{"id":"e095c2ec54370fb2a9a3bc5d97c472cf08f758c74076450ce0728af50508d6f1","title":"Ethan Mollick argues in One Useful Thing that AI’s shape is becoming clearer","argument":"Ethan Mollick published this essay on March 12, 2026. He argues that AI work is shifting from people working back and forth with AI to people managing AI agents: systems given work that can return useful results in minutes. He points to rapid gains on tests, while stressing that the tests have flaws, AI remains “jagged,” and most organizations have changed little so far.\n\nMollick describes StrongDM’s Software Factory as a radical experiment: AI agents write and test production software without human involvement, while people review the finished product. He argues that the experiment matters less for its particular design than for what it signals: organizations may need to experiment with new ways of working as AI improves.\n\nHe connects advances in capability with market reactions, job impacts and conflicts between AI companies and governments, while noting that recent examples were not always what they first appeared to be. Mollick’s central concern is recursive self-improvement: AI systems helping build better AI systems. He says the loop may encounter bottlenecks and its endpoint is uncertain, but argues that companies are actively pursuing it. His conclusion is that uncertainty is not helplessness: choices made by individuals and organizations can shape how AI is used at work, in schools and in government, and those choices may set precedents for others.","topicIds":["work-and-business--jobs-and-occupations","job-displacement-and-creation","artificial-general-intelligence"],"domain":"other","claimType":"attributed reading","limits":"Text as retrieved on October 24, 2026; originally published March 12, 2026. Mollick notes that the AI tests discussed have flaws and that AI remains capable of some tasks while failing at others. The essay says the potential limits and endpoint of recursive self-improvement are unknown.","speaker":{"name":"One Useful Thing","roleAtSource":"Publisher"},"source":{"title":"The Shape of the Thing","publisher":"One Useful Thing","format":"Article","url":"https://www.oneusefulthing.org/p/the-shape-of-the-thing","publishedAt":"2026-03-12T14:10:07.000Z","dateBasis":"Date supplied by the publisher’s feed."},"evidence":[{"text":"Ethan Mollick argues that AI work is shifting from people working with AI to people managing AI agents.","excerpt":"This is an era of managing AIs, rather than working with them.","url":"https://www.oneusefulthing.org/p/the-shape-of-the-thing","relationship":"Source-stated"},{"text":"Mollick says the tests show rapid gains, but have flaws and do not mean most organizations have already changed substantially.","excerpt":"all of these tests have their own flaws, and that AI remains jagged","url":"https://www.oneusefulthing.org/p/the-shape-of-the-thing","relationship":"Source-stated"},{"text":"Mollick describes StrongDM’s Software Factory as having AI write and test production software without human involvement, with people reviewing the finished product.","excerpt":"AI to write, test, and ship production software without human involvement.","url":"https://www.oneusefulthing.org/p/the-shape-of-the-thing","relationship":"Source-stated"},{"text":"Mollick argues that recursive self-improvement could accelerate AI progress, while noting possible bottlenecks and an uncertain endpoint.","excerpt":"If the loop does close, the exponential curves we’ve been watching would get steeper, with an uncertain endpoint.","url":"https://www.oneusefulthing.org/p/the-shape-of-the-thing","relationship":"Source-stated"},{"text":"Mollick concludes that people and organizations can still shape how AI is used, and that their choices can set precedents for others.","excerpt":"every organization figuring out a good way to use AI right now is setting a precedent for everyone else.","url":"https://www.oneusefulthing.org/p/the-shape-of-the-thing","relationship":"Source-stated"},{"text":"The essay is dated March 12, 2026.","excerpt":"Mar 12, 2026","url":"https://www.oneusefulthing.org/p/the-shape-of-the-thing","relationship":"Source-stated"}],"sourceMedia":{"status":"unchecked","sourceUrl":"https://www.oneusefulthing.org/p/the-shape-of-the-thing","checkedAt":null,"items":[]},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"What bottlenecks, if any, will limit AI systems’ ability to help build better AI systems?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"Ethan Mollick argues in One Useful Thing that AI’s shape is becoming clearer","summary":"Ethan Mollick published this essay on March 12, 2026. He argues that AI work is shifting from people working back and forth with AI to people managing AI agents: systems given work that can return useful results in minutes. He points to rapid gains on tests, while stressing that the tests have flaws, AI remains “jagged,” and most organizations have changed little so far.\n\nMollick describes StrongDM’s Software Factory as a radical experiment: AI agents write and test production software without human involvement, while people review the finished product. He argues that the experiment matters less for its particular design than for what it signals: organizations may need to experiment with new ways of working as AI improves.\n\nHe connects advances in capability with market reactions, job impacts and conflicts between AI companies and governments, while noting that recent examples were not always what they first appeared to be. Mollick’s central concern is recursive self-improvement: AI systems helping build better AI systems. He says the loop may encounter bottlenecks and its endpoint is uncertain, but argues that companies are actively pursuing it. His conclusion is that uncertainty is not helplessness: choices made by individuals and organizations can shape how AI is used at work, in schools and in government, and those choices may set precedents for others.","question":"What bottlenecks, if any, will limit AI systems’ ability to help build better AI systems?","claims":[{"text":"Ethan Mollick argues that AI work is shifting from people working with AI to people managing AI agents.","excerpt":"This is an era of managing AIs, rather than working with them.","support":"source-stated"},{"text":"Mollick says the tests show rapid gains, but have flaws and do not mean most organizations have already changed substantially.","excerpt":"all of these tests have their own flaws, and that AI remains jagged","support":"source-stated"},{"text":"Mollick describes StrongDM’s Software Factory as having AI write and test production software without human involvement, with people reviewing the finished product.","excerpt":"AI to write, test, and ship production software without human involvement.","support":"source-stated"},{"text":"Mollick argues that recursive self-improvement could accelerate AI progress, while noting possible bottlenecks and an uncertain endpoint.","excerpt":"If the loop does close, the exponential curves we’ve been watching would get steeper, with an uncertain endpoint.","support":"source-stated"},{"text":"Mollick concludes that people and organizations can still shape how AI is used, and that their choices can set precedents for others.","excerpt":"every organization figuring out a good way to use AI right now is setting a precedent for everyone else.","support":"source-stated"},{"text":"The essay is dated March 12, 2026.","excerpt":"Mar 12, 2026","support":"source-stated"}],"limitations":["Text as retrieved on October 24, 2026; originally published March 12, 2026.","Mollick notes that the AI tests discussed have flaws and that AI remains capable of some tasks while failing at others.","The essay says the potential limits and endpoint of recursive self-improvement are unknown."],"mediaId":null,"correctionReason":null,"id":"e095c2ec54370fb2a9a3bc5d97c472cf08f758c74076450ce0728af50508d6f1","publicationId":"26ad7f1b-6e81-ddbb-7944-04db73586e41","version":1,"sourceUrl":"https://www.oneusefulthing.org/p/the-shape-of-the-thing","sourceTitle":"The Shape of the Thing","publisher":"One Useful Thing","publishedAt":"2026-03-12T14:10:07.000Z","originalPublished":{"value":"2026-03-12T14:10:07.000Z","precision":"instant","basis":"backfill:feed_published"},"firstObservedAt":"2026-09-23T05:37:00.948Z","createdAt":"2026-11-07T22:58:23.751Z","historical":true,"entityIds":["one-useful-thing"],"topicIds":["work-and-business--jobs-and-occupations","job-displacement-and-creation","artificial-general-intelligence"],"media":[],"sourceMedia":{"status":"unchecked","sourceUrl":"https://www.oneusefulthing.org/p/the-shape-of-the-thing","checkedAt":null,"items":[]},"editorialDecision":{"outcome":"accept","mediaId":null,"inputHash":"083b98153a14c945bc6648bcb5557fe28362c70659ce7652a56ed338b206c31f","rubricVersion":"unrecorded","provider":"unrecorded","model":"anthropic/claude-opus-5.5","reasonCodes":["bridge_admitted_by_model","second_opinion_categorical"],"confidence":1,"decidedAt":"2026-10-24T07:09:53.549Z"},"provenance":{"textHash":"b9ce9397b6e22dc01195c9f23210a7fb83cc878052819982f38c8ad3da3d51a9","observationId":"8083db8e432c5ec7fcb9a7e27ab5f76684e0e922616f4b9a67f0ad3353863801","retrievedAt":"2026-10-24T07:11:36.233Z","kind":"publisher-page","preparer":{"model":"openai/gpt-6-luna","runId":"c5a0b5ee-d452-e98e-2e84-44fd5dc2d1b5"},"reviewer":{"model":"anthropic/claude-opus-5.5","runId":"770dd706-3116-50f3-e5af-295af087ed6a"}},"audit":{"status":"passed","reviewedAt":"2026-11-07T22:53:39.169Z","model":"anthropic/claude-opus-5.5"}},"canonicalUrl":"https://escapemeridian.com/readings/ethan-mollick-argues-in-one-useful-thing-that-ai-s-shape-is-e095c2ec","formats":{"html":"/readings/ethan-mollick-argues-in-one-useful-thing-that-ai-s-shape-is-e095c2ec","markdown":"/readings/ethan-mollick-argues-in-one-useful-thing-that-ai-s-shape-is-e095c2ec.md","json":"/readings/ethan-mollick-argues-in-one-useful-thing-that-ai-s-shape-is-e095c2ec.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":[{"id":"term-ai-agent","name":"AI agent","explanation":"An AI system that chooses and carries out actions toward a goal, often using software tools.","referenceUrl":"https://www.ibm.com/think/topics/ai-agents"},{"id":"term-recursive-self-improvement","name":"Recursive self-improvement","explanation":"An AI system improving its own capabilities, such as its code, training methods or reasoning, then using the gains to improve further. Each cycle could hypothetically produce a more capable system.","referenceUrl":"https://hai.stanford.edu/what-is-recursive-self-improvement"}],"entities":[{"id":"ethan-mollick","name":"Ethan Mollick","kind":"person","context":"Ethan Mollick is a Wharton researcher and author exploring how AI changes work, learning and decision-making.","referenceUrls":["https://mgmt.wharton.upenn.edu/profile/emollick/","https://www.oneusefulthing.org/"],"path":"/entities/ethan-mollick","json":"/entities/ethan-mollick.json","markdown":"/entities/ethan-mollick.md"},{"id":"one-useful-thing","name":"One Useful Thing","kind":"organization","context":"One Useful Thing is Ethan Mollick’s publication about AI in work, education and everyday practice.","referenceUrls":["https://www.oneusefulthing.org/"],"path":"/entities/one-useful-thing","json":"/entities/one-useful-thing.json","markdown":"/entities/one-useful-thing.md"}]},"researchQuestions":[],"withdrawn":false,"references":{"schemaVersion":1,"target":{"type":"reading","id":"e095c2ec54370fb2a9a3bc5d97c472cf08f758c74076450ce0728af50508d6f1","revision":"26ad7f1b-6e81-ddbb-7944-04db73586e41","revisionKind":"publication-id","title":"Ethan Mollick argues in One Useful Thing that AI’s shape is becoming clearer","url":"https://escapemeridian.com/readings/ethan-mollick-argues-in-one-useful-thing-that-ai-s-shape-is-e095c2ec"},"source":{"type":"source","id":"https://www.oneusefulthing.org/p/the-shape-of-the-thing","revision":"b9ce9397b6e22dc01195c9f23210a7fb83cc878052819982f38c8ad3da3d51a9","revisionKind":"sha256-source-text","title":"The Shape of the Thing","url":"https://www.oneusefulthing.org/p/the-shape-of-the-thing"},"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":"e095c2ec54370fb2a9a3bc5d97c472cf08f758c74076450ce0728af50508d6f1","title":"Ethan Mollick argues in One Useful Thing that AI’s shape is becoming clearer","argument":"Ethan Mollick published this essay on March 12, 2026. He argues that AI work is shifting from people working back and forth with AI to people managing AI agents: systems given work that can return useful results in minutes. He points to rapid gains on tests, while stressing that the tests have flaws, AI remains “jagged,” and most organizations have changed little so far.\n\nMollick describes StrongDM’s Software Factory as a radical experiment: AI agents write and test production software without human involvement, while people review the finished product. He argues that the experiment matters less for its particular design than for what it signals: organizations may need to experiment with new ways of working as AI improves.\n\nHe connects advances in capability with market reactions, job impacts and conflicts between AI companies and governments, while noting that recent examples were not always what they first appeared to be. Mollick’s central concern is recursive self-improvement: AI systems helping build better AI systems. He says the loop may encounter bottlenecks and its endpoint is uncertain, but argues that companies are actively pursuing it. His conclusion is that uncertainty is not helplessness: choices made by individuals and organizations can shape how AI is used at work, in schools and in government, and those choices may set precedents for others.","topicIds":["work-and-business--jobs-and-occupations","job-displacement-and-creation","artificial-general-intelligence"],"domain":"other","claimType":"attributed reading","limits":"Text as retrieved on October 24, 2026; originally published March 12, 2026. Mollick notes that the AI tests discussed have flaws and that AI remains capable of some tasks while failing at others. The essay says the potential limits and endpoint of recursive self-improvement are unknown.","speaker":{"name":"One Useful Thing","roleAtSource":"Publisher"},"source":{"title":"The Shape of the Thing","publisher":"One Useful Thing","format":"Article","url":"https://www.oneusefulthing.org/p/the-shape-of-the-thing","publishedAt":"2026-03-12T14:10:07.000Z","dateBasis":"Date supplied by the publisher’s feed."},"evidence":[{"text":"Ethan Mollick argues that AI work is shifting from people working with AI to people managing AI agents.","excerpt":"This is an era of managing AIs, rather than working with them.","url":"https://www.oneusefulthing.org/p/the-shape-of-the-thing","relationship":"Source-stated"},{"text":"Mollick says the tests show rapid gains, but have flaws and do not mean most organizations have already changed substantially.","excerpt":"all of these tests have their own flaws, and that AI remains jagged","url":"https://www.oneusefulthing.org/p/the-shape-of-the-thing","relationship":"Source-stated"},{"text":"Mollick describes StrongDM’s Software Factory as having AI write and test production software without human involvement, with people reviewing the finished product.","excerpt":"AI to write, test, and ship production software without human involvement.","url":"https://www.oneusefulthing.org/p/the-shape-of-the-thing","relationship":"Source-stated"},{"text":"Mollick argues that recursive self-improvement could accelerate AI progress, while noting possible bottlenecks and an uncertain endpoint.","excerpt":"If the loop does close, the exponential curves we’ve been watching would get steeper, with an uncertain endpoint.","url":"https://www.oneusefulthing.org/p/the-shape-of-the-thing","relationship":"Source-stated"},{"text":"Mollick concludes that people and organizations can still shape how AI is used, and that their choices can set precedents for others.","excerpt":"every organization figuring out a good way to use AI right now is setting a precedent for everyone else.","url":"https://www.oneusefulthing.org/p/the-shape-of-the-thing","relationship":"Source-stated"},{"text":"The essay is dated March 12, 2026.","excerpt":"Mar 12, 2026","url":"https://www.oneusefulthing.org/p/the-shape-of-the-thing","relationship":"Source-stated"}],"sourceMedia":{"status":"unchecked","sourceUrl":"https://www.oneusefulthing.org/p/the-shape-of-the-thing","checkedAt":null,"items":[]},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"What bottlenecks, if any, will limit AI systems’ ability to help build better AI systems?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"Ethan Mollick argues in One Useful Thing that AI’s shape is becoming clearer","summary":"Ethan Mollick published this essay on March 12, 2026. He argues that AI work is shifting from people working back and forth with AI to people managing AI agents: systems given work that can return useful results in minutes. He points to rapid gains on tests, while stressing that the tests have flaws, AI remains “jagged,” and most organizations have changed little so far.\n\nMollick describes StrongDM’s Software Factory as a radical experiment: AI agents write and test production software without human involvement, while people review the finished product. He argues that the experiment matters less for its particular design than for what it signals: organizations may need to experiment with new ways of working as AI improves.\n\nHe connects advances in capability with market reactions, job impacts and conflicts between AI companies and governments, while noting that recent examples were not always what they first appeared to be. Mollick’s central concern is recursive self-improvement: AI systems helping build better AI systems. He says the loop may encounter bottlenecks and its endpoint is uncertain, but argues that companies are actively pursuing it. His conclusion is that uncertainty is not helplessness: choices made by individuals and organizations can shape how AI is used at work, in schools and in government, and those choices may set precedents for others.","question":"What bottlenecks, if any, will limit AI systems’ ability to help build better AI systems?","claims":[{"text":"Ethan Mollick argues that AI work is shifting from people working with AI to people managing AI agents.","excerpt":"This is an era of managing AIs, rather than working with them.","support":"source-stated"},{"text":"Mollick says the tests show rapid gains, but have flaws and do not mean most organizations have already changed substantially.","excerpt":"all of these tests have their own flaws, and that AI remains jagged","support":"source-stated"},{"text":"Mollick describes StrongDM’s Software Factory as having AI write and test production software without human involvement, with people reviewing the finished product.","excerpt":"AI to write, test, and ship production software without human involvement.","support":"source-stated"},{"text":"Mollick argues that recursive self-improvement could accelerate AI progress, while noting possible bottlenecks and an uncertain endpoint.","excerpt":"If the loop does close, the exponential curves we’ve been watching would get steeper, with an uncertain endpoint.","support":"source-stated"},{"text":"Mollick concludes that people and organizations can still shape how AI is used, and that their choices can set precedents for others.","excerpt":"every organization figuring out a good way to use AI right now is setting a precedent for everyone else.","support":"source-stated"},{"text":"The essay is dated March 12, 2026.","excerpt":"Mar 12, 2026","support":"source-stated"}],"limitations":["Text as retrieved on October 24, 2026; originally published March 12, 2026.","Mollick notes that the AI tests discussed have flaws and that AI remains capable of some tasks while failing at others.","The essay says the potential limits and endpoint of recursive self-improvement are unknown."],"mediaId":null,"correctionReason":null,"id":"e095c2ec54370fb2a9a3bc5d97c472cf08f758c74076450ce0728af50508d6f1","publicationId":"26ad7f1b-6e81-ddbb-7944-04db73586e41","version":1,"sourceUrl":"https://www.oneusefulthing.org/p/the-shape-of-the-thing","sourceTitle":"The Shape of the Thing","publisher":"One Useful Thing","publishedAt":"2026-03-12T14:10:07.000Z","originalPublished":{"value":"2026-03-12T14:10:07.000Z","precision":"instant","basis":"backfill:feed_published"},"firstObservedAt":"2026-09-23T05:37:00.948Z","createdAt":"2026-11-07T22:58:23.751Z","historical":true,"entityIds":["one-useful-thing"],"topicIds":["work-and-business--jobs-and-occupations","job-displacement-and-creation","artificial-general-intelligence"],"media":[],"sourceMedia":{"status":"unchecked","sourceUrl":"https://www.oneusefulthing.org/p/the-shape-of-the-thing","checkedAt":null,"items":[]},"editorialDecision":{"outcome":"accept","mediaId":null,"inputHash":"083b98153a14c945bc6648bcb5557fe28362c70659ce7652a56ed338b206c31f","rubricVersion":"unrecorded","provider":"unrecorded","model":"anthropic/claude-opus-5.5","reasonCodes":["bridge_admitted_by_model","second_opinion_categorical"],"confidence":1,"decidedAt":"2026-10-24T07:09:53.549Z"},"provenance":{"textHash":"b9ce9397b6e22dc01195c9f23210a7fb83cc878052819982f38c8ad3da3d51a9","observationId":"8083db8e432c5ec7fcb9a7e27ab5f76684e0e922616f4b9a67f0ad3353863801","retrievedAt":"2026-10-24T07:11:36.233Z","kind":"publisher-page","preparer":{"model":"openai/gpt-6-luna","runId":"c5a0b5ee-d452-e98e-2e84-44fd5dc2d1b5"},"reviewer":{"model":"anthropic/claude-opus-5.5","runId":"770dd706-3116-50f3-e5af-295af087ed6a"}},"audit":{"status":"passed","reviewedAt":"2026-11-07T22:53:39.169Z","model":"anthropic/claude-opus-5.5"}}}],"audits":[{"publicationId":"26ad7f1b-6e81-ddbb-7944-04db73586e41","verdict":"passed","findings":[{"dimension":"attribution","status":"pass","explanation":"The reading attributes the essay to Ethan Mollick in One Useful Thing, matching the byline and publisher. The arguments stay attributed to him, and statements by StrongDM, Amodei and OpenAI are not misattributed.","excerpt":"The Shape of the Thing\nWhere we are right now, and what likely happens next\nEthan Mollick"},{"dimension":"summary","status":"pass","explanation":"The summary follows the essay's arc: the shift from co-intelligence to managing agents, test gains with caveats, StrongDM's Software Factory, the rolling-disruption week, recursive self-improvement, and the closing call about agency and precedent. Agents and RSI are defined beside the reading. 'Jagged' is glossed in the limitations. Calling RSI his 'central concern' is a mild framing; the source gives it as his main reason not to bet on things settling. The question follows from his stated uncertainty.","excerpt":"One reason is that AI companies are telling us, fairly explicitly, what comes next: recursive self-improvement, or RSI."},{"dimension":"date","status":"pass","explanation":"The originalPublished instant, 2026-03-12T14:10Z, matches the dateline Mar 12, 2026, and the reading states it at day precision. The source is a publisher page, so no archive disclosure is needed. The modification date equals the publication date. There is no hindsight framing, and forecasts such as RSI and future instability remain attributed to Mollick.","excerpt":"Mar 12, 2026"},{"dimension":"support","status":"pass","explanation":"Each claim matches the source text. Mollick says tests are flawed, AI is jagged, and adoption is early, with little change in most organizations. He describes the Software Factory as autonomous AI coding with humans reviewing the finished product. He presents RSI as a possible accelerator, with bottlenecks and an uncertain endpoint. He frames organizations' choices as setting precedents. No benefits or costs are invented.","excerpt":"companies are still very early in adopting AI, meaning that, as of yet, remarkably little has changed in most organizations"},{"dimension":"media","status":"pass","explanation":"No media candidates were supplied and none was selected, which is consistent with the evidence.","excerpt":null}],"createdAt":"2026-11-07T22:53:39.169Z","model":"anthropic/claude-opus-5.5"}],"correctionUrl":"https://escapemeridian.com/corrections/"}