{"id":"fb8b0a580b848acc93e6889e0e9056b031e1514c15ddd5cbdaa9169ca1d1ebcd","title":"Research acceleration: The jagged frontier","argument":"OpenAI reports that coding agents are being used more intensively in its research organization, while their ability to complete work without human intervention varies by task. From March to October 2026, estimated success on tasks expected to take two to four hours rose from 23% to 46%; the report says humans remain central to AI development. OpenAI also finds a widening gap in agent use between median and heaviest users, and reports that Safety & Alignment workweeks increased by 77% from January to September. Researchers described faster implementation and data review, but also poor research ideas, code and data quality, and workflows that still require close human review. OpenAI tracks code changes and experiments as measures of research activity, and presents OpenAI ECI as a more direct measure of model capabilities. The report cautions that increases in activity do not necessarily mean faster research progress, and that its historical trends may not predict future AI-driven progress. OpenAI presents the report as a partial account of changing research practices and says it is not an intention to pursue full RSI. It plans to continue monitoring capability metrics to assess whether capabilities progress is accelerating due to RSI.","topicIds":["intelligence-and-discovery","work-and-agency"],"domain":"work","claimType":"attributed reading","limits":"The report describes internal OpenAI measurements; its task success figures are estimates, and task choices within duration categories may have changed over time. OpenAI notes potential accuracy issues with researchers’ self-reports and the ad-hoc nature of its survey sampling strategy. OpenAI describes its account as partial and says its historical trends and metrics may not predict future AI-driven progress.","speaker":{"name":"OpenAI","roleAtSource":"Publisher"},"source":{"title":"Research acceleration: The jagged frontier","publisher":"OpenAI","format":"Article","url":"https://openai.com/index/research-acceleration-jagged-frontier","publishedAt":"2026-11-18T08:00:00.000Z","dateBasis":"Date supplied by the publisher’s feed."},"evidence":[{"text":"OpenAI reports that estimated agent success on tasks estimated to require two to four hours of human work rose from 23% in March to 46% in October, while stating that humans remain central to AI development.","excerpt":"From March to October, estimated success rose from 23% to 46% for tasks estimated to require two to four hours of human work","url":"https://openai.com/index/research-acceleration-jagged-frontier","relationship":"Source-stated"},{"text":"OpenAI reports that Safety & Alignment documented workweeks increased by 77% from January to September, associated with new hires and existing researchers shifting toward those workstreams.","excerpt":"Safety and Alignment’s documented workweeks increased by 77% from January to September due to both new hires and existing researchers shifting towards these workstreams.","url":"https://openai.com/index/research-acceleration-jagged-frontier","relationship":"Source-stated"},{"text":"OpenAI says agent success varies across tasks, with higher estimated success for monitoring training runs than for writing code for inference and serving stacks.","excerpt":"the autonomous agent success rate was 79% for monitoring training runs, 35% for launching training runs, and 25% for writing code for inference and serving stacks.","url":"https://openai.com/index/research-acceleration-jagged-frontier","relationship":"Source-stated"},{"text":"OpenAI reports that researchers described both faster work with agents and persistent shortcomings that require human oversight.","excerpt":"Workflows require close human involvement to avoid low-quality outputs and wasted cycles","url":"https://openai.com/index/research-acceleration-jagged-frontier","relationship":"Source-stated"},{"text":"OpenAI cautions that increases in tracked research activity do not necessarily mean faster research progress, and that historical trends may not predict future AI-driven progress.","excerpt":"increases in these measures of research activity do not necessarily imply faster research progress.","url":"https://openai.com/index/research-acceleration-jagged-frontier","relationship":"Source-stated"},{"text":"OpenAI says the report is a partial account and that its reporting should not be read as an intention to pursue full RSI.","excerpt":"our reporting should not be read as an intention to pursue full RSI.","url":"https://openai.com/index/research-acceleration-jagged-frontier","relationship":"Source-stated"}],"sourceMedia":{"status":"found","sourceUrl":"https://openai.com/index/research-acceleration-jagged-frontier/","checkedAt":"2026-11-20T23:30:59.070Z","items":[{"type":"image","url":"https://images.ctfassets.net/kftzwdyauwt9/4GWchdof3kw2l5OtMIvGt4/56144e08b813d6af76ae93adc2cefc1b/introducing-medchembench-cover.png?w=1200&q=90&fm=webp","sourceUrl":"https://openai.com/index/research-acceleration-jagged-frontier/","attribution":"OpenAI","basis":"publisher-embed","availability":"verified","label":"MedChemBench: Benchmarking AI for medicinal chemistry — Cover","verifiedAt":"2026-11-20T23:31:00.020Z"}],"note":"Media declared by the original publisher; availability checked separately from the editorial reading."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"What would OpenAI’s continued monitoring need to show for it to determine that capabilities progress is accelerating due to RSI?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"Research acceleration: The jagged frontier","summary":"OpenAI reports that coding agents are being used more intensively in its research organization, while their ability to complete work without human intervention varies by task. From March to October 2026, estimated success on tasks expected to take two to four hours rose from 23% to 46%; the report says humans remain central to AI development. OpenAI also finds a widening gap in agent use between median and heaviest users, and reports that Safety & Alignment workweeks increased by 77% from January to September. Researchers described faster implementation and data review, but also poor research ideas, code and data quality, and workflows that still require close human review. OpenAI tracks code changes and experiments as measures of research activity, and presents OpenAI ECI as a more direct measure of model capabilities. The report cautions that increases in activity do not necessarily mean faster research progress, and that its historical trends may not predict future AI-driven progress. OpenAI presents the report as a partial account of changing research practices and says it is not an intention to pursue full RSI. It plans to continue monitoring capability metrics to assess whether capabilities progress is accelerating due to RSI.","question":"What would OpenAI’s continued monitoring need to show for it to determine that capabilities progress is accelerating due to RSI?","claims":[{"text":"OpenAI reports that estimated agent success on tasks estimated to require two to four hours of human work rose from 23% in March to 46% in October, while stating that humans remain central to AI development.","excerpt":"From March to October, estimated success rose from 23% to 46% for tasks estimated to require two to four hours of human work","support":"source-stated"},{"text":"OpenAI reports that Safety & Alignment documented workweeks increased by 77% from January to September, associated with new hires and existing researchers shifting toward those workstreams.","excerpt":"Safety and Alignment’s documented workweeks increased by 77% from January to September due to both new hires and existing researchers shifting towards these workstreams.","support":"source-stated"},{"text":"OpenAI says agent success varies across tasks, with higher estimated success for monitoring training runs than for writing code for inference and serving stacks.","excerpt":"the autonomous agent success rate was 79% for monitoring training runs, 35% for launching training runs, and 25% for writing code for inference and serving stacks.","support":"source-stated"},{"text":"OpenAI reports that researchers described both faster work with agents and persistent shortcomings that require human oversight.","excerpt":"Workflows require close human involvement to avoid low-quality outputs and wasted cycles","support":"source-stated"},{"text":"OpenAI cautions that increases in tracked research activity do not necessarily mean faster research progress, and that historical trends may not predict future AI-driven progress.","excerpt":"increases in these measures of research activity do not necessarily imply faster research progress.","support":"source-stated"},{"text":"OpenAI says the report is a partial account and that its reporting should not be read as an intention to pursue full RSI.","excerpt":"our reporting should not be read as an intention to pursue full RSI.","support":"source-stated"}],"limitations":["The report describes internal OpenAI measurements; its task success figures are estimates, and task choices within duration categories may have changed over time.","OpenAI notes potential accuracy issues with researchers’ self-reports and the ad-hoc nature of its survey sampling strategy.","OpenAI describes its account as partial and says its historical trends and metrics may not predict future AI-driven progress."],"mediaId":"58da586369f33999eb920dc2411502d48d9dc9ca37a4ab1a788d83ffd176938a","correctionReason":null,"id":"fb8b0a580b848acc93e6889e0e9056b031e1514c15ddd5cbdaa9169ca1d1ebcd","publicationId":"6c4440b7-ba46-4320-8988-6f147021eaf1","version":1,"sourceUrl":"https://openai.com/index/research-acceleration-jagged-frontier","sourceTitle":"Research acceleration: The jagged frontier","publisher":"OpenAI","publishedAt":"2026-11-18T08:00:00.000Z","originalPublished":{"value":"2026-11-18T08:00:00.000Z","precision":"instant","basis":"feed_entry"},"firstObservedAt":"2026-11-18T23:28:09.412Z","createdAt":"2026-11-20T23:32:34.254Z","historical":false,"entityIds":["openai"],"topicIds":["intelligence-and-discovery","work-and-agency"],"media":[{"id":"58da586369f33999eb920dc2411502d48d9dc9ca37a4ab1a788d83ffd176938a","kind":"article","url":"https://openai.com/index/research-acceleration-jagged-frontier","sourceUrl":"https://openai.com/index/research-acceleration-jagged-frontier","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-20T23:31:00.020Z","method":"acquired_source","contentType":"text/html"},"recommendationReason":"Recommended by Jev for understanding this source; checked by Luna."},{"id":"100e78a9c03e6b4c3628ffaaf865aa6c797174129d1457049f8e2de8f4fd6e05","kind":"image","url":"https://images.ctfassets.net/kftzwdyauwt9/4GWchdof3kw2l5OtMIvGt4/56144e08b813d6af76ae93adc2cefc1b/introducing-medchembench-cover.png?w=1200&q=90&fm=webp","sourceUrl":"https://openai.com/index/research-acceleration-jagged-frontier","label":"MedChemBench: Benchmarking AI for medicinal chemistry — Cover","provenance":"publisher_page","verification":{"status":"verified","checkedAt":"2026-11-20T23:31:00.020Z","method":"publisher_association_and_head","contentType":"image/webp","cacheUntil":1795217760020},"recommendationReason":"Observed on the acquired publisher source; availability checked."}],"sourceMedia":{"status":"found","sourceUrl":"https://openai.com/index/research-acceleration-jagged-frontier/","checkedAt":"2026-11-20T23:30:59.070Z","items":[{"type":"image","url":"https://images.ctfassets.net/kftzwdyauwt9/4GWchdof3kw2l5OtMIvGt4/56144e08b813d6af76ae93adc2cefc1b/introducing-medchembench-cover.png?w=1200&q=90&fm=webp","sourceUrl":"https://openai.com/index/research-acceleration-jagged-frontier/","attribution":"OpenAI","basis":"publisher-embed","availability":"verified","label":"MedChemBench: Benchmarking AI for medicinal chemistry — Cover","verifiedAt":"2026-11-20T23:31:00.020Z"}],"note":"Media declared by the original publisher; availability checked separately from the editorial reading."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v9","mediaId":"58da586369f33999eb920dc2411502d48d9dc9ca37a4ab1a788d83ffd176938a","confidence":0.9500000000000001,"provider":"typesafe","model":"jev-1.13.0","inputHash":"15fbf8401294cc2f1e8871b3ad2734ce59043d8c91548d191dad441268cbb93c","decidedAt":"2026-11-20T23:31:02.628Z"},"provenance":{"textHash":"e789a8b8455677437398cd0b333ca7b693317f40b30b31daf4ec8e5739ad266f","observationId":"6f8df0f0d1588443556b5e32fc8a9333d53ff27fa198e95e52e57ec47e79ae21","retrievedAt":"2026-11-20T23:30:59.070Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"9bcde391-4f76-4435-8795-14f1396b84b8"},"reviewer":{"model":"gpt-6-luna","runId":"f8313c70-4920-47d0-824d-ef9343eb53a0"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-20T23:32:34.254Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":null},"canonicalUrl":"https://escapemeridian.com/readings/research-acceleration-the-jagged-frontier-fb8b0a58","formats":{"html":"/readings/research-acceleration-the-jagged-frontier-fb8b0a58","markdown":"/readings/research-acceleration-the-jagged-frontier-fb8b0a58.md","json":"/readings/research-acceleration-the-jagged-frontier-fb8b0a58.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"},{"id":"time","name":"TIME","kind":"organization","context":"TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.","referenceUrls":["https://time.com/"],"path":"/entities/time","json":"/entities/time.json","markdown":"/entities/time.md"},{"id":"nature","name":"Nature","kind":"organization","context":"Nature publishes research, news, and commentary across the sciences.","referenceUrls":["https://www.nature.com/","https://www.nature.com/nature.rss"],"path":"/entities/nature","json":"/entities/nature.json","markdown":"/entities/nature.md"}]},"researchQuestions":[],"withdrawn":false,"references":{"schemaVersion":1,"target":{"type":"reading","id":"fb8b0a580b848acc93e6889e0e9056b031e1514c15ddd5cbdaa9169ca1d1ebcd","revision":"6c4440b7-ba46-4320-8988-6f147021eaf1","revisionKind":"publication-id","title":"Research acceleration: The jagged frontier","url":"https://escapemeridian.com/readings/research-acceleration-the-jagged-frontier-fb8b0a58"},"source":{"type":"source","id":"https://openai.com/index/research-acceleration-jagged-frontier","revision":"e789a8b8455677437398cd0b333ca7b693317f40b30b31daf4ec8e5739ad266f","revisionKind":"sha256-source-text","title":"Research acceleration: The jagged frontier","url":"https://openai.com/index/research-acceleration-jagged-frontier"},"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":"fb8b0a580b848acc93e6889e0e9056b031e1514c15ddd5cbdaa9169ca1d1ebcd","title":"Research acceleration: The jagged frontier","argument":"OpenAI reports that coding agents are being used more intensively in its research organization, while their ability to complete work without human intervention varies by task. From March to October 2026, estimated success on tasks expected to take two to four hours rose from 23% to 46%; the report says humans remain central to AI development. OpenAI also finds a widening gap in agent use between median and heaviest users, and reports that Safety & Alignment workweeks increased by 77% from January to September. Researchers described faster implementation and data review, but also poor research ideas, code and data quality, and workflows that still require close human review. OpenAI tracks code changes and experiments as measures of research activity, and presents OpenAI ECI as a more direct measure of model capabilities. The report cautions that increases in activity do not necessarily mean faster research progress, and that its historical trends may not predict future AI-driven progress. OpenAI presents the report as a partial account of changing research practices and says it is not an intention to pursue full RSI. It plans to continue monitoring capability metrics to assess whether capabilities progress is accelerating due to RSI.","topicIds":["intelligence-and-discovery","work-and-agency"],"domain":"work","claimType":"attributed reading","limits":"The report describes internal OpenAI measurements; its task success figures are estimates, and task choices within duration categories may have changed over time. OpenAI notes potential accuracy issues with researchers’ self-reports and the ad-hoc nature of its survey sampling strategy. OpenAI describes its account as partial and says its historical trends and metrics may not predict future AI-driven progress.","speaker":{"name":"OpenAI","roleAtSource":"Publisher"},"source":{"title":"Research acceleration: The jagged frontier","publisher":"OpenAI","format":"Article","url":"https://openai.com/index/research-acceleration-jagged-frontier","publishedAt":"2026-11-18T08:00:00.000Z","dateBasis":"Date supplied by the publisher’s feed."},"evidence":[{"text":"OpenAI reports that estimated agent success on tasks estimated to require two to four hours of human work rose from 23% in March to 46% in October, while stating that humans remain central to AI development.","excerpt":"From March to October, estimated success rose from 23% to 46% for tasks estimated to require two to four hours of human work","url":"https://openai.com/index/research-acceleration-jagged-frontier","relationship":"Source-stated"},{"text":"OpenAI reports that Safety & Alignment documented workweeks increased by 77% from January to September, associated with new hires and existing researchers shifting toward those workstreams.","excerpt":"Safety and Alignment’s documented workweeks increased by 77% from January to September due to both new hires and existing researchers shifting towards these workstreams.","url":"https://openai.com/index/research-acceleration-jagged-frontier","relationship":"Source-stated"},{"text":"OpenAI says agent success varies across tasks, with higher estimated success for monitoring training runs than for writing code for inference and serving stacks.","excerpt":"the autonomous agent success rate was 79% for monitoring training runs, 35% for launching training runs, and 25% for writing code for inference and serving stacks.","url":"https://openai.com/index/research-acceleration-jagged-frontier","relationship":"Source-stated"},{"text":"OpenAI reports that researchers described both faster work with agents and persistent shortcomings that require human oversight.","excerpt":"Workflows require close human involvement to avoid low-quality outputs and wasted cycles","url":"https://openai.com/index/research-acceleration-jagged-frontier","relationship":"Source-stated"},{"text":"OpenAI cautions that increases in tracked research activity do not necessarily mean faster research progress, and that historical trends may not predict future AI-driven progress.","excerpt":"increases in these measures of research activity do not necessarily imply faster research progress.","url":"https://openai.com/index/research-acceleration-jagged-frontier","relationship":"Source-stated"},{"text":"OpenAI says the report is a partial account and that its reporting should not be read as an intention to pursue full RSI.","excerpt":"our reporting should not be read as an intention to pursue full RSI.","url":"https://openai.com/index/research-acceleration-jagged-frontier","relationship":"Source-stated"}],"sourceMedia":{"status":"found","sourceUrl":"https://openai.com/index/research-acceleration-jagged-frontier/","checkedAt":"2026-11-20T23:30:59.070Z","items":[{"type":"image","url":"https://images.ctfassets.net/kftzwdyauwt9/4GWchdof3kw2l5OtMIvGt4/56144e08b813d6af76ae93adc2cefc1b/introducing-medchembench-cover.png?w=1200&q=90&fm=webp","sourceUrl":"https://openai.com/index/research-acceleration-jagged-frontier/","attribution":"OpenAI","basis":"publisher-embed","availability":"verified","label":"MedChemBench: Benchmarking AI for medicinal chemistry — Cover","verifiedAt":"2026-11-20T23:31:00.020Z"}],"note":"Media declared by the original publisher; availability checked separately from the editorial reading."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"What would OpenAI’s continued monitoring need to show for it to determine that capabilities progress is accelerating due to RSI?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"Research acceleration: The jagged frontier","summary":"OpenAI reports that coding agents are being used more intensively in its research organization, while their ability to complete work without human intervention varies by task. From March to October 2026, estimated success on tasks expected to take two to four hours rose from 23% to 46%; the report says humans remain central to AI development. OpenAI also finds a widening gap in agent use between median and heaviest users, and reports that Safety & Alignment workweeks increased by 77% from January to September. Researchers described faster implementation and data review, but also poor research ideas, code and data quality, and workflows that still require close human review. OpenAI tracks code changes and experiments as measures of research activity, and presents OpenAI ECI as a more direct measure of model capabilities. The report cautions that increases in activity do not necessarily mean faster research progress, and that its historical trends may not predict future AI-driven progress. OpenAI presents the report as a partial account of changing research practices and says it is not an intention to pursue full RSI. It plans to continue monitoring capability metrics to assess whether capabilities progress is accelerating due to RSI.","question":"What would OpenAI’s continued monitoring need to show for it to determine that capabilities progress is accelerating due to RSI?","claims":[{"text":"OpenAI reports that estimated agent success on tasks estimated to require two to four hours of human work rose from 23% in March to 46% in October, while stating that humans remain central to AI development.","excerpt":"From March to October, estimated success rose from 23% to 46% for tasks estimated to require two to four hours of human work","support":"source-stated"},{"text":"OpenAI reports that Safety & Alignment documented workweeks increased by 77% from January to September, associated with new hires and existing researchers shifting toward those workstreams.","excerpt":"Safety and Alignment’s documented workweeks increased by 77% from January to September due to both new hires and existing researchers shifting towards these workstreams.","support":"source-stated"},{"text":"OpenAI says agent success varies across tasks, with higher estimated success for monitoring training runs than for writing code for inference and serving stacks.","excerpt":"the autonomous agent success rate was 79% for monitoring training runs, 35% for launching training runs, and 25% for writing code for inference and serving stacks.","support":"source-stated"},{"text":"OpenAI reports that researchers described both faster work with agents and persistent shortcomings that require human oversight.","excerpt":"Workflows require close human involvement to avoid low-quality outputs and wasted cycles","support":"source-stated"},{"text":"OpenAI cautions that increases in tracked research activity do not necessarily mean faster research progress, and that historical trends may not predict future AI-driven progress.","excerpt":"increases in these measures of research activity do not necessarily imply faster research progress.","support":"source-stated"},{"text":"OpenAI says the report is a partial account and that its reporting should not be read as an intention to pursue full RSI.","excerpt":"our reporting should not be read as an intention to pursue full RSI.","support":"source-stated"}],"limitations":["The report describes internal OpenAI measurements; its task success figures are estimates, and task choices within duration categories may have changed over time.","OpenAI notes potential accuracy issues with researchers’ self-reports and the ad-hoc nature of its survey sampling strategy.","OpenAI describes its account as partial and says its historical trends and metrics may not predict future AI-driven progress."],"mediaId":"58da586369f33999eb920dc2411502d48d9dc9ca37a4ab1a788d83ffd176938a","correctionReason":null,"id":"fb8b0a580b848acc93e6889e0e9056b031e1514c15ddd5cbdaa9169ca1d1ebcd","publicationId":"6c4440b7-ba46-4320-8988-6f147021eaf1","version":1,"sourceUrl":"https://openai.com/index/research-acceleration-jagged-frontier","sourceTitle":"Research acceleration: The jagged frontier","publisher":"OpenAI","publishedAt":"2026-11-18T08:00:00.000Z","originalPublished":{"value":"2026-11-18T08:00:00.000Z","precision":"instant","basis":"feed_entry"},"firstObservedAt":"2026-11-18T23:28:09.412Z","createdAt":"2026-11-20T23:32:34.254Z","historical":false,"entityIds":["openai"],"topicIds":["intelligence-and-discovery","work-and-agency"],"media":[{"id":"58da586369f33999eb920dc2411502d48d9dc9ca37a4ab1a788d83ffd176938a","kind":"article","url":"https://openai.com/index/research-acceleration-jagged-frontier","sourceUrl":"https://openai.com/index/research-acceleration-jagged-frontier","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-20T23:31:00.020Z","method":"acquired_source","contentType":"text/html"},"recommendationReason":"Recommended by Jev for understanding this source; checked by Luna."},{"id":"100e78a9c03e6b4c3628ffaaf865aa6c797174129d1457049f8e2de8f4fd6e05","kind":"image","url":"https://images.ctfassets.net/kftzwdyauwt9/4GWchdof3kw2l5OtMIvGt4/56144e08b813d6af76ae93adc2cefc1b/introducing-medchembench-cover.png?w=1200&q=90&fm=webp","sourceUrl":"https://openai.com/index/research-acceleration-jagged-frontier","label":"MedChemBench: Benchmarking AI for medicinal chemistry — Cover","provenance":"publisher_page","verification":{"status":"verified","checkedAt":"2026-11-20T23:31:00.020Z","method":"publisher_association_and_head","contentType":"image/webp","cacheUntil":1795217760020},"recommendationReason":"Observed on the acquired publisher source; availability checked."}],"sourceMedia":{"status":"found","sourceUrl":"https://openai.com/index/research-acceleration-jagged-frontier/","checkedAt":"2026-11-20T23:30:59.070Z","items":[{"type":"image","url":"https://images.ctfassets.net/kftzwdyauwt9/4GWchdof3kw2l5OtMIvGt4/56144e08b813d6af76ae93adc2cefc1b/introducing-medchembench-cover.png?w=1200&q=90&fm=webp","sourceUrl":"https://openai.com/index/research-acceleration-jagged-frontier/","attribution":"OpenAI","basis":"publisher-embed","availability":"verified","label":"MedChemBench: Benchmarking AI for medicinal chemistry — Cover","verifiedAt":"2026-11-20T23:31:00.020Z"}],"note":"Media declared by the original publisher; availability checked separately from the editorial reading."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v9","mediaId":"58da586369f33999eb920dc2411502d48d9dc9ca37a4ab1a788d83ffd176938a","confidence":0.9500000000000001,"provider":"typesafe","model":"jev-1.13.0","inputHash":"15fbf8401294cc2f1e8871b3ad2734ce59043d8c91548d191dad441268cbb93c","decidedAt":"2026-11-20T23:31:02.628Z"},"provenance":{"textHash":"e789a8b8455677437398cd0b333ca7b693317f40b30b31daf4ec8e5739ad266f","observationId":"6f8df0f0d1588443556b5e32fc8a9333d53ff27fa198e95e52e57ec47e79ae21","retrievedAt":"2026-11-20T23:30:59.070Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"9bcde391-4f76-4435-8795-14f1396b84b8"},"reviewer":{"model":"gpt-6-luna","runId":"f8313c70-4920-47d0-824d-ef9343eb53a0"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-20T23:32:34.254Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":null}}],"audits":[{"publicationId":"6c4440b7-ba46-4320-8988-6f147021eaf1","verdict":"uncertain","findings":[{"dimension":"attribution","status":"pass","explanation":"The draft consistently attributes the measurements, survey results, cautions and plans to OpenAI as source claims rather than presenting them as established fact. The interview themes are correctly presented as what OpenAI reports its researchers described.","excerpt":"Today, we’re providing an update on the state of research acceleration at OpenAI, based on internal measurements"},{"dimension":"summary","status":"uncertain","explanation":"The central term RSI is never expanded or explained, yet both the summary's closing sentences and the question depend on it. A general reader cannot tell what RSI is or why OpenAI tracks it. The draft also omits OpenAI's stated reason for publishing, its view that labs should be required to report RSI indicators, which is the report's core significance. OpenAI ECI is likewise unexplained. Smallest repair: spell out recursive self-improvement (RSI), using the source's definition, and add OpenAI's stated public-reporting rationale. Also suggest a term-library tooltip for RSI and ECI.","excerpt":"recursive self-improvement (RSI)—the process by which AI accelerates the development of future AI systems."},{"dimension":"date","status":"pass","explanation":"The originalPublished date of 2026-11-18 matches the date shown on the publisher page. The draft's March–October, January–September and September–October figures match the source's scopes.","excerpt":"November 18, 2026"},{"dimension":"support","status":"pass","explanation":"Each claim matches the source text: the 23% to 46% success rates, the 77% increase in Safety and Alignment workweeks, the 79% vs 25% task-level rates, the need for close human involvement, the activity-versus-progress caution, and the disclaimer about not intending full RSI. The draft adds no unsupported benefits, costs or forecasts. Its limitations, including estimated rates, shifting task mix, self-report caveats and the partial picture, are drawn from the source.","excerpt":"changes in zero-intervention success rates within these buckets may reflect both capability improvements and researchers’ changing choices"},{"dimension":"media","status":"pass","explanation":"The selected media is the verified direct link to the original OpenAI report, which is the most relevant choice. The alternative image is the cover for an unrelated MedChemBench post and was correctly not chosen.","excerpt":null}],"createdAt":"2026-11-20T23:32:34.254Z","model":"anthropic/claude-opus-5.5"}],"correctionUrl":"https://escapemeridian.com/corrections/"}