sora would simulate the physical world beyond the demonstrated cases.","topicIds":["artificial-general-intelligence"],"domain":"other","claimType":"attributed reading","limits":"Text as retrieved on November 10, 2026; originally published February 15, 2024. This is an archived copy. The authors say model and implementation details are not included in the report; the reading therefore cannot assess those details.","speaker":{"name":"OpenAI","roleAtSource":"Publisher"},"source":{"title":"Video generation models as world simulators","publisher":"OpenAI","format":"Article","url":"https://openai.com/index/video-generation-models-as-world-simulators","publishedAt":"2024-02-15T08:00:00.000Z","dateBasis":"Date supplied by the publisher’s feed."},"evidence":[{"text":"Tim Brooks and coauthors argue that scaling video generation models could be a path towards simulators of the physical world.","excerpt":"Our results suggest that scaling video generation models is a promising path towards building general purpose simulators of the physical world.","url":"https://openai.com/index/video-generation-models-as-world-simulators","relationship":"Source-stated"},{"text":"The authors say Sora can show some consistency across scene changes and simulate simple interactions, but these capabilities have limits.","excerpt":"These capabilities enable Sora to simulate some aspects of people, animals and environments from the physical world.","url":"https://openai.com/index/video-generation-models-as-world-simulators","relationship":"Source-stated"},{"text":"The authors report that Sora does not accurately model the physics of many basic interactions.","excerpt":"it does not accurately model the physics of many basic interactions, like glass shattering.","url":"https://openai.com/index/video-generation-models-as-world-simulators","relationship":"Source-stated"},{"text":"The report does not provide model or implementation details.","excerpt":"Model and implementation details are not included in this report.","url":"https://openai.com/index/video-generation-models-as-world-simulators","relationship":"Source-stated"},{"text":"Escape Meridian reads the examples as evidence of capabilities demonstrated in selected samples, not of reliable physical simulation beyond them.","excerpt":"Sora currently exhibits numerous limitations as a simulator.","url":"https://openai.com/index/video-generation-models-as-world-simulators","relationship":"Escape Meridian interpretation"}],"sourceMedia":{"status":"unchecked","sourceUrl":"https://openai.com/index/video-generation-models-as-world-simulators","checkedAt":null,"items":[]},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How reliably would Sora simulate physical interactions beyond the examples in the report, given the failures the authors describe?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"Tim Brooks and coauthors argue in OpenAI that video models may become world simulators","summary":"The report was published on February 15, 2024. Tim Brooks and coauthors describe Sora, a video generation model trained on videos and images represented as spacetime patches. They argue that scaling video models is a promising path towards building simulators of the physical and digital world. Their case rests on examples in which Sora keeps people and objects consistent as scenes change, simulates simple interactions, and renders a Minecraft world while controlling its player. The authors also report limitations: Sora does not accurately model the physics of many basic interactions, and changes to objects after actions such as eating food are not always correct. They conclude that continued scaling is a promising path towards capable simulators, while noting failures such as incoherence in long samples and objects appearing spontaneously. The report does not include model or implementation details, and its examples do not establish how reliably Sora would simulate the physical world beyond the demonstrated cases.","question":"How reliably would Sora simulate physical interactions beyond the examples in the report, given the failures the authors describe?","claims":[{"text":"Tim Brooks and coauthors argue that scaling video generation models could be a path towards simulators of the physical world.","excerpt":"Our results suggest that scaling video generation models is a promising path towards building general purpose simulators of the physical world.","support":"source-stated"},{"text":"The authors say Sora can show some consistency across scene changes and simulate simple interactions, but these capabilities have limits.","excerpt":"These capabilities enable Sora to simulate some aspects of people, animals and environments from the physical world.","support":"source-stated"},{"text":"The authors report that Sora does not accurately model the physics of many basic interactions.","excerpt":"it does not accurately model the physics of many basic interactions, like glass shattering.","support":"source-stated"},{"text":"The report does not provide model or implementation details.","excerpt":"Model and implementation details are not included in this report.","support":"source-stated"},{"text":"Escape Meridian reads the examples as evidence of capabilities demonstrated in selected samples, not of reliable physical simulation beyond them.","excerpt":"Sora currently exhibits numerous limitations as a simulator.","support":"interpretation"}],"limitations":["Text as retrieved on November 10, 2026; originally published February 15, 2024. This is an archived copy.","The authors say model and implementation details are not included in the report; the reading therefore cannot assess those details."],"mediaId":null,"correctionReason":null,"id":"22765aef934fc06c7610ff0c9cba9038defc201d3541a6497b934a41dc685333","publicationId":"188567d7-40d4-86c0-9dda-0b7222397f57","version":1,"sourceUrl":"https://openai.com/index/video-generation-models-as-world-simulators","sourceTitle":"Video generation models as world simulators","publisher":"OpenAI","publishedAt":"2024-02-15T08:00:00.000Z","originalPublished":{"value":"2024-02-15T08:00:00.000Z","precision":"instant","basis":"backfill:feed_published"},"firstObservedAt":"2026-11-10T16:44:51.214Z","createdAt":"2026-11-21T15:28:10.187Z","historical":true,"entityIds":["openai"],"topicIds":["artificial-general-intelligence"],"media":[],"sourceMedia":{"status":"unchecked","sourceUrl":"https://openai.com/index/video-generation-models-as-world-simulators","checkedAt":null,"items":[]},"editorialDecision":{"outcome":"accept","mediaId":null,"inputHash":"096e3681bdbe8d2ff9b9e4111a2923c9eda3af8238642da3b98e6a04b1b50994","rubricVersion":"unrecorded","provider":"typesafe","model":"jev-1.13.0","reasonCodes":["bridge_admitted_by_model"],"confidence":0.56,"decidedAt":"2026-11-10T16:44:50.731Z"},"provenance":{"textHash":"56e16a4b000f9f6bd90b619f0a2b0e2ab154fb0b9c959456833838c3d3857c5e","observationId":"733cf281c4158fab1dcd26b10a72deb62da0dbfb8257fad4520081a1a2be0287","retrievedAt":"2026-11-10T17:31:52.887Z","kind":"archive-copy","preparer":{"model":"openai/gpt-6-luna","runId":"c47bf9ff-96fa-d054-2fc4-572ea0cb6641"},"reviewer":{"model":"anthropic/claude-opus-5.5","runId":"975eb1f9-858a-2d52-e226-ecd59240536c"}},"audit":{"status":"passed","reviewedAt":"2026-11-21T15:23:39.096Z","model":"anthropic/claude-opus-5.5"}}}],"audits":[{"publicationId":"188567d7-40d4-86c0-9dda-0b7222397f57","verdict":"passed","findings":[{"dimension":"attribution","status":"pass","explanation":"The reading attributes the report to Tim Brooks and coauthors, published by OpenAI, which matches the author list. Its claims about Sora are kept as the authors' assertions ('argue', 'report', 'say'). The text is a single technical report, not a listing.","excerpt":"Tim Brooks, Bill Peebles, Connor Holmes, Will DePue, Yufei Guo, Li Jing, David Schnurr, Joe Taylor, Troy Luhman, Eric Luhman, Clarence Ng, Ricky Wang, Aditya Ramesh"},{"dimension":"summary","status":"pass","explanation":"The summary leads with the actor and the thesis that scaling video models is a promising path to simulators of the physical and digital world. It gives the supporting examples and preserves the authors' stated limitations and conclusion. 'Spacetime patches' goes unexplained, but it is background rather than the central argument, so a general reader can still follow the reading. The question asks about reliability beyond the shown examples, which follows from the evidence without leading the reader.","excerpt":"We believe the capabilities Sora has today demonstrate that continued scaling of video models is a promising path towards the development of capable simulators of the physical and digital world"},{"dimension":"date","status":"pass","explanation":"The stated date, February 15, 2024, matches the dateline and the UTC day of originalPublished. The reading discloses that it is an archived copy. It uses no hindsight and presents no later events. The related-article list predates publication and is not relied on. The page was edited after publication (archive capture May 2024), but every cited passage is plainly part of the original report body.","excerpt":"February 15, 2024\nVideo generation models as world simulators"},{"dimension":"support","status":"pass","explanation":"Each claim matches its excerpt: the scaling thesis, the simulation capabilities, the physics limitations and the missing implementation details. The Minecraft and consistency examples are stated in the source. The final claim is explicitly labeled as Escape Meridian interpretation and keeps the gap between the demonstrated samples and reliable simulation. The edit flag does not affect these clearly original passages.","excerpt":"Sora can simultaneously control the player in Minecraft with a basic policy while also rendering the world and its dynamics in high fidelity."},{"dimension":"media","status":"pass","explanation":"No media was selected and no verified media candidates were supplied, so there is no media identity or relevance issue.","excerpt":null}],"createdAt":"2026-11-21T15:23:39.096Z","model":"anthropic/claude-opus-5.5"}],"correctionUrl":"https://escapemeridian.com/corrections/"}