{"id":"86f08b607409ae6a43b3649a40bdb73f3d66f2bc0c4e38c00619c5fd8caab0e0","title":"Erik Hoel argues in The Intrinsic Perspective that the Dumbo feather may be the point","argument":"Erik Hoel’s essay was published May 23, 2026. He calls OpenAI’s internal model’s disproof of an Erdős geometry conjecture historic: it is a mathematical contribution he considers genuinely new, and the proof needed human rewriting to meet mathematical standards. But Hoel argues that the achievement alone cannot show that the new model is more capable than earlier ones. A publicly available ChatGPT model reportedly reproduced the result with little prompting, while the full details of both efforts remain unknown. Hoel’s Dumbo metaphor captures his challenge to the announcement’s implied progress: perhaps the new model removed a “magic feather”—a prompt telling it the problem was already solved—rather than demonstrating a new level of ability. He allows that AI capabilities may keep increasing, but says this result does not decide the matter. He also leaves open whether AI will automate whole intellectual fields, approximate their work with persistently mediocre results, or automate only parts of them. His conclusion is a demand for controlled comparisons: organizations should test whether earlier models can reproduce celebrated discoveries, or the results may reflect attention and search devoted to a neglected problem rather than a new capability. For Hoel, that comparison is a scientific necessity, not a reason to dismiss the achievement.","topicIds":["artificial-general-intelligence"],"domain":"other","claimType":"attributed reading","limits":"Text as retrieved on October 23, 2026; originally published May 23, 2026. Hoel says the details of both the public-model reproduction and OpenAI’s process are incomplete; this reading cannot establish how much search, prompting, compute or other assistance was involved. The essay reports the model’s disproof and describes the proof as rewritten by humans; this evidence does not independently assess the mathematical result.","speaker":{"name":"Erik Hoel","roleAtSource":"Publisher"},"source":{"title":"But Dumbo could already fly?","publisher":"Erik Hoel","format":"Article","url":"https://www.theintrinsicperspective.com/p/dumbo-could-already-fly","publishedAt":"2026-05-23T12:28:22.000Z","dateBasis":"Date supplied by the publisher’s feed."},"evidence":[{"text":"Erik Hoel describes the disproof as a historic mathematical contribution, while noting that humans rewrote the proof to make it clear and up to standard.","excerpt":"It’s a historic moment. OpenAI released a disproof of a geometry conjecture first proposed by Paul Erdős 80 years ago, discovered by an unnamed internal model.","url":"https://www.theintrinsicperspective.com/p/dumbo-could-already-fly","relationship":"Source-stated"},{"text":"Hoel argues that the result does not by itself establish that the new model is more capable than earlier models; he says key details about the efforts remain unknown.","excerpt":"We don’t know everything about Xiao Ma’s result, but in general the rediscovery by ChatGPT 5.5 seemed unsurprising to many key players; and, importantly, we also don’t know everything about OpenAI’s results","url":"https://www.theintrinsicperspective.com/p/dumbo-could-already-fly","relationship":"Source-stated"},{"text":"Hoel says organizations should rigorously test whether previous models can replicate celebrated discoveries, so the results can provide information about new capabilities.","excerpt":"at minimum, we should demand that organizations like OpenAI and Anthropic rigorously control for whether their new much-hyped novel discoveries can actually be replicated with previously existing models. That’s a scientific necessity.","url":"https://www.theintrinsicperspective.com/p/dumbo-could-already-fly","relationship":"Source-stated"},{"text":"Hoel considers several possible futures for AI in intellectual work: full automation, full approximation, or partial automation.","excerpt":"Those three futures (full automation, full approximation, partial automation) would all look like extremely AI-bullish predictions five years ago","url":"https://www.theintrinsicperspective.com/p/dumbo-could-already-fly","relationship":"Source-stated"}],"sourceMedia":{"status":"unchecked","sourceUrl":"https://www.theintrinsicperspective.com/p/dumbo-could-already-fly","checkedAt":null,"items":[]},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"What controlled comparisons would show whether earlier models can reproduce the result, and how much of the discovery depends on search, prompting or other assistance?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"Erik Hoel argues in The Intrinsic Perspective that the Dumbo feather may be the point","summary":"Erik Hoel’s essay was published May 23, 2026. He calls OpenAI’s internal model’s disproof of an Erdős geometry conjecture historic: it is a mathematical contribution he considers genuinely new, and the proof needed human rewriting to meet mathematical standards. But Hoel argues that the achievement alone cannot show that the new model is more capable than earlier ones. A publicly available ChatGPT model reportedly reproduced the result with little prompting, while the full details of both efforts remain unknown. Hoel’s Dumbo metaphor captures his challenge to the announcement’s implied progress: perhaps the new model removed a “magic feather”—a prompt telling it the problem was already solved—rather than demonstrating a new level of ability. He allows that AI capabilities may keep increasing, but says this result does not decide the matter. He also leaves open whether AI will automate whole intellectual fields, approximate their work with persistently mediocre results, or automate only parts of them. His conclusion is a demand for controlled comparisons: organizations should test whether earlier models can reproduce celebrated discoveries, or the results may reflect attention and search devoted to a neglected problem rather than a new capability. For Hoel, that comparison is a scientific necessity, not a reason to dismiss the achievement.","question":"What controlled comparisons would show whether earlier models can reproduce the result, and how much of the discovery depends on search, prompting or other assistance?","claims":[{"text":"Erik Hoel describes the disproof as a historic mathematical contribution, while noting that humans rewrote the proof to make it clear and up to standard.","excerpt":"It’s a historic moment. OpenAI released a disproof of a geometry conjecture first proposed by Paul Erdős 80 years ago, discovered by an unnamed internal model.","support":"source-stated"},{"text":"Hoel argues that the result does not by itself establish that the new model is more capable than earlier models; he says key details about the efforts remain unknown.","excerpt":"We don’t know everything about Xiao Ma’s result, but in general the rediscovery by ChatGPT 5.5 seemed unsurprising to many key players; and, importantly, we also don’t know everything about OpenAI’s results","support":"source-stated"},{"text":"Hoel says organizations should rigorously test whether previous models can replicate celebrated discoveries, so the results can provide information about new capabilities.","excerpt":"at minimum, we should demand that organizations like OpenAI and Anthropic rigorously control for whether their new much-hyped novel discoveries can actually be replicated with previously existing models. That’s a scientific necessity.","support":"source-stated"},{"text":"Hoel considers several possible futures for AI in intellectual work: full automation, full approximation, or partial automation.","excerpt":"Those three futures (full automation, full approximation, partial automation) would all look like extremely AI-bullish predictions five years ago","support":"source-stated"}],"limitations":["Text as retrieved on October 23, 2026; originally published May 23, 2026.","Hoel says the details of both the public-model reproduction and OpenAI’s process are incomplete; this reading cannot establish how much search, prompting, compute or other assistance was involved.","The essay reports the model’s disproof and describes the proof as rewritten by humans; this evidence does not independently assess the mathematical result."],"mediaId":null,"correctionReason":null,"id":"86f08b607409ae6a43b3649a40bdb73f3d66f2bc0c4e38c00619c5fd8caab0e0","publicationId":"a746c28f-6ead-77a3-5420-55f56fd42503","version":1,"sourceUrl":"https://www.theintrinsicperspective.com/p/dumbo-could-already-fly","sourceTitle":"But Dumbo could already fly?","publisher":"Erik Hoel","publishedAt":"2026-05-23T12:28:22.000Z","originalPublished":{"value":"2026-05-23T12:28:22.000Z","precision":"instant","basis":"backfill:feed_published"},"firstObservedAt":"2026-10-23T19:21:57.700Z","createdAt":"2026-11-07T02:58:23.304Z","historical":true,"entityIds":[],"topicIds":["artificial-general-intelligence"],"media":[],"sourceMedia":{"status":"unchecked","sourceUrl":"https://www.theintrinsicperspective.com/p/dumbo-could-already-fly","checkedAt":null,"items":[]},"editorialDecision":{"outcome":"accept","mediaId":null,"inputHash":"ac378e5dcc7e0b90776e0946f0321af6c8a7b7744fe5cb47362c2c5031e57a8c","rubricVersion":"unrecorded","provider":"unrecorded","model":"anthropic/claude-opus-5.5","reasonCodes":["bridge_admitted_by_model","second_opinion_categorical"],"confidence":1,"decidedAt":"2026-10-23T19:21:55.141Z"},"provenance":{"textHash":"4bc589dd8aa13a6396703e4f0291f792c53fd332c5858d9584eca968c7ce250a","observationId":"d1078b392d775adc834f417c1ee9cc8193c08d72773a5a4a17d4b50efbf9a6fe","retrievedAt":"2026-10-23T19:23:37.566Z","kind":"publisher-page","preparer":{"model":"openai/gpt-6-luna","runId":"60d393f5-9f12-19cb-6251-8cbf2e3f864a"},"reviewer":{"model":"anthropic/claude-opus-5.5","runId":"ba4db379-eb1e-af9f-58fc-c4939cc79b50"}},"audit":{"status":"passed","reviewedAt":"2026-11-07T02:53:44.190Z","model":"anthropic/claude-opus-5.5"}},"canonicalUrl":"https://escapemeridian.com/readings/erik-hoel-argues-in-the-intrinsic-perspective-that-the-dumbo-86f08b60","formats":{"html":"/readings/erik-hoel-argues-in-the-intrinsic-perspective-that-the-dumbo-86f08b60","markdown":"/readings/erik-hoel-argues-in-the-intrinsic-perspective-that-the-dumbo-86f08b60.md","json":"/readings/erik-hoel-argues-in-the-intrinsic-perspective-that-the-dumbo-86f08b60.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":[],"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":"86f08b607409ae6a43b3649a40bdb73f3d66f2bc0c4e38c00619c5fd8caab0e0","revision":"a746c28f-6ead-77a3-5420-55f56fd42503","revisionKind":"publication-id","title":"Erik Hoel argues in The Intrinsic Perspective that the Dumbo feather may be the point","url":"https://escapemeridian.com/readings/erik-hoel-argues-in-the-intrinsic-perspective-that-the-dumbo-86f08b60"},"source":{"type":"source","id":"https://www.theintrinsicperspective.com/p/dumbo-could-already-fly","revision":"4bc589dd8aa13a6396703e4f0291f792c53fd332c5858d9584eca968c7ce250a","revisionKind":"sha256-source-text","title":"But Dumbo could already fly?","url":"https://www.theintrinsicperspective.com/p/dumbo-could-already-fly"},"status":"available","items":[],"next":null,"scope":"Explicit citations in public readings and public file attachments to this reading. Shared topics are not citations. Other artifact types need an explicit reference adapter."},"revisions":[{"id":"86f08b607409ae6a43b3649a40bdb73f3d66f2bc0c4e38c00619c5fd8caab0e0","title":"Erik Hoel argues in The Intrinsic Perspective that the Dumbo feather may be the point","argument":"Erik Hoel’s essay was published May 23, 2026. He calls OpenAI’s internal model’s disproof of an Erdős geometry conjecture historic: it is a mathematical contribution he considers genuinely new, and the proof needed human rewriting to meet mathematical standards. But Hoel argues that the achievement alone cannot show that the new model is more capable than earlier ones. A publicly available ChatGPT model reportedly reproduced the result with little prompting, while the full details of both efforts remain unknown. Hoel’s Dumbo metaphor captures his challenge to the announcement’s implied progress: perhaps the new model removed a “magic feather”—a prompt telling it the problem was already solved—rather than demonstrating a new level of ability. He allows that AI capabilities may keep increasing, but says this result does not decide the matter. He also leaves open whether AI will automate whole intellectual fields, approximate their work with persistently mediocre results, or automate only parts of them. His conclusion is a demand for controlled comparisons: organizations should test whether earlier models can reproduce celebrated discoveries, or the results may reflect attention and search devoted to a neglected problem rather than a new capability. For Hoel, that comparison is a scientific necessity, not a reason to dismiss the achievement.","topicIds":["artificial-general-intelligence"],"domain":"other","claimType":"attributed reading","limits":"Text as retrieved on October 23, 2026; originally published May 23, 2026. Hoel says the details of both the public-model reproduction and OpenAI’s process are incomplete; this reading cannot establish how much search, prompting, compute or other assistance was involved. The essay reports the model’s disproof and describes the proof as rewritten by humans; this evidence does not independently assess the mathematical result.","speaker":{"name":"Erik Hoel","roleAtSource":"Publisher"},"source":{"title":"But Dumbo could already fly?","publisher":"Erik Hoel","format":"Article","url":"https://www.theintrinsicperspective.com/p/dumbo-could-already-fly","publishedAt":"2026-05-23T12:28:22.000Z","dateBasis":"Date supplied by the publisher’s feed."},"evidence":[{"text":"Erik Hoel describes the disproof as a historic mathematical contribution, while noting that humans rewrote the proof to make it clear and up to standard.","excerpt":"It’s a historic moment. OpenAI released a disproof of a geometry conjecture first proposed by Paul Erdős 80 years ago, discovered by an unnamed internal model.","url":"https://www.theintrinsicperspective.com/p/dumbo-could-already-fly","relationship":"Source-stated"},{"text":"Hoel argues that the result does not by itself establish that the new model is more capable than earlier models; he says key details about the efforts remain unknown.","excerpt":"We don’t know everything about Xiao Ma’s result, but in general the rediscovery by ChatGPT 5.5 seemed unsurprising to many key players; and, importantly, we also don’t know everything about OpenAI’s results","url":"https://www.theintrinsicperspective.com/p/dumbo-could-already-fly","relationship":"Source-stated"},{"text":"Hoel says organizations should rigorously test whether previous models can replicate celebrated discoveries, so the results can provide information about new capabilities.","excerpt":"at minimum, we should demand that organizations like OpenAI and Anthropic rigorously control for whether their new much-hyped novel discoveries can actually be replicated with previously existing models. That’s a scientific necessity.","url":"https://www.theintrinsicperspective.com/p/dumbo-could-already-fly","relationship":"Source-stated"},{"text":"Hoel considers several possible futures for AI in intellectual work: full automation, full approximation, or partial automation.","excerpt":"Those three futures (full automation, full approximation, partial automation) would all look like extremely AI-bullish predictions five years ago","url":"https://www.theintrinsicperspective.com/p/dumbo-could-already-fly","relationship":"Source-stated"}],"sourceMedia":{"status":"unchecked","sourceUrl":"https://www.theintrinsicperspective.com/p/dumbo-could-already-fly","checkedAt":null,"items":[]},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"What controlled comparisons would show whether earlier models can reproduce the result, and how much of the discovery depends on search, prompting or other assistance?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"Erik Hoel argues in The Intrinsic Perspective that the Dumbo feather may be the point","summary":"Erik Hoel’s essay was published May 23, 2026. He calls OpenAI’s internal model’s disproof of an Erdős geometry conjecture historic: it is a mathematical contribution he considers genuinely new, and the proof needed human rewriting to meet mathematical standards. But Hoel argues that the achievement alone cannot show that the new model is more capable than earlier ones. A publicly available ChatGPT model reportedly reproduced the result with little prompting, while the full details of both efforts remain unknown. Hoel’s Dumbo metaphor captures his challenge to the announcement’s implied progress: perhaps the new model removed a “magic feather”—a prompt telling it the problem was already solved—rather than demonstrating a new level of ability. He allows that AI capabilities may keep increasing, but says this result does not decide the matter. He also leaves open whether AI will automate whole intellectual fields, approximate their work with persistently mediocre results, or automate only parts of them. His conclusion is a demand for controlled comparisons: organizations should test whether earlier models can reproduce celebrated discoveries, or the results may reflect attention and search devoted to a neglected problem rather than a new capability. For Hoel, that comparison is a scientific necessity, not a reason to dismiss the achievement.","question":"What controlled comparisons would show whether earlier models can reproduce the result, and how much of the discovery depends on search, prompting or other assistance?","claims":[{"text":"Erik Hoel describes the disproof as a historic mathematical contribution, while noting that humans rewrote the proof to make it clear and up to standard.","excerpt":"It’s a historic moment. OpenAI released a disproof of a geometry conjecture first proposed by Paul Erdős 80 years ago, discovered by an unnamed internal model.","support":"source-stated"},{"text":"Hoel argues that the result does not by itself establish that the new model is more capable than earlier models; he says key details about the efforts remain unknown.","excerpt":"We don’t know everything about Xiao Ma’s result, but in general the rediscovery by ChatGPT 5.5 seemed unsurprising to many key players; and, importantly, we also don’t know everything about OpenAI’s results","support":"source-stated"},{"text":"Hoel says organizations should rigorously test whether previous models can replicate celebrated discoveries, so the results can provide information about new capabilities.","excerpt":"at minimum, we should demand that organizations like OpenAI and Anthropic rigorously control for whether their new much-hyped novel discoveries can actually be replicated with previously existing models. That’s a scientific necessity.","support":"source-stated"},{"text":"Hoel considers several possible futures for AI in intellectual work: full automation, full approximation, or partial automation.","excerpt":"Those three futures (full automation, full approximation, partial automation) would all look like extremely AI-bullish predictions five years ago","support":"source-stated"}],"limitations":["Text as retrieved on October 23, 2026; originally published May 23, 2026.","Hoel says the details of both the public-model reproduction and OpenAI’s process are incomplete; this reading cannot establish how much search, prompting, compute or other assistance was involved.","The essay reports the model’s disproof and describes the proof as rewritten by humans; this evidence does not independently assess the mathematical result."],"mediaId":null,"correctionReason":null,"id":"86f08b607409ae6a43b3649a40bdb73f3d66f2bc0c4e38c00619c5fd8caab0e0","publicationId":"a746c28f-6ead-77a3-5420-55f56fd42503","version":1,"sourceUrl":"https://www.theintrinsicperspective.com/p/dumbo-could-already-fly","sourceTitle":"But Dumbo could already fly?","publisher":"Erik Hoel","publishedAt":"2026-05-23T12:28:22.000Z","originalPublished":{"value":"2026-05-23T12:28:22.000Z","precision":"instant","basis":"backfill:feed_published"},"firstObservedAt":"2026-10-23T19:21:57.700Z","createdAt":"2026-11-07T02:58:23.304Z","historical":true,"entityIds":[],"topicIds":["artificial-general-intelligence"],"media":[],"sourceMedia":{"status":"unchecked","sourceUrl":"https://www.theintrinsicperspective.com/p/dumbo-could-already-fly","checkedAt":null,"items":[]},"editorialDecision":{"outcome":"accept","mediaId":null,"inputHash":"ac378e5dcc7e0b90776e0946f0321af6c8a7b7744fe5cb47362c2c5031e57a8c","rubricVersion":"unrecorded","provider":"unrecorded","model":"anthropic/claude-opus-5.5","reasonCodes":["bridge_admitted_by_model","second_opinion_categorical"],"confidence":1,"decidedAt":"2026-10-23T19:21:55.141Z"},"provenance":{"textHash":"4bc589dd8aa13a6396703e4f0291f792c53fd332c5858d9584eca968c7ce250a","observationId":"d1078b392d775adc834f417c1ee9cc8193c08d72773a5a4a17d4b50efbf9a6fe","retrievedAt":"2026-10-23T19:23:37.566Z","kind":"publisher-page","preparer":{"model":"openai/gpt-6-luna","runId":"60d393f5-9f12-19cb-6251-8cbf2e3f864a"},"reviewer":{"model":"anthropic/claude-opus-5.5","runId":"ba4db379-eb1e-af9f-58fc-c4939cc79b50"}},"audit":{"status":"passed","reviewedAt":"2026-11-07T02:53:44.190Z","model":"anthropic/claude-opus-5.5"}}}],"audits":[{"publicationId":"a746c28f-6ead-77a3-5420-55f56fd42503","verdict":"passed","findings":[{"dimension":"attribution","status":"pass","explanation":"The reading correctly attributes the essay to Erik Hoel in The Intrinsic Perspective. It keeps the claims of Gowers, Litt, Ma and OpenAI attributed or marked as reported, and presents Hoel's positions as his argument.","excerpt":"100% pure human copium about OpenAI solving Erdős problems"},{"dimension":"summary","status":"pass","explanation":"The summary covers Hoel's full arc: the result is historic, it needed human rewriting, a public model reportedly reproduced it, and the Dumbo feather metaphor applies. It ends with his three futures and his demand for controlled comparisons with earlier models. The title reflects his 'feather after feather' musing. Terms are understandable without specialist background. The self-mocking 'copium' voice is somewhat muted, but the meaning is not distorted. The question follows directly from the essay.","excerpt":"Maybe that’s what progress on intelligence really is: just feather after feather being removed."},{"dimension":"date","status":"pass","explanation":"originalPublished is 2026-05-23T12:28Z. That matches the essay's dateline of May 23, 2026, and the reading states the date at day precision. The evidence is a publisher page, so no archive disclosure is needed. modifiedAt equals the publication time. The reading presents no hindsight or later events.","excerpt":"Erik Hoel\nMay 23, 2026"},{"dimension":"support","status":"pass","explanation":"Each claim is grounded in the source: the historic contribution and human rewriting, the unknowns about both efforts, the call for controls, and the three futures. Describing the problem as 'neglected' is a fair rendering of Hoel's 'attention-bottlenecked' framing. The reading adds no benefits, costs or forecasts beyond the source.","excerpt":"The AI’s insight behind the finding is elegant (although the proof needed re-writing by humans to be clear and up-to-standard)."},{"dimension":"media","status":"pass","explanation":"No media was selected, and no verified candidates were supplied, so there is no media identity or relevance issue.","excerpt":null}],"createdAt":"2026-11-07T02:53:44.190Z","model":"anthropic/claude-opus-5.5"}],"correctionUrl":"https://escapemeridian.com/corrections/"}