{"id":"c484f1a006df5982bd13e2c2409ed9016c3f037db7b96e9a00f9e60a69a6e97c","title":"Microsoft Research presents TokenEconomizer, which learns to cut coding-agent costs","argument":"Microsoft Research presents TokenEconomizer, an open-source plugin that learns from completed coding-agent sessions in a repository. Its premise is that agents often repeat work across sessions: rediscovering facts, loading unused tools and revisiting failures. The plugin adds repository-specific “micro-skills” after repeated evidence, prunes tools history suggests will go unused, and uses a judge model to identify failure loops and suggest a different approach. It works without modifying the agent or its underlying model; removed tools can be restored, and if the agent remains stuck after a suggestion, the plugin alerts the developer.\n\nIn evaluations on 277 SWE-Bench tasks from six repositories and five agent–model combinations, the researchers report average cost reductions of 18.4–30.9%, without performance loss in the tested repository-combination pairs. Early tasks could cost more while the plugin learns, but cumulative costs fell below the default in every repository by the sixth task. Reported savings include the plugin’s own model calls. The researchers also report lower costs on sequential web-development and security-auditing tasks. They say lowering reasoning effort saved similar or larger amounts but substantially degraded performance.","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The reported evaluation covers 277 SWE-Bench tasks from six repositories and five agent–model combinations; results beyond those tested settings are not established here. The source reports cost and performance outcomes from evaluations, not a guarantee of savings or task success in other use. The source says early tasks can cost more while TokenEconomizer learns, before cumulative costs fall below the default.","speaker":{"name":"Brenda Potts","roleAtSource":"Article author"},"source":{"title":"TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient - Microsoft Research","publisher":"Microsoft Research","format":"Article","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","publishedAt":"2026-11-16T17:01:35+00:00","dateBasis":"Date stated by the publisher."},"evidence":[{"text":"TokenEconomizer is an open-source plugin that learns from an agent’s past sessions in a repository and reduces waste without changing the agent or its underlying model.","excerpt":"TokenEconomizer is an open-source plugin that learns from an agent’s past sessions in a repository to reduce waste in future sessions, without modifying the agent or its underlying model.","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","relationship":"Source-stated"},{"text":"The plugin uses micro-skills, tool pruning, and failure-loop intervention to guide later sessions, with safeguards for restoring tools and alerting developers.","excerpt":"Tool pruning removes tools that repository history suggests the agent is unlikely to use, while always retaining core tools for running commands and editing files. Removed tools can be restored if needed.","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","relationship":"Source-stated"},{"text":"In evaluations on 277 SWE-Bench tasks, researchers report cost reductions of 18.4–30.9% across five agent–model combinations, with no performance loss in the tested repository-combination pairs.","excerpt":"TokenEconomizer lowered the average cost by 18.4–30.9%, as shown in Figure 3, and saved the most in all 30 repository-combination pairs without performance loss.","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","relationship":"Source-stated"},{"text":"The researchers report that cumulative costs fell below the default by the sixth task in every repository, although early tasks could cost more while the plugin learned.","excerpt":"Early tasks can cost more while TokenEconomizer learns, but cumulative costs fell below the default in every repository by the sixth task","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","relationship":"Source-stated"},{"text":"The researchers report lower costs on sequential web-development and security-auditing tasks as well.","excerpt":"The savings also carried over to sequential web development, with costs 16.0–21.2% lower on Web-Bench. On security auditing, where every task is in a new repository, costs were 6.4–19.6% lower.","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","relationship":"Source-stated"},{"text":"The results suggest that repository-specific experience can reduce token use without reducing model reasoning effort; the researchers report that lowering reasoning effort saved similar or larger amounts but substantially degraded performance.","excerpt":"By contrast, lowering the model’s reasoning effort saved similar or larger amounts but led to substantial performance degradation.","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","relationship":"Escape Meridian interpretation"}],"sourceMedia":{"status":"found","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","checkedAt":"2026-11-16T18:18:07.904Z","items":[{"type":"image","url":"https://www.microsoft.com/en-us/research/wp-content/uploads/2026/11/TokenEconomizer-BlogHeroFeature-1400x788-1.jpg","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","attribution":"Microsoft Research","basis":"publisher-embed","availability":"verified","label":"Six coding sessions with shrinking cost bars send their logs to TokenEconomizer, which uses micro-skills, tool pruning and failure-loop intervention to guide a…","verifiedAt":"2026-11-16T18:18:08.220Z"}],"note":"Media declared by the original publisher; availability checked separately from the editorial reading."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How would TokenEconomizer’s cost savings and task performance vary across repositories and coding tasks beyond those evaluated?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"Microsoft Research presents TokenEconomizer, which learns to cut coding-agent costs","summary":"Microsoft Research presents TokenEconomizer, an open-source plugin that learns from completed coding-agent sessions in a repository. Its premise is that agents often repeat work across sessions: rediscovering facts, loading unused tools and revisiting failures. The plugin adds repository-specific “micro-skills” after repeated evidence, prunes tools history suggests will go unused, and uses a judge model to identify failure loops and suggest a different approach. It works without modifying the agent or its underlying model; removed tools can be restored, and if the agent remains stuck after a suggestion, the plugin alerts the developer.\n\nIn evaluations on 277 SWE-Bench tasks from six repositories and five agent–model combinations, the researchers report average cost reductions of 18.4–30.9%, without performance loss in the tested repository-combination pairs. Early tasks could cost more while the plugin learns, but cumulative costs fell below the default in every repository by the sixth task. Reported savings include the plugin’s own model calls. The researchers also report lower costs on sequential web-development and security-auditing tasks. They say lowering reasoning effort saved similar or larger amounts but substantially degraded performance.","question":"How would TokenEconomizer’s cost savings and task performance vary across repositories and coding tasks beyond those evaluated?","claims":[{"text":"TokenEconomizer is an open-source plugin that learns from an agent’s past sessions in a repository and reduces waste without changing the agent or its underlying model.","excerpt":"TokenEconomizer is an open-source plugin that learns from an agent’s past sessions in a repository to reduce waste in future sessions, without modifying the agent or its underlying model.","support":"source-stated"},{"text":"The plugin uses micro-skills, tool pruning, and failure-loop intervention to guide later sessions, with safeguards for restoring tools and alerting developers.","excerpt":"Tool pruning removes tools that repository history suggests the agent is unlikely to use, while always retaining core tools for running commands and editing files. Removed tools can be restored if needed.","support":"source-stated"},{"text":"In evaluations on 277 SWE-Bench tasks, researchers report cost reductions of 18.4–30.9% across five agent–model combinations, with no performance loss in the tested repository-combination pairs.","excerpt":"TokenEconomizer lowered the average cost by 18.4–30.9%, as shown in Figure 3, and saved the most in all 30 repository-combination pairs without performance loss.","support":"source-stated"},{"text":"The researchers report that cumulative costs fell below the default by the sixth task in every repository, although early tasks could cost more while the plugin learned.","excerpt":"Early tasks can cost more while TokenEconomizer learns, but cumulative costs fell below the default in every repository by the sixth task","support":"source-stated"},{"text":"The researchers report lower costs on sequential web-development and security-auditing tasks as well.","excerpt":"The savings also carried over to sequential web development, with costs 16.0–21.2% lower on Web-Bench. On security auditing, where every task is in a new repository, costs were 6.4–19.6% lower.","support":"source-stated"},{"text":"The results suggest that repository-specific experience can reduce token use without reducing model reasoning effort; the researchers report that lowering reasoning effort saved similar or larger amounts but substantially degraded performance.","excerpt":"By contrast, lowering the model’s reasoning effort saved similar or larger amounts but led to substantial performance degradation.","support":"interpretation"}],"limitations":["The reported evaluation covers 277 SWE-Bench tasks from six repositories and five agent–model combinations; results beyond those tested settings are not established here.","The source reports cost and performance outcomes from evaluations, not a guarantee of savings or task success in other use.","The source says early tasks can cost more while TokenEconomizer learns, before cumulative costs fall below the default."],"mediaId":"97087fe3c15c5fa6c7d3911339af6ce91c2f0de12dcbe96c0b0ffd4aac16a8bb","correctionReason":"The summary was revised to foreground how the plugin learns and intervenes, retain the comparison with reduced reasoning effort, and clarify that reported results come from evaluated settings.","id":"c484f1a006df5982bd13e2c2409ed9016c3f037db7b96e9a00f9e60a69a6e97c","publicationId":"b768e55d-2685-413f-8c98-487d0bddfdba","version":3,"sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","sourceTitle":"TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient - Microsoft Research","publisher":"Microsoft Research","publishedAt":"2026-11-16T17:01:35+00:00","originalPublished":{"value":"2026-11-16T17:01:35+00:00","precision":"instant","basis":"publisher_page"},"firstObservedAt":"2026-11-16T17:03:40.126Z","createdAt":"2026-11-16T18:19:23.243Z","historical":false,"entityIds":["microsoft-research"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"97087fe3c15c5fa6c7d3911339af6ce91c2f0de12dcbe96c0b0ffd4aac16a8bb","kind":"article","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-16T18:18:08.220Z","method":"acquired_source","contentType":"text/html"},"recommendationReason":"Recommended by Jev for understanding this source; checked by Luna."},{"id":"901c26545d3dbc17a65b6a21bb95ca7f009582d7a5ae3f0b4d3fdaa4d52ba419","kind":"image","url":"https://www.microsoft.com/en-us/research/wp-content/uploads/2026/11/TokenEconomizer-BlogHeroFeature-1400x788-1.jpg","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","label":"Six coding sessions with shrinking cost bars send their logs to TokenEconomizer, which uses micro-sk","provenance":"publisher_page","verification":{"status":"verified","checkedAt":"2026-11-16T18:18:08.220Z","method":"publisher_association_and_head","contentType":"image/jpeg","cacheUntil":1794853088220},"recommendationReason":"Observed on the acquired publisher source; availability checked."}],"sourceMedia":{"status":"found","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","checkedAt":"2026-11-16T18:18:07.904Z","items":[{"type":"image","url":"https://www.microsoft.com/en-us/research/wp-content/uploads/2026/11/TokenEconomizer-BlogHeroFeature-1400x788-1.jpg","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","attribution":"Microsoft Research","basis":"publisher-embed","availability":"verified","label":"Six coding sessions with shrinking cost bars send their logs to TokenEconomizer, which uses micro-skills, tool pruning and failure-loop intervention to guide a…","verifiedAt":"2026-11-16T18:18:08.220Z"}],"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":"97087fe3c15c5fa6c7d3911339af6ce91c2f0de12dcbe96c0b0ffd4aac16a8bb","confidence":0.9,"provider":"typesafe","model":"jev-1.13.0","inputHash":"cef3364d1f76cb06ad68af0a4a1f39079908404c2e2765ba9477e690f753d7db","decidedAt":"2026-11-16T18:18:10.018Z"},"provenance":{"textHash":"d47030acdb0819f4a749fb53d26ebc482404aa7571a298ca1649adec222a50b8","observationId":"426a136326fe74c2f54beaf8ee4925c619d7c5a2b67b0d434abcde7f79cd77a1","retrievedAt":"2026-11-16T18:18:07.904Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"ba425e15-8b4c-4c5a-8d92-455e344ccec3"},"reviewer":{"model":"gpt-6-luna","runId":"25373ea2-730d-4fb9-8c8f-b1668198cef5"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-16T18:19:23.243Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Brenda Potts","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Brenda Potts","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Brenda Potts","status":"stated","provenance":{"basis":"article-jsonld","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","locator":"document/html[1]/head[0]/script[71]@3114.@graph[0].author"},"assertionId":"f0140ca607810369fce25557c81cfa904145a08ca3d075c138850d3349bf5328:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-16T18:18:07.904Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"f0140ca607810369fce25557c81cfa904145a08ca3d075c138850d3349bf5328","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","contentSha256":"778d1f2c5b9b6404011258404b37099143f9e4da58d20c42dc18accf1bc97315","textHash":"d47030acdb0819f4a749fb53d26ebc482404aa7571a298ca1649adec222a50b8","observedAt":"2026-11-16T18:18:07.904Z"}}],"participationCoverage":{"total":2,"returned":1,"complete":false,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","snapshotId":"f0140ca607810369fce25557c81cfa904145a08ca3d075c138850d3349bf5328","contentSha256":"778d1f2c5b9b6404011258404b37099143f9e4da58d20c42dc18accf1bc97315","textHash":"d47030acdb0819f4a749fb53d26ebc482404aa7571a298ca1649adec222a50b8","observedAt":"2026-11-16T18:18:07.904Z","basis":"publisher-page-metadata","sourceLinks":13,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/c484f1a006df5982bd13e2c2409ed9016c3f037db7b96e9a00f9e60a69a6e97c/source"}},"canonicalUrl":"https://escapemeridian.com/readings/microsoft-research-presents-tokeneconomizer-which-learns-to-cut-c484f1a0","formats":{"html":"/readings/microsoft-research-presents-tokeneconomizer-which-learns-to-cut-c484f1a0","markdown":"/readings/microsoft-research-presents-tokeneconomizer-which-learns-to-cut-c484f1a0.md","json":"/readings/microsoft-research-presents-tokeneconomizer-which-learns-to-cut-c484f1a0.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":"microsoft-research","name":"Microsoft Research","kind":"organization","context":"Primary technical research with named authors, papers and implementation details.","referenceUrls":["https://www.dwarkesh.com/p/satya-nadella","https://www.dwarkesh.com/p/grant-sanderson","https://www.microsoft.com/en-us/research/blog/proof-based-general-purpose-computing-with-magnetar/","https://www.microsoft.com/en-us/research/"],"path":"/entities/microsoft-research","json":"/entities/microsoft-research.json","markdown":"/entities/microsoft-research.md"}]},"researchQuestions":[],"withdrawn":false,"references":{"schemaVersion":1,"target":{"type":"reading","id":"c484f1a006df5982bd13e2c2409ed9016c3f037db7b96e9a00f9e60a69a6e97c","revision":"b768e55d-2685-413f-8c98-487d0bddfdba","revisionKind":"publication-id","title":"Microsoft Research presents TokenEconomizer, which learns to cut coding-agent costs","url":"https://escapemeridian.com/readings/microsoft-research-presents-tokeneconomizer-which-learns-to-cut-c484f1a0"},"source":{"type":"source","id":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","revision":"d47030acdb0819f4a749fb53d26ebc482404aa7571a298ca1649adec222a50b8","revisionKind":"sha256-source-text","title":"TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient - Microsoft Research","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/"},"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":"c484f1a006df5982bd13e2c2409ed9016c3f037db7b96e9a00f9e60a69a6e97c","title":"Microsoft Research presents TokenEconomizer, which learns to cut coding-agent costs","argument":"Microsoft Research presents TokenEconomizer, an open-source plugin that learns from completed coding-agent sessions in a repository. Its premise is that agents often repeat work across sessions: rediscovering facts, loading unused tools and revisiting failures. The plugin adds repository-specific “micro-skills” after repeated evidence, prunes tools history suggests will go unused, and uses a judge model to identify failure loops and suggest a different approach. It works without modifying the agent or its underlying model; removed tools can be restored, and if the agent remains stuck after a suggestion, the plugin alerts the developer.\n\nIn evaluations on 277 SWE-Bench tasks from six repositories and five agent–model combinations, the researchers report average cost reductions of 18.4–30.9%, without performance loss in the tested repository-combination pairs. Early tasks could cost more while the plugin learns, but cumulative costs fell below the default in every repository by the sixth task. Reported savings include the plugin’s own model calls. The researchers also report lower costs on sequential web-development and security-auditing tasks. They say lowering reasoning effort saved similar or larger amounts but substantially degraded performance.","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The reported evaluation covers 277 SWE-Bench tasks from six repositories and five agent–model combinations; results beyond those tested settings are not established here. The source reports cost and performance outcomes from evaluations, not a guarantee of savings or task success in other use. The source says early tasks can cost more while TokenEconomizer learns, before cumulative costs fall below the default.","speaker":{"name":"Brenda Potts","roleAtSource":"Article author"},"source":{"title":"TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient - Microsoft Research","publisher":"Microsoft Research","format":"Article","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","publishedAt":"2026-11-16T17:01:35+00:00","dateBasis":"Date stated by the publisher."},"evidence":[{"text":"TokenEconomizer is an open-source plugin that learns from an agent’s past sessions in a repository and reduces waste without changing the agent or its underlying model.","excerpt":"TokenEconomizer is an open-source plugin that learns from an agent’s past sessions in a repository to reduce waste in future sessions, without modifying the agent or its underlying model.","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","relationship":"Source-stated"},{"text":"The plugin uses micro-skills, tool pruning, and failure-loop intervention to guide later sessions, with safeguards for restoring tools and alerting developers.","excerpt":"Tool pruning removes tools that repository history suggests the agent is unlikely to use, while always retaining core tools for running commands and editing files. Removed tools can be restored if needed.","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","relationship":"Source-stated"},{"text":"In evaluations on 277 SWE-Bench tasks, researchers report cost reductions of 18.4–30.9% across five agent–model combinations, with no performance loss in the tested repository-combination pairs.","excerpt":"TokenEconomizer lowered the average cost by 18.4–30.9%, as shown in Figure 3, and saved the most in all 30 repository-combination pairs without performance loss.","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","relationship":"Source-stated"},{"text":"The researchers report that cumulative costs fell below the default by the sixth task in every repository, although early tasks could cost more while the plugin learned.","excerpt":"Early tasks can cost more while TokenEconomizer learns, but cumulative costs fell below the default in every repository by the sixth task","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","relationship":"Source-stated"},{"text":"The researchers report lower costs on sequential web-development and security-auditing tasks as well.","excerpt":"The savings also carried over to sequential web development, with costs 16.0–21.2% lower on Web-Bench. On security auditing, where every task is in a new repository, costs were 6.4–19.6% lower.","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","relationship":"Source-stated"},{"text":"The results suggest that repository-specific experience can reduce token use without reducing model reasoning effort; the researchers report that lowering reasoning effort saved similar or larger amounts but substantially degraded performance.","excerpt":"By contrast, lowering the model’s reasoning effort saved similar or larger amounts but led to substantial performance degradation.","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","relationship":"Escape Meridian interpretation"}],"sourceMedia":{"status":"found","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","checkedAt":"2026-11-16T18:18:07.904Z","items":[{"type":"image","url":"https://www.microsoft.com/en-us/research/wp-content/uploads/2026/11/TokenEconomizer-BlogHeroFeature-1400x788-1.jpg","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","attribution":"Microsoft Research","basis":"publisher-embed","availability":"verified","label":"Six coding sessions with shrinking cost bars send their logs to TokenEconomizer, which uses micro-skills, tool pruning and failure-loop intervention to guide a…","verifiedAt":"2026-11-16T18:18:08.220Z"}],"note":"Media declared by the original publisher; availability checked separately from the editorial reading."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How would TokenEconomizer’s cost savings and task performance vary across repositories and coding tasks beyond those evaluated?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"Microsoft Research presents TokenEconomizer, which learns to cut coding-agent costs","summary":"Microsoft Research presents TokenEconomizer, an open-source plugin that learns from completed coding-agent sessions in a repository. Its premise is that agents often repeat work across sessions: rediscovering facts, loading unused tools and revisiting failures. The plugin adds repository-specific “micro-skills” after repeated evidence, prunes tools history suggests will go unused, and uses a judge model to identify failure loops and suggest a different approach. It works without modifying the agent or its underlying model; removed tools can be restored, and if the agent remains stuck after a suggestion, the plugin alerts the developer.\n\nIn evaluations on 277 SWE-Bench tasks from six repositories and five agent–model combinations, the researchers report average cost reductions of 18.4–30.9%, without performance loss in the tested repository-combination pairs. Early tasks could cost more while the plugin learns, but cumulative costs fell below the default in every repository by the sixth task. Reported savings include the plugin’s own model calls. The researchers also report lower costs on sequential web-development and security-auditing tasks. They say lowering reasoning effort saved similar or larger amounts but substantially degraded performance.","question":"How would TokenEconomizer’s cost savings and task performance vary across repositories and coding tasks beyond those evaluated?","claims":[{"text":"TokenEconomizer is an open-source plugin that learns from an agent’s past sessions in a repository and reduces waste without changing the agent or its underlying model.","excerpt":"TokenEconomizer is an open-source plugin that learns from an agent’s past sessions in a repository to reduce waste in future sessions, without modifying the agent or its underlying model.","support":"source-stated"},{"text":"The plugin uses micro-skills, tool pruning, and failure-loop intervention to guide later sessions, with safeguards for restoring tools and alerting developers.","excerpt":"Tool pruning removes tools that repository history suggests the agent is unlikely to use, while always retaining core tools for running commands and editing files. Removed tools can be restored if needed.","support":"source-stated"},{"text":"In evaluations on 277 SWE-Bench tasks, researchers report cost reductions of 18.4–30.9% across five agent–model combinations, with no performance loss in the tested repository-combination pairs.","excerpt":"TokenEconomizer lowered the average cost by 18.4–30.9%, as shown in Figure 3, and saved the most in all 30 repository-combination pairs without performance loss.","support":"source-stated"},{"text":"The researchers report that cumulative costs fell below the default by the sixth task in every repository, although early tasks could cost more while the plugin learned.","excerpt":"Early tasks can cost more while TokenEconomizer learns, but cumulative costs fell below the default in every repository by the sixth task","support":"source-stated"},{"text":"The researchers report lower costs on sequential web-development and security-auditing tasks as well.","excerpt":"The savings also carried over to sequential web development, with costs 16.0–21.2% lower on Web-Bench. On security auditing, where every task is in a new repository, costs were 6.4–19.6% lower.","support":"source-stated"},{"text":"The results suggest that repository-specific experience can reduce token use without reducing model reasoning effort; the researchers report that lowering reasoning effort saved similar or larger amounts but substantially degraded performance.","excerpt":"By contrast, lowering the model’s reasoning effort saved similar or larger amounts but led to substantial performance degradation.","support":"interpretation"}],"limitations":["The reported evaluation covers 277 SWE-Bench tasks from six repositories and five agent–model combinations; results beyond those tested settings are not established here.","The source reports cost and performance outcomes from evaluations, not a guarantee of savings or task success in other use.","The source says early tasks can cost more while TokenEconomizer learns, before cumulative costs fall below the default."],"mediaId":"97087fe3c15c5fa6c7d3911339af6ce91c2f0de12dcbe96c0b0ffd4aac16a8bb","correctionReason":"The summary was revised to foreground how the plugin learns and intervenes, retain the comparison with reduced reasoning effort, and clarify that reported results come from evaluated settings.","id":"c484f1a006df5982bd13e2c2409ed9016c3f037db7b96e9a00f9e60a69a6e97c","publicationId":"b768e55d-2685-413f-8c98-487d0bddfdba","version":3,"sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","sourceTitle":"TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient - Microsoft Research","publisher":"Microsoft Research","publishedAt":"2026-11-16T17:01:35+00:00","originalPublished":{"value":"2026-11-16T17:01:35+00:00","precision":"instant","basis":"publisher_page"},"firstObservedAt":"2026-11-16T17:03:40.126Z","createdAt":"2026-11-16T18:19:23.243Z","historical":false,"entityIds":["microsoft-research"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"97087fe3c15c5fa6c7d3911339af6ce91c2f0de12dcbe96c0b0ffd4aac16a8bb","kind":"article","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-16T18:18:08.220Z","method":"acquired_source","contentType":"text/html"},"recommendationReason":"Recommended by Jev for understanding this source; checked by Luna."},{"id":"901c26545d3dbc17a65b6a21bb95ca7f009582d7a5ae3f0b4d3fdaa4d52ba419","kind":"image","url":"https://www.microsoft.com/en-us/research/wp-content/uploads/2026/11/TokenEconomizer-BlogHeroFeature-1400x788-1.jpg","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","label":"Six coding sessions with shrinking cost bars send their logs to TokenEconomizer, which uses micro-sk","provenance":"publisher_page","verification":{"status":"verified","checkedAt":"2026-11-16T18:18:08.220Z","method":"publisher_association_and_head","contentType":"image/jpeg","cacheUntil":1794853088220},"recommendationReason":"Observed on the acquired publisher source; availability checked."}],"sourceMedia":{"status":"found","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","checkedAt":"2026-11-16T18:18:07.904Z","items":[{"type":"image","url":"https://www.microsoft.com/en-us/research/wp-content/uploads/2026/11/TokenEconomizer-BlogHeroFeature-1400x788-1.jpg","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","attribution":"Microsoft Research","basis":"publisher-embed","availability":"verified","label":"Six coding sessions with shrinking cost bars send their logs to TokenEconomizer, which uses micro-skills, tool pruning and failure-loop intervention to guide a…","verifiedAt":"2026-11-16T18:18:08.220Z"}],"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":"97087fe3c15c5fa6c7d3911339af6ce91c2f0de12dcbe96c0b0ffd4aac16a8bb","confidence":0.9,"provider":"typesafe","model":"jev-1.13.0","inputHash":"cef3364d1f76cb06ad68af0a4a1f39079908404c2e2765ba9477e690f753d7db","decidedAt":"2026-11-16T18:18:10.018Z"},"provenance":{"textHash":"d47030acdb0819f4a749fb53d26ebc482404aa7571a298ca1649adec222a50b8","observationId":"426a136326fe74c2f54beaf8ee4925c619d7c5a2b67b0d434abcde7f79cd77a1","retrievedAt":"2026-11-16T18:18:07.904Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"ba425e15-8b4c-4c5a-8d92-455e344ccec3"},"reviewer":{"model":"gpt-6-luna","runId":"25373ea2-730d-4fb9-8c8f-b1668198cef5"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-16T18:19:23.243Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Brenda Potts","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Brenda Potts","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Brenda Potts","status":"stated","provenance":{"basis":"article-jsonld","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","locator":"document/html[1]/head[0]/script[71]@3114.@graph[0].author"},"assertionId":"f0140ca607810369fce25557c81cfa904145a08ca3d075c138850d3349bf5328:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-16T18:18:07.904Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"f0140ca607810369fce25557c81cfa904145a08ca3d075c138850d3349bf5328","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","contentSha256":"778d1f2c5b9b6404011258404b37099143f9e4da58d20c42dc18accf1bc97315","textHash":"d47030acdb0819f4a749fb53d26ebc482404aa7571a298ca1649adec222a50b8","observedAt":"2026-11-16T18:18:07.904Z"}}],"participationCoverage":{"total":2,"returned":1,"complete":false,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","snapshotId":"f0140ca607810369fce25557c81cfa904145a08ca3d075c138850d3349bf5328","contentSha256":"778d1f2c5b9b6404011258404b37099143f9e4da58d20c42dc18accf1bc97315","textHash":"d47030acdb0819f4a749fb53d26ebc482404aa7571a298ca1649adec222a50b8","observedAt":"2026-11-16T18:18:07.904Z","basis":"publisher-page-metadata","sourceLinks":13,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/c484f1a006df5982bd13e2c2409ed9016c3f037db7b96e9a00f9e60a69a6e97c/source"}}},{"id":"c484f1a006df5982bd13e2c2409ed9016c3f037db7b96e9a00f9e60a69a6e97c","title":"Microsoft Research says TokenEconomizer learns from coding sessions to reduce costs","argument":"Microsoft Research describes TokenEconomizer, an open-source plugin that learns from completed AI coding-agent sessions in a repository to reduce waste in later sessions, without changing the agent or its model. It adds repository-specific “micro-skills” after repeated evidence, prunes tools history suggests will go unused, and intervenes when a judge model identifies a failure loop. Removed tools can be restored; if the agent remains stuck after a suggested change, the plugin alerts the developer.\n\nIn evaluations spanning 277 SWE-Bench tasks, six repositories and five agent–model combinations, the researchers report average cost reductions of 18.4–30.9%, without performance loss in the tested repository-combination pairs. Cumulative costs fell below the default in every repository by the sixth task, though early tasks could cost more while the plugin learned. The reported savings include the plugin’s own model calls. The researchers also report lower costs on sequential web-development and security-auditing tasks. Escape Meridian interpretation: the comparison suggests that reducing reasoning effort is not equivalent to learning repository-specific ways to save costs; the researchers say reduced reasoning effort saved similar or larger amounts but substantially degraded performance.","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The reported evaluation covers 277 SWE-Bench tasks from six repositories and five agent–model combinations; results beyond those tested settings are not established here. The source reports cost and performance outcomes from evaluations, not a guarantee of savings or task success in other use. The source says early tasks can cost more while TokenEconomizer learns, before cumulative costs fall below the default.","speaker":{"name":"Brenda Potts","roleAtSource":"Article author"},"source":{"title":"TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient - Microsoft Research","publisher":"Microsoft Research","format":"Article","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","publishedAt":"2026-11-16T17:01:35+00:00","dateBasis":"Date stated by the publisher."},"evidence":[{"text":"TokenEconomizer is an open-source plugin that learns from an agent’s past sessions in a repository and reduces waste without changing the agent or its underlying model.","excerpt":"TokenEconomizer is an open-source plugin that learns from an agent’s past sessions in a repository to reduce waste in future sessions, without modifying the agent or its underlying model.","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","relationship":"Source-stated"},{"text":"The plugin uses micro-skills, tool pruning, and failure-loop intervention to guide later sessions, with safeguards for restoring tools and alerting developers.","excerpt":"Tool pruning removes tools that repository history suggests the agent is unlikely to use, while always retaining core tools for running commands and editing files. Removed tools can be restored if needed.","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","relationship":"Source-stated"},{"text":"In evaluations on 277 SWE-Bench tasks, researchers report cost reductions of 18.4–30.9% across five agent–model combinations, with no performance loss in the tested repository-combination pairs.","excerpt":"TokenEconomizer lowered the average cost by 18.4–30.9%, as shown in Figure 3, and saved the most in all 30 repository-combination pairs without performance loss.","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","relationship":"Source-stated"},{"text":"The researchers report that cumulative costs fell below the default by the sixth task in every repository, although early tasks could cost more while the plugin learned.","excerpt":"Early tasks can cost more while TokenEconomizer learns, but cumulative costs fell below the default in every repository by the sixth task","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","relationship":"Source-stated"},{"text":"The researchers report lower costs on sequential web-development and security-auditing tasks as well.","excerpt":"The savings also carried over to sequential web development, with costs 16.0–21.2% lower on Web-Bench. On security auditing, where every task is in a new repository, costs were 6.4–19.6% lower.","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","relationship":"Source-stated"},{"text":"The results suggest that repository-specific experience can reduce token use without reducing model reasoning effort; the researchers report that lowering reasoning effort saved similar or larger amounts but substantially degraded performance.","excerpt":"By contrast, lowering the model’s reasoning effort saved similar or larger amounts but led to substantial performance degradation.","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","relationship":"Escape Meridian interpretation"}],"sourceMedia":{"status":"found","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","checkedAt":"2026-11-16T18:02:10.275Z","items":[{"type":"image","url":"https://www.microsoft.com/en-us/research/wp-content/uploads/2026/11/TokenEconomizer-BlogHeroFeature-1400x788-1.jpg","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","attribution":"Microsoft Research","basis":"publisher-embed","availability":"verified","label":"Six coding sessions with shrinking cost bars send their logs to TokenEconomizer, which uses micro-skills, tool pruning and failure-loop intervention to guide a…","verifiedAt":"2026-11-16T18:02:10.716Z"}],"note":"Media declared by the original publisher; availability checked separately from the editorial reading."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How would TokenEconomizer’s cost savings and task performance vary across repositories and coding tasks beyond those evaluated?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"Microsoft Research says TokenEconomizer learns from coding sessions to reduce costs","summary":"Microsoft Research describes TokenEconomizer, an open-source plugin that learns from completed AI coding-agent sessions in a repository to reduce waste in later sessions, without changing the agent or its model. It adds repository-specific “micro-skills” after repeated evidence, prunes tools history suggests will go unused, and intervenes when a judge model identifies a failure loop. Removed tools can be restored; if the agent remains stuck after a suggested change, the plugin alerts the developer.\n\nIn evaluations spanning 277 SWE-Bench tasks, six repositories and five agent–model combinations, the researchers report average cost reductions of 18.4–30.9%, without performance loss in the tested repository-combination pairs. Cumulative costs fell below the default in every repository by the sixth task, though early tasks could cost more while the plugin learned. The reported savings include the plugin’s own model calls. The researchers also report lower costs on sequential web-development and security-auditing tasks. Escape Meridian interpretation: the comparison suggests that reducing reasoning effort is not equivalent to learning repository-specific ways to save costs; the researchers say reduced reasoning effort saved similar or larger amounts but substantially degraded performance.","question":"How would TokenEconomizer’s cost savings and task performance vary across repositories and coding tasks beyond those evaluated?","claims":[{"text":"TokenEconomizer is an open-source plugin that learns from an agent’s past sessions in a repository and reduces waste without changing the agent or its underlying model.","excerpt":"TokenEconomizer is an open-source plugin that learns from an agent’s past sessions in a repository to reduce waste in future sessions, without modifying the agent or its underlying model.","support":"source-stated"},{"text":"The plugin uses micro-skills, tool pruning, and failure-loop intervention to guide later sessions, with safeguards for restoring tools and alerting developers.","excerpt":"Tool pruning removes tools that repository history suggests the agent is unlikely to use, while always retaining core tools for running commands and editing files. Removed tools can be restored if needed.","support":"source-stated"},{"text":"In evaluations on 277 SWE-Bench tasks, researchers report cost reductions of 18.4–30.9% across five agent–model combinations, with no performance loss in the tested repository-combination pairs.","excerpt":"TokenEconomizer lowered the average cost by 18.4–30.9%, as shown in Figure 3, and saved the most in all 30 repository-combination pairs without performance loss.","support":"source-stated"},{"text":"The researchers report that cumulative costs fell below the default by the sixth task in every repository, although early tasks could cost more while the plugin learned.","excerpt":"Early tasks can cost more while TokenEconomizer learns, but cumulative costs fell below the default in every repository by the sixth task","support":"source-stated"},{"text":"The researchers report lower costs on sequential web-development and security-auditing tasks as well.","excerpt":"The savings also carried over to sequential web development, with costs 16.0–21.2% lower on Web-Bench. On security auditing, where every task is in a new repository, costs were 6.4–19.6% lower.","support":"source-stated"},{"text":"The results suggest that repository-specific experience can reduce token use without reducing model reasoning effort; the researchers report that lowering reasoning effort saved similar or larger amounts but substantially degraded performance.","excerpt":"By contrast, lowering the model’s reasoning effort saved similar or larger amounts but led to substantial performance degradation.","support":"interpretation"}],"limitations":["The reported evaluation covers 277 SWE-Bench tasks from six repositories and five agent–model combinations; results beyond those tested settings are not established here.","The source reports cost and performance outcomes from evaluations, not a guarantee of savings or task success in other use.","The source says early tasks can cost more while TokenEconomizer learns, before cumulative costs fall below the default."],"mediaId":"97087fe3c15c5fa6c7d3911339af6ce91c2f0de12dcbe96c0b0ffd4aac16a8bb","correctionReason":"The summary was revised to foreground how the plugin learns and intervenes, retain the comparison with reduced reasoning effort, and clarify that reported results come from evaluated settings.","id":"c484f1a006df5982bd13e2c2409ed9016c3f037db7b96e9a00f9e60a69a6e97c","publicationId":"3262db45-ae3a-46e8-8aad-1d3f54749f45","version":2,"sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","sourceTitle":"TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient - Microsoft Research","publisher":"Microsoft Research","publishedAt":"2026-11-16T17:01:35+00:00","originalPublished":{"value":"2026-11-16T17:01:35+00:00","precision":"instant","basis":"publisher_page"},"firstObservedAt":"2026-11-16T17:03:40.126Z","createdAt":"2026-11-16T18:03:37.632Z","historical":false,"entityIds":["microsoft-research"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"97087fe3c15c5fa6c7d3911339af6ce91c2f0de12dcbe96c0b0ffd4aac16a8bb","kind":"article","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-16T18:02:10.716Z","method":"acquired_source","contentType":"text/html"},"recommendationReason":"Recommended by Jev for understanding this source; checked by Luna."},{"id":"901c26545d3dbc17a65b6a21bb95ca7f009582d7a5ae3f0b4d3fdaa4d52ba419","kind":"image","url":"https://www.microsoft.com/en-us/research/wp-content/uploads/2026/11/TokenEconomizer-BlogHeroFeature-1400x788-1.jpg","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","label":"Six coding sessions with shrinking cost bars send their logs to TokenEconomizer, which uses micro-sk","provenance":"publisher_page","verification":{"status":"verified","checkedAt":"2026-11-16T18:02:10.716Z","method":"publisher_association_and_head","contentType":"image/jpeg","cacheUntil":1794852130716},"recommendationReason":"Observed on the acquired publisher source; availability checked."}],"sourceMedia":{"status":"found","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","checkedAt":"2026-11-16T18:02:10.275Z","items":[{"type":"image","url":"https://www.microsoft.com/en-us/research/wp-content/uploads/2026/11/TokenEconomizer-BlogHeroFeature-1400x788-1.jpg","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","attribution":"Microsoft Research","basis":"publisher-embed","availability":"verified","label":"Six coding sessions with shrinking cost bars send their logs to TokenEconomizer, which uses micro-skills, tool pruning and failure-loop intervention to guide a…","verifiedAt":"2026-11-16T18:02:10.716Z"}],"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":"97087fe3c15c5fa6c7d3911339af6ce91c2f0de12dcbe96c0b0ffd4aac16a8bb","confidence":0.86,"provider":"typesafe","model":"jev-1.13.0","inputHash":"ddc350d06be13efd20243c83e70176bce12ed7268be7687414303178fdcadf07","decidedAt":"2026-11-16T18:02:13.191Z"},"provenance":{"textHash":"328cc46dccc4b09b21c9c398359542c67f73751230644778f36fc9c9f15d3792","observationId":"6bc1818e245388399e53c619a021397abe3ebd2bb81917c282f701880ca16cc8","retrievedAt":"2026-11-16T18:02:10.275Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"8a42d195-879b-4398-8596-fa696389a5d8"},"reviewer":{"model":"gpt-6-luna","runId":"bca313dc-23b2-46ed-8ca1-4200e1843f43"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-16T18:03:37.632Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Brenda Potts","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Brenda Potts","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Brenda Potts","status":"stated","provenance":{"basis":"article-jsonld","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","locator":"document/html[1]/head[0]/script[71]@3114.@graph[0].author"},"assertionId":"cb8f6dc6f92dc96dcdd3590963edafc007fdae5fbec8a543a0c7daab35c16407:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-16T18:02:10.275Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"cb8f6dc6f92dc96dcdd3590963edafc007fdae5fbec8a543a0c7daab35c16407","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","contentSha256":"80802a81059a53944a93dbdd45cae9ad484c77ecfdc739121ae0a5c40145b8be","textHash":"328cc46dccc4b09b21c9c398359542c67f73751230644778f36fc9c9f15d3792","observedAt":"2026-11-16T18:02:10.275Z"}}],"participationCoverage":{"total":2,"returned":1,"complete":false,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","snapshotId":"cb8f6dc6f92dc96dcdd3590963edafc007fdae5fbec8a543a0c7daab35c16407","contentSha256":"80802a81059a53944a93dbdd45cae9ad484c77ecfdc739121ae0a5c40145b8be","textHash":"328cc46dccc4b09b21c9c398359542c67f73751230644778f36fc9c9f15d3792","observedAt":"2026-11-16T18:02:10.275Z","basis":"publisher-page-metadata","sourceLinks":13,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/c484f1a006df5982bd13e2c2409ed9016c3f037db7b96e9a00f9e60a69a6e97c/source"}}},{"id":"c484f1a006df5982bd13e2c2409ed9016c3f037db7b96e9a00f9e60a69a6e97c","title":"Microsoft Research presents TokenEconomizer, which learns from coding sessions to cut costs","argument":"Microsoft Research describes TokenEconomizer, an open-source plugin that uses an AI coding agent’s past work in a repository to reduce repeated effort in later sessions. It does not modify the agent or its underlying model. The plugin records repository-specific patterns as “micro-skills,” removes tools history suggests are unlikely to be used, and intervenes when a judge model identifies a failure loop. Developers can restore removed tools; if an agent remains stuck after a suggested change in approach, the plugin alerts them.\n\nIn tests on 277 SWE-Bench tasks across six repositories and five agent–model combinations, the researchers report average cost reductions of 18.4–30.9%. They say costs fell in every tested repository-combination pair without performance loss, while Pass@1 rose in four combinations. Savings include TokenEconomizer’s own model calls. Early tasks can cost more as the plugin learns, but cumulative costs fell below the default by the sixth task in every repository.\n\nThe researchers also report lower costs on sequential web-development tasks and security-auditing tasks, including tasks in new repositories. They say lowering the model’s reasoning effort saved similar or larger amounts but substantially degraded performance.","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The reported evaluation covers 277 SWE-Bench tasks from six repositories and five agent–model combinations; results beyond those tested settings are not established here. The source reports cost and performance outcomes from evaluations, not a guarantee of savings or task success in other use. The source says early tasks can cost more while TokenEconomizer learns, before cumulative costs fall below the default.","speaker":{"name":"Brenda Potts","roleAtSource":"Article author"},"source":{"title":"TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient - Microsoft Research","publisher":"Microsoft Research","format":"Article","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","publishedAt":"2026-11-16T17:01:35+00:00","dateBasis":"Date stated by the publisher."},"evidence":[{"text":"TokenEconomizer is an open-source plugin that learns from an agent’s past sessions in a repository and reduces waste without changing the agent or its underlying model.","excerpt":"TokenEconomizer is an open-source plugin that learns from an agent’s past sessions in a repository to reduce waste in future sessions, without modifying the agent or its underlying model.","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","relationship":"Source-stated"},{"text":"The plugin uses micro-skills, tool pruning, and failure-loop intervention to guide later sessions, with safeguards for restoring tools and alerting developers.","excerpt":"Tool pruning removes tools that repository history suggests the agent is unlikely to use, while always retaining core tools for running commands and editing files. Removed tools can be restored if needed.","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","relationship":"Source-stated"},{"text":"In evaluations on 277 SWE-Bench tasks, researchers report cost reductions of 18.4–30.9% across five agent–model combinations, with no performance loss in the tested repository-combination pairs.","excerpt":"TokenEconomizer lowered the average cost by 18.4–30.9%, as shown in Figure 3, and saved the most in all 30 repository-combination pairs without performance loss.","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","relationship":"Source-stated"},{"text":"The researchers report that cumulative costs fell below the default by the sixth task in every repository, although early tasks could cost more while the plugin learned.","excerpt":"Early tasks can cost more while TokenEconomizer learns, but cumulative costs fell below the default in every repository by the sixth task","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","relationship":"Source-stated"},{"text":"The researchers report lower costs on sequential web-development and security-auditing tasks as well.","excerpt":"The savings also carried over to sequential web development, with costs 16.0–21.2% lower on Web-Bench. On security auditing, where every task is in a new repository, costs were 6.4–19.6% lower.","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","relationship":"Source-stated"},{"text":"The results suggest that repository-specific experience can reduce token use without reducing model reasoning effort; the researchers report that lowering reasoning effort saved similar or larger amounts but substantially degraded performance.","excerpt":"By contrast, lowering the model’s reasoning effort saved similar or larger amounts but led to substantial performance degradation.","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","relationship":"Escape Meridian interpretation"}],"sourceMedia":{"status":"found","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","checkedAt":"2026-11-16T17:15:52.808Z","items":[{"type":"image","url":"https://www.microsoft.com/en-us/research/wp-content/uploads/2026/11/TokenEconomizer-BlogHeroFeature-1400x788-1.jpg","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","attribution":"Microsoft Research","basis":"publisher-embed","availability":"verified","label":"Six coding sessions with shrinking cost bars send their logs to TokenEconomizer, which uses micro-skills, tool pruning and failure-loop intervention to guide a…","verifiedAt":"2026-11-16T17:15:53.717Z"}],"note":"Media declared by the original publisher; availability checked separately from the editorial reading."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How would TokenEconomizer’s cost savings and task performance vary across repositories and coding tasks beyond those evaluated?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"Microsoft Research presents TokenEconomizer, which learns from coding sessions to cut costs","summary":"Microsoft Research describes TokenEconomizer, an open-source plugin that uses an AI coding agent’s past work in a repository to reduce repeated effort in later sessions. It does not modify the agent or its underlying model. The plugin records repository-specific patterns as “micro-skills,” removes tools history suggests are unlikely to be used, and intervenes when a judge model identifies a failure loop. Developers can restore removed tools; if an agent remains stuck after a suggested change in approach, the plugin alerts them.\n\nIn tests on 277 SWE-Bench tasks across six repositories and five agent–model combinations, the researchers report average cost reductions of 18.4–30.9%. They say costs fell in every tested repository-combination pair without performance loss, while Pass@1 rose in four combinations. Savings include TokenEconomizer’s own model calls. Early tasks can cost more as the plugin learns, but cumulative costs fell below the default by the sixth task in every repository.\n\nThe researchers also report lower costs on sequential web-development tasks and security-auditing tasks, including tasks in new repositories. They say lowering the model’s reasoning effort saved similar or larger amounts but substantially degraded performance.","question":"How would TokenEconomizer’s cost savings and task performance vary across repositories and coding tasks beyond those evaluated?","claims":[{"text":"TokenEconomizer is an open-source plugin that learns from an agent’s past sessions in a repository and reduces waste without changing the agent or its underlying model.","excerpt":"TokenEconomizer is an open-source plugin that learns from an agent’s past sessions in a repository to reduce waste in future sessions, without modifying the agent or its underlying model.","support":"source-stated"},{"text":"The plugin uses micro-skills, tool pruning, and failure-loop intervention to guide later sessions, with safeguards for restoring tools and alerting developers.","excerpt":"Tool pruning removes tools that repository history suggests the agent is unlikely to use, while always retaining core tools for running commands and editing files. Removed tools can be restored if needed.","support":"source-stated"},{"text":"In evaluations on 277 SWE-Bench tasks, researchers report cost reductions of 18.4–30.9% across five agent–model combinations, with no performance loss in the tested repository-combination pairs.","excerpt":"TokenEconomizer lowered the average cost by 18.4–30.9%, as shown in Figure 3, and saved the most in all 30 repository-combination pairs without performance loss.","support":"source-stated"},{"text":"The researchers report that cumulative costs fell below the default by the sixth task in every repository, although early tasks could cost more while the plugin learned.","excerpt":"Early tasks can cost more while TokenEconomizer learns, but cumulative costs fell below the default in every repository by the sixth task","support":"source-stated"},{"text":"The researchers report lower costs on sequential web-development and security-auditing tasks as well.","excerpt":"The savings also carried over to sequential web development, with costs 16.0–21.2% lower on Web-Bench. On security auditing, where every task is in a new repository, costs were 6.4–19.6% lower.","support":"source-stated"},{"text":"The results suggest that repository-specific experience can reduce token use without reducing model reasoning effort; the researchers report that lowering reasoning effort saved similar or larger amounts but substantially degraded performance.","excerpt":"By contrast, lowering the model’s reasoning effort saved similar or larger amounts but led to substantial performance degradation.","support":"interpretation"}],"limitations":["The reported evaluation covers 277 SWE-Bench tasks from six repositories and five agent–model combinations; results beyond those tested settings are not established here.","The source reports cost and performance outcomes from evaluations, not a guarantee of savings or task success in other use.","The source says early tasks can cost more while TokenEconomizer learns, before cumulative costs fall below the default."],"mediaId":"97087fe3c15c5fa6c7d3911339af6ce91c2f0de12dcbe96c0b0ffd4aac16a8bb","correctionReason":null,"id":"c484f1a006df5982bd13e2c2409ed9016c3f037db7b96e9a00f9e60a69a6e97c","publicationId":"793ae8ce-f7f6-4801-895a-19cd19990b74","version":1,"sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","sourceTitle":"TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient - Microsoft Research","publisher":"Microsoft Research","publishedAt":"2026-11-16T17:01:35+00:00","originalPublished":{"value":"2026-11-16T17:01:35+00:00","precision":"instant","basis":"publisher_page"},"firstObservedAt":"2026-11-16T17:03:40.126Z","createdAt":"2026-11-16T17:17:22.218Z","historical":false,"entityIds":["microsoft-research"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"97087fe3c15c5fa6c7d3911339af6ce91c2f0de12dcbe96c0b0ffd4aac16a8bb","kind":"article","url":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-16T17:15:53.717Z","method":"acquired_source","contentType":"text/html"},"recommendationReason":"Recommended by Jev for understanding this source; checked by Luna."},{"id":"901c26545d3dbc17a65b6a21bb95ca7f009582d7a5ae3f0b4d3fdaa4d52ba419","kind":"image","url":"https://www.microsoft.com/en-us/research/wp-content/uploads/2026/11/TokenEconomizer-BlogHeroFeature-1400x788-1.jpg","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","label":"Six coding sessions with shrinking cost bars send their logs to TokenEconomizer, which uses micro-sk","provenance":"publisher_page","verification":{"status":"verified","checkedAt":"2026-11-16T17:15:53.717Z","method":"publisher_association_and_head","contentType":"image/jpeg","cacheUntil":1794849353717},"recommendationReason":"Observed on the acquired publisher source; availability checked."}],"sourceMedia":{"status":"found","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","checkedAt":"2026-11-16T17:15:52.808Z","items":[{"type":"image","url":"https://www.microsoft.com/en-us/research/wp-content/uploads/2026/11/TokenEconomizer-BlogHeroFeature-1400x788-1.jpg","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","attribution":"Microsoft Research","basis":"publisher-embed","availability":"verified","label":"Six coding sessions with shrinking cost bars send their logs to TokenEconomizer, which uses micro-skills, tool pruning and failure-loop intervention to guide a…","verifiedAt":"2026-11-16T17:15:53.717Z"}],"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":"97087fe3c15c5fa6c7d3911339af6ce91c2f0de12dcbe96c0b0ffd4aac16a8bb","confidence":0.9,"provider":"typesafe","model":"jev-1.13.0","inputHash":"cef3364d1f76cb06ad68af0a4a1f39079908404c2e2765ba9477e690f753d7db","decidedAt":"2026-11-16T17:15:55.980Z"},"provenance":{"textHash":"d47030acdb0819f4a749fb53d26ebc482404aa7571a298ca1649adec222a50b8","observationId":"65eb0f2082b9759a22afec59ae678c4376485278237bcc990b89108f64d75a1d","retrievedAt":"2026-11-16T17:15:52.808Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"d395fea8-a680-449e-8fd9-db5a63f83557"},"reviewer":{"model":"gpt-6-luna","runId":"e0893936-8e5c-420b-8ea7-aba455215ba0"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-16T17:17:22.218Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Brenda Potts","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Brenda Potts","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Brenda Potts","status":"stated","provenance":{"basis":"article-jsonld","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","locator":"document/html[1]/head[0]/script[71]@3114.@graph[0].author"},"assertionId":"839caa3668278ae0d19c69e1e2d5f1987a5e55d1fccd529e963656b3830f5dad:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-16T17:15:52.808Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"839caa3668278ae0d19c69e1e2d5f1987a5e55d1fccd529e963656b3830f5dad","sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","contentSha256":"3f0d1174135b4896ad7823a9df99b7b62dee8cf189ea3e1e6c1d9825e61304b0","textHash":"d47030acdb0819f4a749fb53d26ebc482404aa7571a298ca1649adec222a50b8","observedAt":"2026-11-16T17:15:52.808Z"}}],"participationCoverage":{"total":2,"returned":1,"complete":false,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient/","snapshotId":"839caa3668278ae0d19c69e1e2d5f1987a5e55d1fccd529e963656b3830f5dad","contentSha256":"3f0d1174135b4896ad7823a9df99b7b62dee8cf189ea3e1e6c1d9825e61304b0","textHash":"d47030acdb0819f4a749fb53d26ebc482404aa7571a298ca1649adec222a50b8","observedAt":"2026-11-16T17:15:52.808Z","basis":"publisher-page-metadata","sourceLinks":13,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/c484f1a006df5982bd13e2c2409ed9016c3f037db7b96e9a00f9e60a69a6e97c/source"}}}],"audits":[{"publicationId":"793ae8ce-f7f6-4801-895a-19cd19990b74","verdict":"uncertain","findings":[{"dimension":"attribution","status":"pass","explanation":"The reading consistently attributes the plugin description and results to Microsoft Research and its researchers, and source findings remain reported claims. The headline names only Microsoft Research, although the plugin is described as coming from UC Berkeley, Microsoft Research, and Microsoft Security. Because the blog is a Microsoft Research publication and the draft does not claim sole authorship, this omission is minor. Adding the collaborators would be a modest fidelity improvement.","excerpt":"TokenEconomizer, an open-source plugin from UC Berkeley, Microsoft Research, and Microsoft Security"},{"dimension":"summary","status":"uncertain","explanation":"The draft says security-auditing tasks included tasks in new repositories, but the source says every security-auditing task was in a new repository. That detail matters because it shows savings without accumulated repository history. Smallest repair: change it to say every security-auditing task was in a new repository. SWE-Bench and Pass@1 also appear without plain explanation; tooltips or brief glosses would make the evaluation understandable to non-specialists. One possible gloss is that SWE-Bench is a benchmark of real bug fixes in open-source Python projects.","excerpt":"On security auditing, where every task is in a new repository, costs were 6.4–19.6% lower."},{"dimension":"date","status":"pass","explanation":"The evidence gives an original publication date of 2026-11-16 from the publisher page. The feed timestamp matches it, and nothing indicates an earlier first publication.","excerpt":null},{"dimension":"support","status":"uncertain","explanation":"The summary says developers can restore removed tools. The source only says, in passive voice, that removed tools can be restored if needed, so it does not establish who restores them; the controller might do it automatically. Smallest repair: say removed tools can be restored if needed. The draft's no-performance-loss wording follows the source's claim. However, the source also says it saved the most, apparently relative to compared approaches, and its at-a-glance section says success rates remained broadly stable. The draft does not contradict either point. The interpretation claim is labeled and supported.","excerpt":"Removed tools can be restored if needed."},{"dimension":"media","status":"pass","explanation":"The selected media is the verified original Microsoft Research blog article. It provides direct, relevant access to the source.","excerpt":null}],"createdAt":"2026-11-16T17:17:22.218Z","model":"anthropic/claude-opus-5.5"},{"publicationId":"3262db45-ae3a-46e8-8aad-1d3f54749f45","verdict":"uncertain","findings":[{"dimension":"attribution","status":"pass","explanation":"The post is published on Microsoft Research's blog, so 'Microsoft Research describes/says' fits, and results are framed as researcher reports. The source names the plugin as a joint UC Berkeley, Microsoft Research and Microsoft Security project. Mentioning UC Berkeley would be more complete, but its absence does not misattribute the work.","excerpt":"TokenEconomizer, an open-source plugin from UC Berkeley, Microsoft Research, and Microsoft Security, turns completed sessions into guidance for subsequent ones."},{"dimension":"summary","status":"uncertain","explanation":"The mechanisms are explained clearly, but core context for a non-specialist is missing. 'SWE-Bench' is never described as a benchmark of real bug fixes in open-source Python projects. The draft also never says the 'costs' are tokens consumed by every model interaction, which is the core problem the post addresses. Smallest repair: add 'real bug-fixing tasks from SWE-Bench' and one sentence saying agents pay in tokens for instructions, tool descriptions and history sent on each turn. The question is substantive and non-leading.","excerpt":"SWE-Bench, a benchmark of real bug fixes in open-source Python projects"},{"dimension":"date","status":"pass","explanation":"evidence.originalPublished (2026-11-16T17:01:35Z, publisher page) matches the feed entry. Nothing in the text suggests an earlier first publication of this post.","excerpt":null},{"dimension":"support","status":"uncertain","explanation":"The draft's 'without performance loss in the tested repository-combination pairs' stretches an ambiguous source sentence. That sentence could instead mean TokenEconomizer saved the most among methods without performance loss. Elsewhere the source says success rates 'remained broadly stable' and Pass@1 rose in four of five combinations, which leaves room for some decline. Smallest repair: say success rates remained broadly stable, with Pass@1 rising in four of five combinations. The reasoning-effort interpretation is labeled as interpretation and sourced. Other figures (18.4–30.9%, sixth task, Web-Bench, security auditing) match the source.","excerpt":"Costs fell in every repository tested, while task success rates remained broadly stable."},{"dimension":"media","status":"pass","explanation":"The selected media is the verified direct link to the original Microsoft Research blog post. That is the most direct way for readers to reach the source.","excerpt":null}],"createdAt":"2026-11-16T18:03:37.632Z","model":"anthropic/claude-opus-5.5"},{"publicationId":"b768e55d-2685-413f-8c98-487d0bddfdba","verdict":"uncertain","findings":[{"dimension":"attribution","status":"pass","explanation":"Presenting this as a Microsoft Research work is supported because it appears on the Microsoft Research blog. The source, however, calls it a joint plugin from UC Berkeley, Microsoft Research and Microsoft Security. Smallest repair: add 'with UC Berkeley and Microsoft Security' in the first sentence. Results are attributed to the researchers throughout.","excerpt":"TokenEconomizer, an open-source plugin from UC Berkeley, Microsoft Research, and Microsoft Security"},{"dimension":"summary","status":"uncertain","explanation":"The reading is faithful and clear about the mechanisms, but the central evidence term SWE-Bench is never explained. The source defines it as a benchmark of real bug fixes in open-source Python projects. 'Tokens', which drive the cost being cut, and 'reasoning effort' are also unexplained. A non-specialist cannot tell what was tested or why sessions cost money. Smallest repair: add a plain-language gloss of SWE-Bench and a note that costs come from tokens sent to the model each turn. The question is a fair open issue.","excerpt":"SWE-Bench, a benchmark of real bug fixes in open-source Python projects"},{"dimension":"date","status":"pass","explanation":"originalPublished 2026-11-16T17:01:35Z comes from the publisher page and matches the feed entry. Nothing in the source suggests an earlier first publication of this blog post.","excerpt":null},{"dimension":"support","status":"uncertain","explanation":"The draft says there was no performance loss in the tested repository-combination pairs. The source sentence is ambiguous: it says TokenEconomizer saved the most in all 30 pairs, which can be read as a comparison among non-degrading methods. Elsewhere the source says success rates remained 'broadly stable' and Pass@1 rose in four of five combinations. That implies possible small variation in the fifth combination. Smallest repair: say 'task success rates remained broadly stable (Pass@1 rose in four of five combinations)'. Other claims, including the labeled interpretation, are source-grounded.","excerpt":"Costs fell in every repository tested, while task success rates remained broadly stable."},{"dimension":"media","status":"pass","explanation":"The selected media is the verified direct article link to the original Microsoft Research blog post. It is the most direct access to the source.","excerpt":null}],"createdAt":"2026-11-16T18:19:23.243Z","model":"anthropic/claude-opus-5.5"}],"correctionUrl":"https://escapemeridian.com/corrections/"}