{"id":"8cbb5a7ceb81eaf9daa64e04216c1eb3ecc3c77ac6353ad8d499bdb46f1f7a96","title":"Toby Ord argues in “Swarm Scaling” that AI swarms trade efficiency for speed","argument":"In an essay published September 21, 2026, Toby Ord argues that adding AI agents to a swarm can increase capability, but is less compute-efficient than giving one agent more time to reason. Ord estimates that, across three benchmarks, scaling the swarm by a factor of ten yields the performance gain of scaling one agent’s tokens by three to five. He describes this as diminishing returns: the more agents added, the further swarms fall behind single-agent duration scaling in performance for a given compute increase.\n\nOrd’s case for swarms is speed. Agents can run in parallel, so a swarm may finish sooner at greater total cost; he says the measured results make speed the main demonstrated reason to use them. He also raises a possible higher performance plateau for very large swarms, but says this remains unshown. His broader concern is that the measured values of the “stepping on toes” parameter, λ, are close to estimates used in models of recursive self-improvement. In those models, Ord says, higher λ makes an intelligence explosion more likely. He had hoped the values would be lower, and concludes that people should keep tracking them, especially as new orchestration methods emerge. The estimates come from limited benchmark data; whether they hold at much larger swarm sizes remains unknown.","topicIds":["artificial-general-intelligence"],"domain":"other","claimType":"attributed reading","limits":"Text as retrieved on October 27, 2026; originally published September 21, 2026. Ord says there is little data on swarm scaling, especially for frontier systems; the estimates discussed are based on benchmark charts and do not establish how scaling behaves at much larger swarm sizes. Ord’s possible higher performance plateau for a very large swarm is presented as a possibility, not a demonstrated result.","speaker":{"name":"Toby Ord","roleAtSource":"Publisher"},"source":{"title":"Swarm Scaling","publisher":"Toby Ord","format":"Article","url":"https://www.tobyord.com/writing/swarm-scaling","publishedAt":"2026-09-21T18:53:19+01:00","dateBasis":"Date stated on the publisher’s page."},"evidence":[{"text":"Ord argues that increasing swarm size gives less capability gain per unit of compute than increasing one agent’s reasoning time.","excerpt":"The swarm scaling gives a little over half as much gain in capability for the same scale-up of compute.","url":"https://www.tobyord.com/writing/swarm-scaling","relationship":"Source-stated"},{"text":"Ord estimates that the measured λ values imply a tenfold increase in swarm size yields the performance gain of a threefold to fivefold increase in tokens for one agent.","excerpt":"scaling up the number of agents in the swarm by 10x doesn’t get as much performance as using 10x as many tokens with one agent. Instead it gets $10^\\lambda$x as much — which is 3x to 5x.","url":"https://www.tobyord.com/writing/swarm-scaling","relationship":"Source-stated"},{"text":"Ord says speed is the main demonstrated reason to use swarms, though he also raises a possible higher performance plateau for a very large swarm.","excerpt":"for now, the main demonstrated reason is speed.","url":"https://www.tobyord.com/writing/swarm-scaling","relationship":"Source-stated"},{"text":"Ord argues that the measured λ values are close to estimates used in models of recursive self-improvement, where higher λ makes an intelligence explosion more likely.","excerpt":"high $\\lambda$ makes intelligence explosions more likely.","url":"https://www.tobyord.com/writing/swarm-scaling","relationship":"Source-stated"},{"text":"Ord says the estimates may change with task and swarm size, and calls for continued tracking as new estimates and orchestration methods appear.","excerpt":"People should keep tracking this as new estimates for $\\lambda$ appear and (especially) when new orchestration methods increase the value of $\\lambda$ for a given type of task.","url":"https://www.tobyord.com/writing/swarm-scaling","relationship":"Source-stated"}],"sourceMedia":{"status":"unchecked","sourceUrl":"https://www.tobyord.com/writing/swarm-scaling","checkedAt":null,"items":[]},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How does λ change when swarms grow far beyond the sizes tested, and does that change the implications Ord draws for recursive self-improvement?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"Toby Ord argues in “Swarm Scaling” that AI swarms trade efficiency for speed","summary":"In an essay published September 21, 2026, Toby Ord argues that adding AI agents to a swarm can increase capability, but is less compute-efficient than giving one agent more time to reason. Ord estimates that, across three benchmarks, scaling the swarm by a factor of ten yields the performance gain of scaling one agent’s tokens by three to five. He describes this as diminishing returns: the more agents added, the further swarms fall behind single-agent duration scaling in performance for a given compute increase.\n\nOrd’s case for swarms is speed. Agents can run in parallel, so a swarm may finish sooner at greater total cost; he says the measured results make speed the main demonstrated reason to use them. He also raises a possible higher performance plateau for very large swarms, but says this remains unshown. His broader concern is that the measured values of the “stepping on toes” parameter, λ, are close to estimates used in models of recursive self-improvement. In those models, Ord says, higher λ makes an intelligence explosion more likely. He had hoped the values would be lower, and concludes that people should keep tracking them, especially as new orchestration methods emerge. The estimates come from limited benchmark data; whether they hold at much larger swarm sizes remains unknown.","question":"How does λ change when swarms grow far beyond the sizes tested, and does that change the implications Ord draws for recursive self-improvement?","claims":[{"text":"Ord argues that increasing swarm size gives less capability gain per unit of compute than increasing one agent’s reasoning time.","excerpt":"The swarm scaling gives a little over half as much gain in capability for the same scale-up of compute.","support":"source-stated"},{"text":"Ord estimates that the measured λ values imply a tenfold increase in swarm size yields the performance gain of a threefold to fivefold increase in tokens for one agent.","excerpt":"scaling up the number of agents in the swarm by 10x doesn’t get as much performance as using 10x as many tokens with one agent. Instead it gets $10^\\lambda$x as much — which is 3x to 5x.","support":"source-stated"},{"text":"Ord says speed is the main demonstrated reason to use swarms, though he also raises a possible higher performance plateau for a very large swarm.","excerpt":"for now, the main demonstrated reason is speed.","support":"source-stated"},{"text":"Ord argues that the measured λ values are close to estimates used in models of recursive self-improvement, where higher λ makes an intelligence explosion more likely.","excerpt":"high $\\lambda$ makes intelligence explosions more likely.","support":"source-stated"},{"text":"Ord says the estimates may change with task and swarm size, and calls for continued tracking as new estimates and orchestration methods appear.","excerpt":"People should keep tracking this as new estimates for $\\lambda$ appear and (especially) when new orchestration methods increase the value of $\\lambda$ for a given type of task.","support":"source-stated"}],"limitations":["Text as retrieved on October 27, 2026; originally published September 21, 2026.","Ord says there is little data on swarm scaling, especially for frontier systems; the estimates discussed are based on benchmark charts and do not establish how scaling behaves at much larger swarm sizes.","Ord’s possible higher performance plateau for a very large swarm is presented as a possibility, not a demonstrated result."],"mediaId":null,"correctionReason":null,"id":"8cbb5a7ceb81eaf9daa64e04216c1eb3ecc3c77ac6353ad8d499bdb46f1f7a96","publicationId":"803ddfa4-72d4-3939-413f-3e3c316c9869","version":1,"sourceUrl":"https://www.tobyord.com/writing/swarm-scaling","sourceTitle":"Swarm Scaling","publisher":"Toby Ord","publishedAt":"2026-09-21T18:53:19+01:00","originalPublished":{"value":"2026-09-21T18:53:19+01:00","precision":"instant","basis":"backfill:article_published"},"firstObservedAt":"2026-10-27T21:55:28.830Z","createdAt":"2026-11-05T17:28:23.390Z","historical":true,"entityIds":[],"topicIds":["artificial-general-intelligence"],"media":[],"sourceMedia":{"status":"unchecked","sourceUrl":"https://www.tobyord.com/writing/swarm-scaling","checkedAt":null,"items":[]},"editorialDecision":{"outcome":"accept","mediaId":null,"inputHash":"9595e564898a21a6bf5c1dc33a0ad30401e1df5d5d8b50363022ff456f2d8ec7","rubricVersion":"unrecorded","provider":"anthropic","model":"claude-opus-5-5","reasonCodes":["bridge_admitted_by_model","second_opinion_categorical"],"confidence":1,"decidedAt":"2026-10-27T21:55:26.139Z"},"provenance":{"textHash":"d06fb54ed4ef5618e9fa3da24df2e28abd4e2fe2665a9d5d40c004ac890f1c25","observationId":"fa7ebb77a1ef4ef5384e53eccaaf7e0b4314e5e5558b1a1a3b74d4d8d401d9d9","retrievedAt":"2026-10-27T21:57:24.407Z","kind":"publisher-page","preparer":{"model":"openai/gpt-6-luna","runId":"36196654-bfea-a79f-b03f-3898d152a7e6"},"reviewer":{"model":"anthropic/claude-opus-5.5","runId":"bd330171-7ef6-d5e4-7eaa-dbfb41dcbc21"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-05T17:23:50.790Z","model":"anthropic/claude-opus-5.5"}},"canonicalUrl":"https://escapemeridian.com/readings/toby-ord-argues-in-swarm-scaling-that-ai-swarms-trade-8cbb5a7c","formats":{"html":"/readings/toby-ord-argues-in-swarm-scaling-that-ai-swarms-trade-8cbb5a7c","markdown":"/readings/toby-ord-argues-in-swarm-scaling-that-ai-swarms-trade-8cbb5a7c.md","json":"/readings/toby-ord-argues-in-swarm-scaling-that-ai-swarms-trade-8cbb5a7c.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-benchmark","name":"Benchmark","explanation":"A defined set of tests used to compare systems. Results depend on the tasks and measures selected.","referenceUrl":"https://arxiv.org/abs/2211.09110"},{"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":"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"}]},"researchQuestions":[],"withdrawn":false,"references":{"schemaVersion":1,"target":{"type":"reading","id":"8cbb5a7ceb81eaf9daa64e04216c1eb3ecc3c77ac6353ad8d499bdb46f1f7a96","revision":"803ddfa4-72d4-3939-413f-3e3c316c9869","revisionKind":"publication-id","title":"Toby Ord argues in “Swarm Scaling” that AI swarms trade efficiency for speed","url":"https://escapemeridian.com/readings/toby-ord-argues-in-swarm-scaling-that-ai-swarms-trade-8cbb5a7c"},"source":{"type":"source","id":"https://www.tobyord.com/writing/swarm-scaling","revision":"d06fb54ed4ef5618e9fa3da24df2e28abd4e2fe2665a9d5d40c004ac890f1c25","revisionKind":"sha256-source-text","title":"Swarm Scaling","url":"https://www.tobyord.com/writing/swarm-scaling"},"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":"8cbb5a7ceb81eaf9daa64e04216c1eb3ecc3c77ac6353ad8d499bdb46f1f7a96","title":"Toby Ord argues in “Swarm Scaling” that AI swarms trade efficiency for speed","argument":"In an essay published September 21, 2026, Toby Ord argues that adding AI agents to a swarm can increase capability, but is less compute-efficient than giving one agent more time to reason. Ord estimates that, across three benchmarks, scaling the swarm by a factor of ten yields the performance gain of scaling one agent’s tokens by three to five. He describes this as diminishing returns: the more agents added, the further swarms fall behind single-agent duration scaling in performance for a given compute increase.\n\nOrd’s case for swarms is speed. Agents can run in parallel, so a swarm may finish sooner at greater total cost; he says the measured results make speed the main demonstrated reason to use them. He also raises a possible higher performance plateau for very large swarms, but says this remains unshown. His broader concern is that the measured values of the “stepping on toes” parameter, λ, are close to estimates used in models of recursive self-improvement. In those models, Ord says, higher λ makes an intelligence explosion more likely. He had hoped the values would be lower, and concludes that people should keep tracking them, especially as new orchestration methods emerge. The estimates come from limited benchmark data; whether they hold at much larger swarm sizes remains unknown.","topicIds":["artificial-general-intelligence"],"domain":"other","claimType":"attributed reading","limits":"Text as retrieved on October 27, 2026; originally published September 21, 2026. Ord says there is little data on swarm scaling, especially for frontier systems; the estimates discussed are based on benchmark charts and do not establish how scaling behaves at much larger swarm sizes. Ord’s possible higher performance plateau for a very large swarm is presented as a possibility, not a demonstrated result.","speaker":{"name":"Toby Ord","roleAtSource":"Publisher"},"source":{"title":"Swarm Scaling","publisher":"Toby Ord","format":"Article","url":"https://www.tobyord.com/writing/swarm-scaling","publishedAt":"2026-09-21T18:53:19+01:00","dateBasis":"Date stated on the publisher’s page."},"evidence":[{"text":"Ord argues that increasing swarm size gives less capability gain per unit of compute than increasing one agent’s reasoning time.","excerpt":"The swarm scaling gives a little over half as much gain in capability for the same scale-up of compute.","url":"https://www.tobyord.com/writing/swarm-scaling","relationship":"Source-stated"},{"text":"Ord estimates that the measured λ values imply a tenfold increase in swarm size yields the performance gain of a threefold to fivefold increase in tokens for one agent.","excerpt":"scaling up the number of agents in the swarm by 10x doesn’t get as much performance as using 10x as many tokens with one agent. Instead it gets $10^\\lambda$x as much — which is 3x to 5x.","url":"https://www.tobyord.com/writing/swarm-scaling","relationship":"Source-stated"},{"text":"Ord says speed is the main demonstrated reason to use swarms, though he also raises a possible higher performance plateau for a very large swarm.","excerpt":"for now, the main demonstrated reason is speed.","url":"https://www.tobyord.com/writing/swarm-scaling","relationship":"Source-stated"},{"text":"Ord argues that the measured λ values are close to estimates used in models of recursive self-improvement, where higher λ makes an intelligence explosion more likely.","excerpt":"high $\\lambda$ makes intelligence explosions more likely.","url":"https://www.tobyord.com/writing/swarm-scaling","relationship":"Source-stated"},{"text":"Ord says the estimates may change with task and swarm size, and calls for continued tracking as new estimates and orchestration methods appear.","excerpt":"People should keep tracking this as new estimates for $\\lambda$ appear and (especially) when new orchestration methods increase the value of $\\lambda$ for a given type of task.","url":"https://www.tobyord.com/writing/swarm-scaling","relationship":"Source-stated"}],"sourceMedia":{"status":"unchecked","sourceUrl":"https://www.tobyord.com/writing/swarm-scaling","checkedAt":null,"items":[]},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How does λ change when swarms grow far beyond the sizes tested, and does that change the implications Ord draws for recursive self-improvement?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"Toby Ord argues in “Swarm Scaling” that AI swarms trade efficiency for speed","summary":"In an essay published September 21, 2026, Toby Ord argues that adding AI agents to a swarm can increase capability, but is less compute-efficient than giving one agent more time to reason. Ord estimates that, across three benchmarks, scaling the swarm by a factor of ten yields the performance gain of scaling one agent’s tokens by three to five. He describes this as diminishing returns: the more agents added, the further swarms fall behind single-agent duration scaling in performance for a given compute increase.\n\nOrd’s case for swarms is speed. Agents can run in parallel, so a swarm may finish sooner at greater total cost; he says the measured results make speed the main demonstrated reason to use them. He also raises a possible higher performance plateau for very large swarms, but says this remains unshown. His broader concern is that the measured values of the “stepping on toes” parameter, λ, are close to estimates used in models of recursive self-improvement. In those models, Ord says, higher λ makes an intelligence explosion more likely. He had hoped the values would be lower, and concludes that people should keep tracking them, especially as new orchestration methods emerge. The estimates come from limited benchmark data; whether they hold at much larger swarm sizes remains unknown.","question":"How does λ change when swarms grow far beyond the sizes tested, and does that change the implications Ord draws for recursive self-improvement?","claims":[{"text":"Ord argues that increasing swarm size gives less capability gain per unit of compute than increasing one agent’s reasoning time.","excerpt":"The swarm scaling gives a little over half as much gain in capability for the same scale-up of compute.","support":"source-stated"},{"text":"Ord estimates that the measured λ values imply a tenfold increase in swarm size yields the performance gain of a threefold to fivefold increase in tokens for one agent.","excerpt":"scaling up the number of agents in the swarm by 10x doesn’t get as much performance as using 10x as many tokens with one agent. Instead it gets $10^\\lambda$x as much — which is 3x to 5x.","support":"source-stated"},{"text":"Ord says speed is the main demonstrated reason to use swarms, though he also raises a possible higher performance plateau for a very large swarm.","excerpt":"for now, the main demonstrated reason is speed.","support":"source-stated"},{"text":"Ord argues that the measured λ values are close to estimates used in models of recursive self-improvement, where higher λ makes an intelligence explosion more likely.","excerpt":"high $\\lambda$ makes intelligence explosions more likely.","support":"source-stated"},{"text":"Ord says the estimates may change with task and swarm size, and calls for continued tracking as new estimates and orchestration methods appear.","excerpt":"People should keep tracking this as new estimates for $\\lambda$ appear and (especially) when new orchestration methods increase the value of $\\lambda$ for a given type of task.","support":"source-stated"}],"limitations":["Text as retrieved on October 27, 2026; originally published September 21, 2026.","Ord says there is little data on swarm scaling, especially for frontier systems; the estimates discussed are based on benchmark charts and do not establish how scaling behaves at much larger swarm sizes.","Ord’s possible higher performance plateau for a very large swarm is presented as a possibility, not a demonstrated result."],"mediaId":null,"correctionReason":null,"id":"8cbb5a7ceb81eaf9daa64e04216c1eb3ecc3c77ac6353ad8d499bdb46f1f7a96","publicationId":"803ddfa4-72d4-3939-413f-3e3c316c9869","version":1,"sourceUrl":"https://www.tobyord.com/writing/swarm-scaling","sourceTitle":"Swarm Scaling","publisher":"Toby Ord","publishedAt":"2026-09-21T18:53:19+01:00","originalPublished":{"value":"2026-09-21T18:53:19+01:00","precision":"instant","basis":"backfill:article_published"},"firstObservedAt":"2026-10-27T21:55:28.830Z","createdAt":"2026-11-05T17:28:23.390Z","historical":true,"entityIds":[],"topicIds":["artificial-general-intelligence"],"media":[],"sourceMedia":{"status":"unchecked","sourceUrl":"https://www.tobyord.com/writing/swarm-scaling","checkedAt":null,"items":[]},"editorialDecision":{"outcome":"accept","mediaId":null,"inputHash":"9595e564898a21a6bf5c1dc33a0ad30401e1df5d5d8b50363022ff456f2d8ec7","rubricVersion":"unrecorded","provider":"anthropic","model":"claude-opus-5-5","reasonCodes":["bridge_admitted_by_model","second_opinion_categorical"],"confidence":1,"decidedAt":"2026-10-27T21:55:26.139Z"},"provenance":{"textHash":"d06fb54ed4ef5618e9fa3da24df2e28abd4e2fe2665a9d5d40c004ac890f1c25","observationId":"fa7ebb77a1ef4ef5384e53eccaaf7e0b4314e5e5558b1a1a3b74d4d8d401d9d9","retrievedAt":"2026-10-27T21:57:24.407Z","kind":"publisher-page","preparer":{"model":"openai/gpt-6-luna","runId":"36196654-bfea-a79f-b03f-3898d152a7e6"},"reviewer":{"model":"anthropic/claude-opus-5.5","runId":"bd330171-7ef6-d5e4-7eaa-dbfb41dcbc21"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-05T17:23:50.790Z","model":"anthropic/claude-opus-5.5"}}}],"audits":[{"publicationId":"803ddfa4-72d4-3939-413f-3e3c316c9869","verdict":"uncertain","findings":[{"dimension":"attribution","status":"pass","explanation":"The source is a single essay by Toby Ord on his own site. The reading consistently attributes the arguments, estimates and concerns to Ord rather than presenting them as established fact or as Escape Meridian's view.","excerpt":"Swarm Scaling\nSeptember 21, 2026Toby Ord"},{"dimension":"summary","status":"uncertain","explanation":"The reading tracks Ord's argument faithfully, but a non-specialist may not follow it. λ is named only as the 'stepping on toes' parameter; the reading never says it measures how parallelisable a task is. 'Tokens', 'duration scaling', 'intelligence explosion' and 'orchestration methods' are also undefined and have no definitions entry. The reading omits that the estimates are inferred from OpenAI's GPT 5.6 Sol launch charts. Smallest repair: add one plain-language clause explaining λ, add 'tokens' and 'intelligence explosion' to the term library, and name the OpenAI chart data.","excerpt":"Here $\\lambda$ is a parameter measuring how parallelisable the task is. They call it the ‘stepping on toes’ parameter."},{"dimension":"date","status":"pass","explanation":"originalPublished is 2026-09-21T18:53:19+01:00, which is September 21, 2026 in UTC. This matches the source dateline and the reading's day-level date. The page was edited on 27/09/2026 to remove a closing Navier-Stokes section. The reading neither uses nor relies on that edit, and its cited passages are plainly part of the original body. There is no hindsight framing, and this is a publisher page, so no archive disclosure is needed.","excerpt":"Edit (27/09/2026): I originally closed this essay with a speculation on swarm scaling"},{"dimension":"support","status":"pass","explanation":"Each claim matches its excerpt: the efficiency shortfall, the 10x-to-3x–5x conversion, speed as the main demonstrated reason, the possible higher plateau, the link to recursive self-improvement, Ord's hope for lower values and his call for tracking. The λ values come from regressions Ord asked Claude Opus 5 to run, but Ord adopts them, so 'Ord estimates' is acceptable. Despite the later edit, the cited text is plainly original body text.","excerpt":"I’d hoped that the value of $\\lambda$ for AI agents would be lower, making an intelligence explosion less likely, but that appears to not be the case."},{"dimension":"media","status":"pass","explanation":"No media is selected, and no verified candidates were supplied, so a null media choice is appropriate.","excerpt":null}],"createdAt":"2026-11-05T17:23:50.790Z","model":"anthropic/claude-opus-5.5"}],"correctionUrl":"https://escapemeridian.com/corrections/"}