Brenda Potts · Nov 16, 2026
Microsoft Research argues coding agents can learn to spend less in familiar repositories
Markdown · JSON · Original source
presents TokenEconomizer, an open-source plugin developed with UC Berkeley and Microsoft Security. The researchers argue that coding agents returning to the same repository can use experience from earlier sessions to avoid repeatedly rediscovering facts, loading unused tools, and revisiting failures. The plugin learns from session logs and guides later work through micro-skills, tool pruning, and failure-loop intervention, without changing the agent or its underlying model. In 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%, with no performance loss in the tested repository-combination pairs. Reported savings include TokenEconomizer’s own model calls. Early tasks can cost more while it learns, but cumulative costs fell below the default in every repository by the sixth task. The researchers also report lower costs on sequential web-development and security-auditing tasks. They say lowering the model’s reasoning effort saved similar or larger amounts, but substantially degraded performance. Their results support repository experience as a way to reduce repeated computation in these evaluations; they do not establish outcomes in other settings.
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.
Original source
· Article
Source details
Original source date: Nov 16, 2026 (2026-11-16T17:01:35+00:00).
Date basis: Date stated by the publisher.
Added to Escape Meridian: 2026-11-16T23:47:11.152Z. First observed: 2026-11-16T17:03:40.126Z. Retrieved: 2026-11-16T23:46:01.527Z.
Referenced by
Recorded citations in public readings. A reference alone does not establish agreement or influence.
No public references recorded here yet.
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.
Open question
Do TokenEconomizer’s reported cost savings and task performance hold in repositories and agent–model combinations beyond those evaluated?
Source-stated
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.
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.
Source-stated
The plugin uses micro-skills, tool pruning, and failure-loop intervention to guide later sessions, with safeguards for restoring tools and alerting developers.
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.
Source-stated
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.
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.
Source-stated
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.
Early tasks can cost more while TokenEconomizer learns, but cumulative costs fell below the default in every repository by the sixth task
Source-stated
The researchers report lower costs on sequential web-development and security-auditing tasks as well.
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.
Source-stated
The researchers report that lowering the model’s reasoning effort saved similar or larger amounts but substantially degraded performance.
By contrast, lowering the model’s reasoning effort saved similar or larger amounts but led to substantial performance degradation.
Source and contributor context
Brenda Potts · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
Media declared by the original publisher; availability checked separately from the editorial reading.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Microsoft Research
Primary technical research with named authors, papers and implementation details.
Editorial provenance
Version 12. Publication ID: 48bb58df-4e5c-428e-8fe6-7c4a8986cef3.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-16T23:47:11.152Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: 0bf8f6f11e07d7f2570ca3fdba13d1e2bb7d9842d8b7521560d620718e85cbc2. Source text hash: 9eb55f9d664ed5e96fcd4650dcf1a5d907d554b73772b71dbafdc6912bb84c73.
Correction reason: The summary now foregrounds the researchers’ repository-specific cost-control argument and explains the plugin’s safeguards alongside its reported results and limits.
Revision and source-check history
Version 12 · 2026-11-16T23:47:11.152Z · The summary now foregrounds the researchers’ repository-specific cost-control argument and explains the plugin’s safeguards alongside its reported results and limits.
Microsoft Research argues coding agents can learn to spend less in familiar repositories
presents TokenEconomizer, an open-source plugin developed with UC Berkeley and Microsoft Security. The researchers argue that coding agents returning to the same repository can use experience from earlier sessions to avoid repeatedly rediscovering facts, loading unused tools, and revisiting failures. The plugin learns from session logs and guides later work through micro-skills, tool pruning, and failure-loop intervention, without changing the agent or its underlying model. In 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%, with no performance loss in the tested repository-combination pairs. Reported savings include TokenEconomizer’s own model calls. Early tasks can cost more while it learns, but cumulative costs fell below the default in every repository by the sixth task. The researchers also report lower costs on sequential web-development and security-auditing tasks. They say lowering the model’s reasoning effort saved similar or larger amounts, but substantially degraded performance. Their results support repository experience as a way to reduce repeated computation in these evaluations; they do not establish outcomes in other settings.
Original source
· Article
Source details
Original source date: Nov 16, 2026 (2026-11-16T17:01:35+00:00).
Date basis: Date stated by the publisher.
Added to Escape Meridian: 2026-11-16T23:47:11.152Z. First observed: 2026-11-16T17:03:40.126Z. Retrieved: 2026-11-16T23:46:01.527Z.
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.
Open question
Do TokenEconomizer’s reported cost savings and task performance hold in repositories and agent–model combinations beyond those evaluated?
Source-stated
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.
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.
Source-stated
The plugin uses micro-skills, tool pruning, and failure-loop intervention to guide later sessions, with safeguards for restoring tools and alerting developers.
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.
Source-stated
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.
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.
Source-stated
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.
Early tasks can cost more while TokenEconomizer learns, but cumulative costs fell below the default in every repository by the sixth task
Source-stated
The researchers report lower costs on sequential web-development and security-auditing tasks as well.
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.
Source-stated
The researchers report that lowering the model’s reasoning effort saved similar or larger amounts but substantially degraded performance.
By contrast, lowering the model’s reasoning effort saved similar or larger amounts but led to substantial performance degradation.
Source and contributor context
Brenda Potts · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
Media declared by the original publisher; availability checked separately from the editorial reading.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Microsoft Research
Primary technical research with named authors, papers and implementation details.
Editorial provenance
Version 12. Publication ID: 48bb58df-4e5c-428e-8fe6-7c4a8986cef3.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-16T23:47:11.152Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: 0bf8f6f11e07d7f2570ca3fdba13d1e2bb7d9842d8b7521560d620718e85cbc2. Source text hash: 9eb55f9d664ed5e96fcd4650dcf1a5d907d554b73772b71dbafdc6912bb84c73.
Correction reason: The summary now foregrounds the researchers’ repository-specific cost-control argument and explains the plugin’s safeguards alongside its reported results and limits.
Version 11 · 2026-11-16T23:31:40.404Z · The summary now foregrounds the researchers’ repository-specific cost-control argument and explains the plugin’s safeguards alongside its reported results and limits.
Microsoft Research argues coding agents can learn to spend less in familiar repositories
presents TokenEconomizer, an open-source plugin built with UC Berkeley and Microsoft Security. Its argument is that coding agents returning to the same repository can use experience from earlier sessions to avoid repeatedly rediscovering facts, loading unused tools, and revisiting failures. The plugin learns from session logs and guides later work through micro-skills, tool pruning, and failure-loop intervention, without changing the agent or its underlying model. In 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%, with no performance loss in the tested repository-combination pairs. The reported savings include TokenEconomizer’s own model calls. Early tasks can cost more while it learns, but cumulative costs fell below the default in every repository by the sixth task. The researchers also report lower costs on sequential web-development and security-auditing tasks. They say lowering the model’s reasoning effort saved similar or larger amounts, but substantially degraded performance. The results make repository experience a proposed way to reduce repeated computation while keeping the coding agent intact; the reported evaluations do not establish outcomes in other settings.
Original source
· Article
Source details
Original source date: Nov 16, 2026 (2026-11-16T17:01:35+00:00).
Date basis: Date stated by the publisher.
Added to Escape Meridian: 2026-11-16T23:31:40.404Z. First observed: 2026-11-16T17:03:40.126Z. Retrieved: 2026-11-16T23:30:49.179Z.
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.
Open question
Do TokenEconomizer’s reported cost savings and task performance hold in repositories and agent–model combinations beyond those evaluated?
Source-stated
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.
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.
Source-stated
The plugin uses micro-skills, tool pruning, and failure-loop intervention to guide later sessions, with safeguards for restoring tools and alerting developers.
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.
Source-stated
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.
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.
Source-stated
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.
Early tasks can cost more while TokenEconomizer learns, but cumulative costs fell below the default in every repository by the sixth task
Source-stated
The researchers report lower costs on sequential web-development and security-auditing tasks as well.
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.
Source-stated
The researchers report that lowering the model’s reasoning effort saved similar or larger amounts but substantially degraded performance.
By contrast, lowering the model’s reasoning effort saved similar or larger amounts but led to substantial performance degradation.
Source and contributor context
Brenda Potts · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
Media declared by the original publisher; availability checked separately from the editorial reading.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Microsoft Research
Primary technical research with named authors, papers and implementation details.
Editorial provenance
Version 11. Publication ID: 5a9a347c-b61b-46bf-8ccb-7cf36aa5a5d5.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-16T23:31:40.404Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: 7e07bb5f05266911c1e53c7386e6892845f8a4607319ef15bcf743b030d2ee4d. Source text hash: 0cbc0dbef3fef8d851dd10aaf5b0b736233c78e9344c9787e5538db1671188d8.
Correction reason: The summary now foregrounds the researchers’ repository-specific cost-control argument and explains the plugin’s safeguards alongside its reported results and limits.
Version 10 · 2026-11-16T23:17:51.056Z · The summary now foregrounds the researchers’ repository-specific cost-control argument and explains the plugin’s safeguards alongside its reported results and limits.
Microsoft Research presents TokenEconomizer for reducing coding-agent costs
presents TokenEconomizer, an open-source plugin that learns from coding agents’ past sessions in a repository. The researchers’ central idea is that agents can use this experience to avoid wasting effort when developers return to the same codebase, without changing the agent or its underlying model. After each session, TokenEconomizer updates its repository knowledge and applies micro-skills, tool pruning, and failure-loop intervention. Removed tools can be restored, and if an agent remains stuck after a suggested alternative, the plugin alerts the developer rather than stopping the work. In 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%, with no performance loss in the tested repository-combination pairs. The reported savings include the plugin’s own model calls. Early tasks can cost more while it learns; cumulative costs fell below the default in every repository by the sixth task. The researchers also report lower costs on sequential web-development and security-auditing tasks. They say lowering the model’s reasoning effort saved similar or larger amounts, but substantially degraded performance.
Original source
· Article
Source details
Original source date: Nov 16, 2026 (2026-11-16T17:01:35+00:00).
Date basis: Date stated by the publisher.
Added to Escape Meridian: 2026-11-16T23:17:51.056Z. First observed: 2026-11-16T17:03:40.126Z. Retrieved: 2026-11-16T23:17:06.575Z.
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.
Open question
Do TokenEconomizer’s reported cost savings and task performance hold in repositories and agent–model combinations beyond those evaluated?
Source-stated
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.
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.
Source-stated
The plugin uses micro-skills, tool pruning, and failure-loop intervention to guide later sessions, with safeguards for restoring tools and alerting developers.
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.
Source-stated
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.
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.
Source-stated
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.
Early tasks can cost more while TokenEconomizer learns, but cumulative costs fell below the default in every repository by the sixth task
Source-stated
The researchers report lower costs on sequential web-development and security-auditing tasks as well.
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.
Source-stated
The researchers report that lowering the model’s reasoning effort saved similar or larger amounts but substantially degraded performance.
By contrast, lowering the model’s reasoning effort saved similar or larger amounts but led to substantial performance degradation.
Source and contributor context
Brenda Potts · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
Media declared by the original publisher; availability checked separately from the editorial reading.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Microsoft Research
Primary technical research with named authors, papers and implementation details.
Editorial provenance
Version 10. Publication ID: 35142586-352c-405c-8655-941266c5c46e.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-16T23:17:51.056Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: 54b47bffe0e9029d9ae00a0f1d2ea97521624622f94df357c9988e49047c0750. Source text hash: 22258615ecf58352f71edaf0e6097c80d3062b756cc5c47a53812976bd1ececf.
Correction reason: The summary now foregrounds the researchers’ repository-specific cost-control argument and explains the plugin’s safeguards alongside its reported results and limits.
Version 9 · 2026-11-16T23:02:34.669Z · The summary now foregrounds the researchers’ repository-specific cost-control argument and explains the plugin’s safeguards alongside its reported results and limits.
Microsoft Research: TokenEconomizer learns from coding sessions to reduce costs
presents TokenEconomizer, an open-source plugin built around a practical observation: coding agents return to the same repositories but often repeat work from earlier sessions. The researchers argue that cost control should adapt to a repository’s history rather than apply one fixed strategy everywhere. After each session, TokenEconomizer updates what it has learned and uses micro-skills, tool pruning, and failure-loop intervention to guide later work, without modifying the coding agent or its underlying model. Removed tools can be restored. If an agent remains stuck after a suggested alternative, the plugin alerts the developer rather than stopping the work. Across 277 SWE-Bench tasks from six repositories and five agent–model combinations, the researchers report average cost reductions of 18.4–30.9%, with no performance loss in the tested repository-combination pairs. The reported savings include the plugin’s own model calls. Early tasks can cost more while it learns; cumulative costs fell below the default in every repository by the sixth task. The researchers also report lower costs on sequential web-development and security-auditing tasks. In their comparison, lowering the model’s reasoning effort saved similar or larger amounts but substantially degraded performance.
Original source
· Article
Source details
Original source date: Nov 16, 2026 (2026-11-16T17:01:35+00:00).
Date basis: Date stated by the publisher.
Added to Escape Meridian: 2026-11-16T23:02:34.669Z. First observed: 2026-11-16T17:03:40.126Z. Retrieved: 2026-11-16T23:00:59.226Z.
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.
Open question
Do TokenEconomizer’s reported cost savings and task performance hold in repositories and agent–model combinations beyond those evaluated?
Source-stated
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.
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.
Source-stated
The plugin uses micro-skills, tool pruning, and failure-loop intervention to guide later sessions, with safeguards for restoring tools and alerting developers.
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.
Source-stated
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.
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.
Source-stated
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.
Early tasks can cost more while TokenEconomizer learns, but cumulative costs fell below the default in every repository by the sixth task
Source-stated
The researchers report lower costs on sequential web-development and security-auditing tasks as well.
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.
Source-stated
The researchers report that lowering the model’s reasoning effort saved similar or larger amounts but substantially degraded performance.
By contrast, lowering the model’s reasoning effort saved similar or larger amounts but led to substantial performance degradation.
Source and contributor context
Brenda Potts · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
Media declared by the original publisher; availability checked separately from the editorial reading.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Microsoft Research
Primary technical research with named authors, papers and implementation details.
Editorial provenance
Version 9. Publication ID: 18190bc0-0d58-4242-8e21-5b713484197d.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-16T23:02:34.669Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: f633d613fdc80fc6a35551e8dea1f7a1432b65403929d81d1aae11ff24c12abf. Source text hash: 7b97310d4b9e88b1bcdb4b9b99100ce13bf1524e4980e5f3244f13c5da05416c.
Correction reason: The summary now foregrounds the researchers’ repository-specific cost-control argument and explains the plugin’s safeguards alongside its reported results and limits.
Version 8 · 2026-11-16T20:47:16.395Z · The summary was revised to foreground TokenEconomizer’s learning and intervention mechanisms, retain the comparison with lowering reasoning effort, and clarify that reported results come from evaluated settings.
Microsoft Research presents TokenEconomizer for lower-cost coding-agent sessions
presents TokenEconomizer, an open-source plugin that learns from completed sessions in a repository to guide later coding-agent work. The researchers say agents waste tokens rediscovering repository facts, loading unused tools, and repeating failed approaches. TokenEconomizer addresses these patterns with micro-skills, tool pruning, and failure-loop intervention, without modifying the agent or its underlying model. Removed tools can be restored; if an agent remains stuck after a suggested alternative, the plugin alerts the developer. Across 277 SWE-Bench tasks from six repositories and five agent–model combinations, the researchers report average cost reductions of 18.4–30.9%, with no performance loss in the tested repository-combination pairs. These savings include the plugin’s own model calls. Early tasks can cost more while it learns, but cumulative costs fell below the default in every repository by the sixth task. The researchers also report lower costs on sequential web-development and security-auditing tasks. Lowering the model’s reasoning effort saved similar or larger amounts in their comparison, but substantially degraded performance.
Original source
· Article
Source details
Original source date: Nov 16, 2026 (2026-11-16T17:01:35+00:00).
Date basis: Date stated by the publisher.
Added to Escape Meridian: 2026-11-16T20:47:16.395Z. First observed: 2026-11-16T17:03:40.126Z. Retrieved: 2026-11-16T20:46:18.310Z.
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.
Open question
Do TokenEconomizer’s reported cost savings and task performance hold in repositories and agent–model combinations beyond those evaluated?
Source-stated
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.
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.
Source-stated
The plugin uses micro-skills, tool pruning, and failure-loop intervention to guide later sessions, with safeguards for restoring tools and alerting developers.
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.
Source-stated
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.
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.
Source-stated
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.
Early tasks can cost more while TokenEconomizer learns, but cumulative costs fell below the default in every repository by the sixth task
Source-stated
The researchers report lower costs on sequential web-development and security-auditing tasks as well.
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.
Source-stated
The researchers report that lowering the model’s reasoning effort saved similar or larger amounts but substantially degraded performance.
By contrast, lowering the model’s reasoning effort saved similar or larger amounts but led to substantial performance degradation.
Source and contributor context
Brenda Potts · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
Media declared by the original publisher; availability checked separately from the editorial reading.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Microsoft Research
Primary technical research with named authors, papers and implementation details.
Editorial provenance
Version 8. Publication ID: 7e9fb9a6-2bf1-453c-8f79-757c1a955a3f.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-16T20:47:16.395Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: db50fb5c5ab78e017849174967296cacd934ff241b29533bc8228b1b2d73d16a. Source text hash: 618384a71bbbf72a79e300b041870c19a0150b814a58fd8a7ef4a6925f493650.
Correction reason: The summary was revised to foreground TokenEconomizer’s learning and intervention mechanisms, retain the comparison with lowering reasoning effort, and clarify that reported results come from evaluated settings.
Version 7 · 2026-11-16T20:32:24.622Z · The summary was revised to foreground TokenEconomizer’s learning and intervention mechanisms, retain the comparison with lowering reasoning effort, and clarify that reported results come from evaluated settings.
Microsoft Research describes TokenEconomizer’s approach to reducing coding-agent costs
presents TokenEconomizer, an open-source plugin that learns from completed coding-agent sessions in a repository. The researchers argue that agents waste tokens rediscovering facts, loading unused tools, and repeating unsuccessful approaches. The plugin turns session histories into guidance for later work through micro-skills, tool pruning, and failure-loop intervention. It leaves the agent and underlying model unchanged; removed tools can be restored, and developers are alerted if an agent remains stuck after a suggested alternative. In evaluations on 277 SWE-Bench tasks, a of real bug fixes in open-source Python projects, the researchers report average cost reductions of 18.4–30.9% across five agent–model combinations, with no performance loss in the tested repository-combination pairs. Early tasks can cost more while the plugin learns, but cumulative costs fell below the default in every repository by the sixth task. The 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 the model’s reasoning effort saved similar or larger amounts but substantially degraded performance. These results describe the tested settings; whether savings and task performance hold in other repositories and agent–model combinations remains open.
Original source
· Article
Source details
Original source date: Nov 16, 2026 (2026-11-16T17:01:35+00:00).
Date basis: Date stated by the publisher.
Added to Escape Meridian: 2026-11-16T20:32:24.622Z. First observed: 2026-11-16T17:03:40.126Z. Retrieved: 2026-11-16T20:31:01.211Z.
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.
Open question
Do the reported cost savings and task performance hold across repositories and agent–model combinations beyond those evaluated?
Source-stated
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.
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.
Source-stated
The plugin uses micro-skills, tool pruning, and failure-loop intervention to guide later sessions, with safeguards for restoring tools and alerting developers.
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.
Source-stated
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.
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.
Source-stated
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.
Early tasks can cost more while TokenEconomizer learns, but cumulative costs fell below the default in every repository by the sixth task
Source-stated
The researchers report lower costs on sequential web-development and security-auditing tasks as well.
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.
Source and contributor context
Brenda Potts · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
Media declared by the original publisher; availability checked separately from the editorial reading.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Benchmark
A defined set of tests used to compare systems. Results depend on the tasks and measures selected.
Microsoft Research
Primary technical research with named authors, papers and implementation details.
Editorial provenance
Version 7. Publication ID: 4b188ff7-55e7-443e-8b3c-aeb60eeff6e9.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-16T20:32:24.622Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: 9f9ea56cca0acc17689b4d70e4d715fb1e91b027971a8d3a0e6c5e09b2f60894. Source text hash: 7b97310d4b9e88b1bcdb4b9b99100ce13bf1524e4980e5f3244f13c5da05416c.
Correction reason: The summary was revised to foreground TokenEconomizer’s learning and intervention mechanisms, retain the comparison with lowering reasoning effort, and clarify that reported results come from evaluated settings.
Version 6 · 2026-11-16T20:16:44.126Z · The summary was revised to foreground TokenEconomizer’s learning and intervention mechanisms, retain the comparison with lowering reasoning effort, and clarify that reported results come from evaluated settings.
Microsoft Research describes TokenEconomizer’s approach to reducing coding-agent costs
presents TokenEconomizer, an open-source plugin that learns from completed coding-agent sessions in a repository. The researchers argue that agents waste tokens rediscovering facts, loading unused tools, and repeating unsuccessful approaches. TokenEconomizer turns session logs into repository-specific “micro-skills,” prunes tools history suggests are unlikely to be used, and intervenes when an agent appears stuck. Removed tools can be restored; if an agent remains stuck after a suggested alternative, the plugin alerts the developer rather than stopping the work. It does not modify the agent or its underlying model. In 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 can 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 contrast this approach with lowering reasoning effort, which they say saved similar or larger amounts but substantially degraded performance.
Original source
· Article
Source details
Original source date: Nov 16, 2026 (2026-11-16T17:01:35+00:00).
Date basis: Date stated by the publisher.
Added to Escape Meridian: 2026-11-16T20:16:44.126Z. First observed: 2026-11-16T17:03:40.126Z. Retrieved: 2026-11-16T20:15:51.564Z.
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.
Open question
Do the reported cost savings and task performance hold across repositories and agent–model combinations beyond those evaluated?
Source-stated
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.
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.
Source-stated
The plugin uses micro-skills, tool pruning, and failure-loop intervention to guide later sessions, with safeguards for restoring tools and alerting developers.
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.
Source-stated
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.
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.
Source-stated
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.
Early tasks can cost more while TokenEconomizer learns, but cumulative costs fell below the default in every repository by the sixth task
Source-stated
The researchers report lower costs on sequential web-development and security-auditing tasks as well.
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.
Source and contributor context
Brenda Potts · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
Media declared by the original publisher; availability checked separately from the editorial reading.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Microsoft Research
Primary technical research with named authors, papers and implementation details.
Editorial provenance
Version 6. Publication ID: 2aed5c6f-4ce8-48ed-8784-09479760728a.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-16T20:16:44.126Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: 8b6b97347acfc826e619c40bfc4b1199a3c79214418166a40d1d6650d54c7959. Source text hash: 31a7daa750462f9e364638e08be2e9beaf5c179450a295112807021bef27c821.
Correction reason: The summary was revised to foreground TokenEconomizer’s learning and intervention mechanisms, retain the comparison with lowering reasoning effort, and clarify that reported results come from evaluated settings.
Version 5 · 2026-11-16T19:47:20.024Z · 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.
Microsoft Research presents TokenEconomizer, which learns to reduce coding-agent costs
presents TokenEconomizer, an open-source plugin that learns from completed coding-agent sessions in a repository. The researchers argue that agents waste tokens rediscovering familiar facts, loading unused tools, and repeating failed approaches. The plugin uses session logs to build repository-specific “micro-skills,” prune tools, and detect when an agent is stuck; it can suggest another approach without modifying the agent or its underlying model. Removed tools can be restored, and if an agent remains stuck after a suggestion, TokenEconomizer alerts the developer. In 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 can cost more while the plugin learns, but cumulative costs fell below the default in every repository by the sixth task. The reported savings include TokenEconomizer’s own model calls. 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.
Original source
· Article
Source details
Original source date: Nov 16, 2026 (2026-11-16T17:01:35+00:00).
Date basis: Date stated by the publisher.
Added to Escape Meridian: 2026-11-16T19:47:20.024Z. First observed: 2026-11-16T17:03:40.126Z. Retrieved: 2026-11-16T19:46:24.952Z.
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.
Open question
How would TokenEconomizer’s cost savings and task performance vary across repositories and coding tasks beyond those evaluated?
Source-stated
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.
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.
Source-stated
The plugin uses micro-skills, tool pruning, and failure-loop intervention to guide later sessions, with safeguards for restoring tools and alerting developers.
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.
Source-stated
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.
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.
Source-stated
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.
Early tasks can cost more while TokenEconomizer learns, but cumulative costs fell below the default in every repository by the sixth task
Source-stated
The researchers report lower costs on sequential web-development and security-auditing tasks as well.
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.
Source and contributor context
Brenda Potts · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
Media declared by the original publisher; availability checked separately from the editorial reading.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Microsoft Research
Primary technical research with named authors, papers and implementation details.
Editorial provenance
Version 5. Publication ID: d995de36-d69f-43ac-8cd9-86b4b5ae9df4.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-16T19:47:20.024Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: 2128ecfa9bb98f6cfaa60cae2fdf1806e873098f3831cf86b5b40815a5dd6f6b. Source text hash: 85f9453bf1206984ea028f15d00f77ff90858fada28765c7cda01bde44042098.
Correction reason: 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.
Version 4 · 2026-11-16T19:04:17.117Z · 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.
Microsoft Research presents TokenEconomizer, which learns to reduce coding-agent costs
presents TokenEconomizer, an open-source plugin that learns from completed coding-agent sessions in a repository. The researchers’ premise is that agents can waste tokens rediscovering repository facts, loading tools they do not use, and repeating failed approaches. TokenEconomizer turns session logs into repository-specific “micro-skills,” prunes tools the agent is unlikely to need, and uses a judge model to detect when an agent is stuck and suggest another approach. It works without modifying the agent or its underlying model. Removed tools can be restored; if an agent remains stuck after a suggestion, the plugin alerts the developer. In 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 can 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.
Original source
· Article
Source details
Original source date: Nov 16, 2026 (2026-11-16T17:01:35+00:00).
Date basis: Date stated by the publisher.
Added to Escape Meridian: 2026-11-16T19:04:17.117Z. First observed: 2026-11-16T17:03:40.126Z. Retrieved: 2026-11-16T19:02:44.005Z.
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.
Open question
How would TokenEconomizer’s cost savings and task performance vary across repositories and coding tasks beyond those evaluated?
Source-stated
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.
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.
Source-stated
The plugin uses micro-skills, tool pruning, and failure-loop intervention to guide later sessions, with safeguards for restoring tools and alerting developers.
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.
Source-stated
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.
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.
Source-stated
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.
Early tasks can cost more while TokenEconomizer learns, but cumulative costs fell below the default in every repository by the sixth task
Source-stated
The researchers report lower costs on sequential web-development and security-auditing tasks as well.
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.
Source and contributor context
Brenda Potts · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
Media declared by the original publisher; availability checked separately from the editorial reading.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Microsoft Research
Primary technical research with named authors, papers and implementation details.
Editorial provenance
Version 4. Publication ID: bc1db08d-9e03-4bb9-88e0-23b26bd75552.
Prepared by gpt-6-luna; stream-sentence-removal-v1: removed claim 6; grounding review by gpt-6-luna.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-16T19:04:17.117Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: e9ff3bb4067a49fdbf6ffc7c02d9b1f8a92c0ff787ff507036981389341af706. Source text hash: 0802273a3be4df59b02d613e401b77b3c89ef1d2b76325fdd184abb125659dd8.
Correction reason: 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.
Version 3 · 2026-11-16T18:19:23.243Z · 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.
Microsoft Research presents TokenEconomizer, which learns to cut coding-agent costs
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. In 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.
Original source
· Article
Source details
Original source date: Nov 16, 2026 (2026-11-16T17:01:35+00:00).
Date basis: Date stated by the publisher.
Added to Escape Meridian: 2026-11-16T18:19:23.243Z. First observed: 2026-11-16T17:03:40.126Z. Retrieved: 2026-11-16T18:18:07.904Z.
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.
Open question
How would TokenEconomizer’s cost savings and task performance vary across repositories and coding tasks beyond those evaluated?
Source-stated
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.
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.
Source-stated
The plugin uses micro-skills, tool pruning, and failure-loop intervention to guide later sessions, with safeguards for restoring tools and alerting developers.
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.
Source-stated
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.
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.
Source-stated
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.
Early tasks can cost more while TokenEconomizer learns, but cumulative costs fell below the default in every repository by the sixth task
Source-stated
The researchers report lower costs on sequential web-development and security-auditing tasks as well.
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.
Escape Meridian interpretation
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.
By contrast, lowering the model’s reasoning effort saved similar or larger amounts but led to substantial performance degradation.
Source and contributor context
Brenda Potts · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
Media declared by the original publisher; availability checked separately from the editorial reading.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Microsoft Research
Primary technical research with named authors, papers and implementation details.
Editorial provenance
Version 3. Publication ID: b768e55d-2685-413f-8c98-487d0bddfdba.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-16T18:19:23.243Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: 426a136326fe74c2f54beaf8ee4925c619d7c5a2b67b0d434abcde7f79cd77a1. Source text hash: d47030acdb0819f4a749fb53d26ebc482404aa7571a298ca1649adec222a50b8.
Correction reason: 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.
Version 2 · 2026-11-16T18:03:37.632Z · 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.
Microsoft Research says TokenEconomizer learns from coding sessions to reduce costs
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. In 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.
Original source
· Article
Source details
Original source date: Nov 16, 2026 (2026-11-16T17:01:35+00:00).
Date basis: Date stated by the publisher.
Added to Escape Meridian: 2026-11-16T18:03:37.632Z. First observed: 2026-11-16T17:03:40.126Z. Retrieved: 2026-11-16T18:02:10.275Z.
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.
Open question
How would TokenEconomizer’s cost savings and task performance vary across repositories and coding tasks beyond those evaluated?
Source-stated
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.
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.
Source-stated
The plugin uses micro-skills, tool pruning, and failure-loop intervention to guide later sessions, with safeguards for restoring tools and alerting developers.
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.
Source-stated
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.
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.
Source-stated
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.
Early tasks can cost more while TokenEconomizer learns, but cumulative costs fell below the default in every repository by the sixth task
Source-stated
The researchers report lower costs on sequential web-development and security-auditing tasks as well.
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.
Escape Meridian interpretation
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.
By contrast, lowering the model’s reasoning effort saved similar or larger amounts but led to substantial performance degradation.
Source and contributor context
Brenda Potts · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
Media declared by the original publisher; availability checked separately from the editorial reading.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Microsoft Research
Primary technical research with named authors, papers and implementation details.
Editorial provenance
Version 2. Publication ID: 3262db45-ae3a-46e8-8aad-1d3f54749f45.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-16T18:03:37.632Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: 6bc1818e245388399e53c619a021397abe3ebd2bb81917c282f701880ca16cc8. Source text hash: 328cc46dccc4b09b21c9c398359542c67f73751230644778f36fc9c9f15d3792.
Correction reason: 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.
Version 1 · 2026-11-16T17:17:22.218Z
Microsoft Research presents TokenEconomizer, which learns from coding sessions to cut costs
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. In 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. The 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.
Original source
· Article
Source details
Original source date: Nov 16, 2026 (2026-11-16T17:01:35+00:00).
Date basis: Date stated by the publisher.
Added to Escape Meridian: 2026-11-16T17:17:22.218Z. First observed: 2026-11-16T17:03:40.126Z. Retrieved: 2026-11-16T17:15:52.808Z.
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.
Open question
How would TokenEconomizer’s cost savings and task performance vary across repositories and coding tasks beyond those evaluated?
Source-stated
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.
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.
Source-stated
The plugin uses micro-skills, tool pruning, and failure-loop intervention to guide later sessions, with safeguards for restoring tools and alerting developers.
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.
Source-stated
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.
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.
Source-stated
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.
Early tasks can cost more while TokenEconomizer learns, but cumulative costs fell below the default in every repository by the sixth task
Source-stated
The researchers report lower costs on sequential web-development and security-auditing tasks as well.
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.
Escape Meridian interpretation
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.
By contrast, lowering the model’s reasoning effort saved similar or larger amounts but led to substantial performance degradation.
Source and contributor context
Brenda Potts · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
Media declared by the original publisher; availability checked separately from the editorial reading.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Microsoft Research
Primary technical research with named authors, papers and implementation details.
Editorial provenance
Version 1. Publication ID: 793ae8ce-f7f6-4801-895a-19cd19990b74.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-16T17:17:22.218Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: 65eb0f2082b9759a22afec59ae678c4376485278237bcc990b89108f64d75a1d. Source text hash: d47030acdb0819f4a749fb53d26ebc482404aa7571a298ca1649adec222a50b8.
uncertain · 2026-11-16T17:17:22.218Z · anthropic/claude-opus-5.5
Publication: 793ae8ce-f7f6-4801-895a-19cd19990b74
attribution — pass: 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.
TokenEconomizer, an open-source plugin from UC Berkeley, Microsoft Research, and Microsoft Security
summary — uncertain: 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.
On security auditing, where every task is in a new repository, costs were 6.4–19.6% lower.
date — pass: 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.
support — uncertain: 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.
Removed tools can be restored if needed.
media — pass: The selected media is the verified original Microsoft Research blog article. It provides direct, relevant access to the source.
uncertain · 2026-11-16T18:03:37.632Z · anthropic/claude-opus-5.5
Publication: 3262db45-ae3a-46e8-8aad-1d3f54749f45
attribution — pass: 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.
TokenEconomizer, an open-source plugin from UC Berkeley, Microsoft Research, and Microsoft Security, turns completed sessions into guidance for subsequent ones.
summary — uncertain: 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.
SWE-Bench, a benchmark of real bug fixes in open-source Python projects
date — pass: 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.
support — uncertain: 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.
Costs fell in every repository tested, while task success rates remained broadly stable.
media — pass: 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.
uncertain · 2026-11-16T18:19:23.243Z · anthropic/claude-opus-5.5
Publication: b768e55d-2685-413f-8c98-487d0bddfdba
attribution — pass: 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.
TokenEconomizer, an open-source plugin from UC Berkeley, Microsoft Research, and Microsoft Security
summary — uncertain: 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.
SWE-Bench, a benchmark of real bug fixes in open-source Python projects
date — pass: 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.
support — uncertain: 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.
Costs fell in every repository tested, while task success rates remained broadly stable.
media — pass: 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.
passed · 2026-11-16T19:04:17.117Z · anthropic/claude-opus-5.5
Publication: bc1db08d-9e03-4bb9-88e0-23b26bd75552
attribution — pass: The reading presents TokenEconomizer as described on the Microsoft Research blog, and it attributes findings to 'the researchers.' The source also credits UC Berkeley and Microsoft Security. Naming them would be more complete, but presenting it as a Microsoft Research presentation is not wrong.
TokenEconomizer, an open-source plugin from UC Berkeley, Microsoft Research, and Microsoft Security, turns completed sessions into guidance for subsequent ones.
summary — pass: The summary leads with the actor and the development. It explains the three waste patterns and the three mechanisms in plain language, and it keeps the conditions: early tasks cost more, the plugin's own model calls are counted, and the reduced-reasoning comparison is included. Terms like tokens and SWE-Bench can be handled by tooltips. A small optional repair: noting that security-auditing tasks each used a new repository would clarify the smaller savings there. The question is a genuine open issue about results beyond the tested settings.
Early tasks can cost more while TokenEconomizer learns, but cumulative costs fell below the default in every repository by the sixth task
date — pass: originalPublished is 2026-11-16T17:01:35Z, based on the publisher page. It matches the feed timestamp, and no earlier first publication is indicated.
support — pass: Each claim matches the source text: the plugin's description, the mechanisms and safeguards, the 18.4–30.9% savings on 277 tasks, the sixth-task cumulative result, and the Web-Bench and security-auditing figures. The reading reports no performance loss across the 30 pairs, consistent with the source. The limitations correctly confine results to the evaluated settings without inventing benefits.
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.
media — pass: The selected media is the verified direct link to the original Microsoft Research blog post. That is the most direct access to the source.
uncertain · 2026-11-16T19:47:20.024Z · anthropic/claude-opus-5.5
Publication: d995de36-d69f-43ac-8cd9-86b4b5ae9df4
attribution — uncertain: The title and lead say 'Microsoft Research presents TokenEconomizer.' The post does appear on the Microsoft Research blog, but the source describes the plugin as a joint work of UC Berkeley, Microsoft Research and Microsoft Security. The authors listed include Berkeley professors. Smallest repair: say 'an open-source plugin from UC Berkeley, Microsoft Research and Microsoft Security' in the lead. Results remain correctly framed as the researchers' reports.
TokenEconomizer, an open-source plugin from UC Berkeley, Microsoft Research, and Microsoft Security, turns completed sessions into guidance for subsequent ones.
summary — uncertain: The reading is faithful, but several central terms are left unexplained for a general reader. 'Micro-skills' are one-line repository notes, SWE-Bench is a benchmark of real bug fixes in Python projects, and 'reasoning effort' and 'tokens' are not explained. The draft also drops the profiling evidence behind the argument: 30% repeated actions, 16 of 23 tools never called, and failed tasks taking 32% of spending. Smallest repair: briefly define micro-skills and SWE-Bench, and restore one or two of those profiling figures.
Micro-skills capture repository-specific knowledge in one-line notes.
date — pass: The originalPublished value of 2026-11-16T17:01:35Z comes from the publisher page, matches the feed entry, and is consistent with first observation shortly afterward. No earlier publication is indicated.
support — pass: Each claim matches a source statement. These include the cost reductions of 18.4–30.9%, 'without performance loss' across 30 pairs, cumulative savings by the sixth task, and the web-development and security-auditing figures. They also include the inclusion of the plugin's own model calls and the reasoning-effort comparison. The limitations and the question stay within the evidence and do not overclaim real-world reliability.
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.
media — pass: The selected media is the verified link to the original Microsoft Research blog post. It gives readers the most direct access to the actual source.
uncertain · 2026-11-16T20:16:44.126Z · anthropic/claude-opus-5.5
Publication: 2aed5c6f-4ce8-48ed-8784-09479760728a
attribution — pass: The draft attributes the work and its results to Microsoft Research and 'the researchers', and frames evaluation outcomes as reported. The source names a UC Berkeley, Microsoft Research and Microsoft Security collaboration. The draft does not mention that, but this is a minor omission rather than a misattribution, since the Microsoft Research blog is the publisher and presenter.
TokenEconomizer, an open-source plugin from UC Berkeley, Microsoft Research, and Microsoft Security, turns completed sessions into guidance for subsequent ones.
summary — uncertain: The summary is faithful, but key terms stay opaque for a general reader. "Micro-skills" is the source's own coinage, quoted without explanation and unlikely to be in a term library; the source defines micro-skills as one-line notes added to the starting context once confirmed. "Tokens" is not tied to cost, although the source says each model interaction consumes them. Smallest repair: briefly gloss micro-skills as confirmed one-line repository notes, and say that tokens are the metered model usage that drives cost. The title and question are neutral and supported.
Micro-skills capture repository-specific knowledge in one-line notes.
date — pass: The originalPublished value is 2026-11-16T17:01:35Z, taken from the publisher page. It matches the feed entry, and nothing in the source indicates an earlier first publication. The draft adds no conflicting date.
support — pass: Each claim is source-stated and matches its excerpt: the 18.4–30.9% savings, no performance loss across 30 pairs, cumulative costs below default by the sixth task, the Web-Bench and security-auditing figures, the comparison with lowering reasoning effort, and the inclusion of the plugin's own model calls. The limitations correctly confine the results to the tested settings. The draft makes no unsupported benefit or cost claims.
Reported savings also account for the cost of TokenEconomizer’s own model calls.
media — pass: The selected media is the verified direct link to the original Microsoft Research blog post. This is the most direct access to the source.
uncertain · 2026-11-16T20:32:24.622Z · anthropic/claude-opus-5.5
Publication: 4b188ff7-55e7-443e-8b3c-aeb60eeff6e9
attribution — pass: The reading attributes the blog presentation to Microsoft Research and keeps the findings as the researchers' reported results. The source names the plugin as a joint effort of UC Berkeley, Microsoft Research and Microsoft Security. 'Microsoft Research presents' fits the publisher role. Optional repair: mention the collaborators.
TokenEconomizer, an open-source plugin from UC Berkeley, Microsoft Research, and Microsoft Security
summary — uncertain: The structure is faithful, but central terms are not explained for a general reader. 'Tokens' are not described as the metered units of text that drive model cost. Micro-skills, tool pruning and failure-loop intervention are only named, so readers cannot tell what each mechanism does, such as one-line repository notes, dropping unused tool descriptions, or a judge model spotting stuck loops. Smallest repair: give a short plain-language gloss for tokens and each of the three mechanisms. The title and question are fair and not leading.
every interaction with their models consumes tokens. In each session, an agent sends instructions, tool descriptions, and session history to its model.
date — pass: The originalPublished value of 2026-11-16T17:01:35Z is based on the publisher page and matches the feed entry. The draft makes no conflicting date claim.
support — pass: The claims are source-stated and attributed: 277 tasks; 18.4–30.9% savings; the 'without performance loss' wording; cumulative savings by the sixth task; savings that include the plugin's own model calls; Web-Bench and security results; and the reasoning-effort comparison. The limitations and the question stay within the evidence. A minor scope point is that the developer alert applies specifically 'in the open-source plugin', but this is not material.
saved the most in all 30 repository-combination pairs without performance loss
media — pass: The selected media is the verified original Microsoft Research article. This is the direct primary source and the most accessible one.
uncertain · 2026-11-16T20:47:16.395Z · anthropic/claude-opus-5.5
Publication: 7e9fb9a6-2bf1-453c-8f79-757c1a955a3f
attribution — uncertain: The reading credits the plugin to Microsoft Research, the blog's publisher, and properly frames its results as the researchers' reports. However, the source describes TokenEconomizer as coming from UC Berkeley, Microsoft Research and Microsoft Security. Smallest repair: say Microsoft Research describes TokenEconomizer, an open-source plugin from UC Berkeley, Microsoft Research and Microsoft Security.
TokenEconomizer, an open-source plugin from UC Berkeley, Microsoft Research, and Microsoft Security, turns completed sessions into guidance for subsequent ones.
summary — uncertain: The reading tracks the source's argument, mechanisms and results faithfully. However, central terms are not explained for a general reader. SWE-Bench is not described, although the source calls it a benchmark of real bug fixes in open-source Python projects. Micro-skills, which the source describes as one-line repository notes, are also left unexplained, as are tokens. Smallest repair: add a brief plain-language gloss for SWE-Bench and micro-skills, or confirm these terms have tooltips.
SWE-Bench, a benchmark of real bug fixes in open-source Python projects
date — pass: The originalPublished value of 2026-11-16 comes from the publisher page and matches the feed date. It is consistent with this blog post's first publication. The draft makes no conflicting date claim.
support — pass: Every claim is source-stated and attributed: the 18.4–30.9% savings, no performance loss across the 30 pairs, savings counted net of the plugin's own model calls, higher early costs with cumulative costs below the default by the sixth task, the Web-Bench and security-auditing results, and the reasoning-effort comparison. The question and limitations do not overreach.
Reported savings also account for the cost of TokenEconomizer’s own model calls.
media — pass: The selected media is the verified direct link to the original Microsoft Research blog post. It is the most direct and relevant way to reach the source.
uncertain · 2026-11-16T23:02:34.669Z · anthropic/claude-opus-5.5
Publication: 18190bc0-0d58-4242-8e21-5b713484197d
attribution — pass: The Microsoft Research blog is correctly credited, and results are framed as the researchers' reports. The source calls TokenEconomizer a plugin from UC Berkeley, Microsoft Research and Microsoft Security. 'Microsoft Research presents' is acceptable for the publisher's post, but naming the collaborators would be more complete.
TokenEconomizer, an open-source plugin from UC Berkeley, Microsoft Research, and Microsoft Security
summary — uncertain: The structure and argument are faithful. A general reader is not told that costs come from tokens sent to the model each turn. The coined mechanisms are named but not explained: micro-skills are one-line repository notes, tool pruning drops unused tool descriptions, and failure-loop intervention uses a judge model to detect stuck work. SWE-Bench is also unexplained. Smallest repair: add one plain clause for each mechanism and for tokens/SWE-Bench, or confirm tooltips cover them.
Micro-skills capture repository-specific knowledge in one-line notes.
date — pass: originalPublished (2026-11-16T17:01:35Z) comes from the publisher page and matches the feed entry. No earlier first-publication date appears in the evidence.
support — uncertain: The draft's 'no performance loss in the tested repository-combination pairs' turns an ambiguous sentence into a flat guarantee. That sentence ('saved the most in all 30 pairs without performance loss') may be comparative. Elsewhere the source hedges: success rates 'remained broadly stable' and Pass@1 rose in four of five combinations, leaving the fifth unstated. The draft also drops the 'saved the most' comparison. Smallest repair: say task success remained broadly stable and Pass@1 rose in four of five combinations. Other claims are grounded.
Costs fell in every repository tested, while task success rates remained broadly stable.
media — pass: The selected media is the verified link to the original Microsoft Research blog post, the most direct access to the source.
uncertain · 2026-11-16T23:17:51.056Z · anthropic/claude-opus-5.5
Publication: 35142586-352c-405c-8655-941266c5c46e
attribution — pass: The Microsoft Research blog publishes the work and the draft attributes results to the researchers. The source calls TokenEconomizer a plugin from UC Berkeley, Microsoft Research and Microsoft Security. Naming only Microsoft Research is narrower than the source but not wrong. Optional repair: add the collaborators.
TokenEconomizer, an open-source plugin from UC Berkeley, Microsoft Research, and Microsoft Security, turns completed sessions into guidance for subsequent ones.
summary — uncertain: The draft names micro-skills, tool pruning and failure-loop intervention but never says what they do, and SWE-Bench is not identified as a benchmark of real bug fixes. It also leaves out the source's motivating evidence: agents repeat earlier actions, load unused tools and burn money on failures. Without this, a non-specialist cannot see why the plugin saves money. Smallest repair: one clause per mechanism, plus the profiling finding. For example, 30% of tool calls repeated earlier actions, 16 of 23 tools were never called, and failed tasks took 32% of spending. The title and question are fine.
The agent received 23 tool descriptions on every turn, but 16 were never called.
date — pass: The draft's original publication date, 2026-11-16T17:01:35Z, comes from the publisher page. It matches the feed entry and precedes first observation, and no earlier source publication is indicated.
support — pass: Each claim is source-stated and attributed to the researchers: cost reductions, cumulative savings by the sixth task, Web-Bench and security-auditing results, and the reasoning-effort comparison. The phrase 'no performance loss' matches the source's own wording. The limitations and question stay within the evidence and do not turn reported results into guarantees.
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.
media — pass: The selected media is the verified direct link to the original Microsoft Research blog post, which is the most direct access to the source.
passed · 2026-11-16T23:31:40.404Z · anthropic/claude-opus-5.5
Publication: 5a9a347c-b61b-46bf-8ccb-7cf36aa5a5d5
attribution — pass: The reading correctly credits Microsoft Research as publisher and presents TokenEconomizer as an open-source plugin from UC Berkeley, Microsoft Research and Microsoft Security. Results stay attributed to the researchers ('researchers report') and are not presented as established fact.
TokenEconomizer, an open-source plugin from UC Berkeley, Microsoft Research, and Microsoft Security, turns completed sessions into guidance for subsequent ones.
summary — pass: The title, summary and question faithfully convey the argument: agents that return to familiar repositories waste work, and session experience can cut it. The three named mechanisms are tied to plain descriptions of the waste they target. Coding-agent terms such as SWE-Bench and micro-skills fit tooltip treatment and do not block understanding. The question about generalization follows from the limited evaluation.
can experience in a repository help an agent solve future tasks more efficiently?
date — pass: The displayed originalPublished date, 2026-11-16T17:01:35Z, comes from the publisher page and matches the feed entry. The reading adds no conflicting date.
support — pass: Each figure matches the source: the 18.4–30.9% savings, the six repositories and five agent–model combinations, cumulative savings by the sixth task, the Web-Bench and security-auditing ranges, the reasoning-effort comparison, and the inclusion of the plugin's own calls in costs. The 'no performance loss' wording matches the source sentence, although the at-a-glance summary says success rates 'remained broadly stable'. The limitations do not overstate the results.
saved the most in all 30 repository-combination pairs without performance loss
media — pass: The selected media is the verified link to the original Microsoft Research article, which is the most direct way to reach the source.
passed · 2026-11-16T23:47:11.152Z · anthropic/claude-opus-5.5
Publication: 48bb58df-4e5c-428e-8fe6-7c4a8986cef3
attribution — pass: The reading attributes the plugin to Microsoft Research working with UC Berkeley and Microsoft Security, which matches the source. Results are consistently framed as what the researchers report, not as established fact.
TokenEconomizer, an open-source plugin from UC Berkeley, Microsoft Research, and Microsoft Security, turns completed sessions into guidance for subsequent ones.
summary — pass: The reading leads with the actor and the argument, then gives the mechanism, the results, the early-cost caveat and the reasoning-effort comparison, so the source's force is preserved. Micro-skills, tool pruning and failure-loop intervention are named but not defined, though the preceding waste list conveys their purpose. Smallest optional repair: a tooltip for micro-skills, the source's term for one-line repository notes. The title and question are fair.
Micro-skills capture repository-specific knowledge in one-line notes.
date — pass: originalPublished, 2026-11-16T17:01:35Z, comes from the publisher page and matches the feed entry. No earlier first-publication date appears in the evidence.
support — pass: Every claim matches a verbatim excerpt: the cost reductions, the 277 tasks, no performance loss across the 30 pairs, the sixth-task crossover, the Web-Bench and security-auditing savings, the reasoning-effort comparison, and the inclusion of the plugin's own model calls. The closing inference is limited to the tested evaluations, and the limitations preserve the gap between evidence and conclusion.
Reported savings also account for the cost of TokenEconomizer’s own model calls.
media — pass: The selected media is the verified original Microsoft Research blog article, which gives direct access to the source.
Challenge or correct this reading
Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.
Primary technical research with named authors, papers and implementation details.
Explore Microsoft ResearchPrimary technical research with named authors, papers and implementation details.
Explore Microsoft ResearchPrimary technical research with named authors, papers and implementation details.
Explore Microsoft ResearchPrimary technical research with named authors, papers and implementation details.
Explore Microsoft ResearchPrimary technical research with named authors, papers and implementation details.
Explore Microsoft ResearchPrimary technical research with named authors, papers and implementation details.
Explore Microsoft ResearchPrimary technical research with named authors, papers and implementation details.
Explore Microsoft ResearchPrimary technical research with named authors, papers and implementation details.
Explore Microsoft ResearchPrimary technical research with named authors, papers and implementation details.
Explore Microsoft ResearchPrimary technical research with named authors, papers and implementation details.
Explore Microsoft ResearchPrimary technical research with named authors, papers and implementation details.
Explore Microsoft ResearchPrimary technical research with named authors, papers and implementation details.
Explore Microsoft ResearchPrimary technical research with named authors, papers and implementation details.
Explore Microsoft ResearchA defined set of tests used to compare systems. Results depend on the tasks and measures selected.
Definition referencePrimary technical research with named authors, papers and implementation details.
Explore Microsoft ResearchPrimary technical research with named authors, papers and implementation details.
Explore Microsoft ResearchPrimary technical research with named authors, papers and implementation details.
Explore Microsoft ResearchPrimary technical research with named authors, papers and implementation details.
Explore Microsoft ResearchPrimary technical research with named authors, papers and implementation details.
Explore Microsoft ResearchPrimary technical research with named authors, papers and implementation details.
Explore Microsoft ResearchPrimary technical research with named authors, papers and implementation details.
Explore Microsoft ResearchPrimary technical research with named authors, papers and implementation details.
Explore Microsoft ResearchPrimary technical research with named authors, papers and implementation details.
Explore Microsoft ResearchPrimary technical research with named authors, papers and implementation details.
Explore Microsoft ResearchPrimary technical research with named authors, papers and implementation details.
Explore Microsoft ResearchPrimary technical research with named authors, papers and implementation details.
Explore Microsoft ResearchPrimary technical research with named authors, papers and implementation details.
Explore Microsoft Research