Wu, Zhiyao, Liang, Zi, Hu, Haibo · Nov 14, 2026
arXiv cs.AI: When the Mythical Man-Month Meets Agentic Web Systems
Markdown · JSON · Original source
Authors Zhiyao Wu, Zi Liang and Haibo Hu argue that adding agents to complex Web tasks does not necessarily improve performance: communication, synchronization, redundant work and coordination overhead can offset the benefits of greater parallelism and task coverage. Their experiments on complex Web-oriented software development tasks find a non-linear scaling pattern: per-agent token consumption initially falls as the agent population grows, then shows diminishing returns. In response, they introduce the Surgical Team Model (STM), which organizes agents through role specialization, constrained inter-agent communication and coordinator-mediated task aggregation. On full-stack Web application development tasks, the authors report that STM achieves competitive task completion quality and comparable token consumption relative to representative baselines, while incurring higher execution . The paper therefore frames agent coordination as a tradeoff, not a case where simply increasing the number of agents guarantees better results. Its abstract does not give quantitative results or details of the representative baselines, so the strength and scope of these comparisons cannot be assessed from this evidence alone.
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
cs.AI · Article
Source details
Original source date: Nov 14, 2026 (2026-11-14).
Date basis: Date basis not provided.
Added to Escape Meridian: 2026-11-17T13:34:46.800Z. First observed: 2026-11-17T08:28:29.280Z. Retrieved: 2026-11-17T13:33:52.393Z.
Referenced by
Recorded citations in public readings. A reference alone does not establish agreement or influence.
No public references recorded here yet.
Limits
The abstract provides no quantitative results, so the size of the reported differences in task completion quality, token consumption and execution is unclear. The abstract does not identify the representative baselines or provide details of the full-stack Web application development tasks, limiting assessment of how broadly the comparisons apply.
Open question
How much higher was STM’s execution than the representative baselines on the full-stack Web application development tasks?
Source-stated
The authors say that adding agents can bring parallelism and task coverage, but coordination costs may offset those benefits.
increasing the number of agents does not necessarily improve system performance, as communication, synchronization, redundant work, and coordination overhead may offset the benefits
Source-stated
In the authors’ experiments, per-agent token consumption initially decreased as the agent population grew, then showed diminishing returns.
per-agent token consumption initially decreases as the agent population grows, but eventually exhibits diminishing returns.
Source-stated
The authors introduce STM, a framework using role specialization, constrained inter-agent communication and coordinator-mediated task aggregation.
a structured coordination framework that organizes agents through role specialization, constrained inter-agent communication, and coordinator-mediated task aggregation
Source-stated
For full-stack Web application development tasks, the authors report competitive task completion quality and comparable token consumption against representative baselines, with higher execution latency.
STM achieves competitive task completion quality and comparable token consumption relative to representative baselines, while incurring higher execution latency.
Source and contributor context
Wu, Zhiyao, Liang, Zi, Hu, Haibo · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
No suitable media was found by the supported page extraction.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Agentic AI
AI systems that plan and carry out a series of steps toward a goal, often by using other software, rather than giving one answer.
Latency
The delay between sending a request and receiving the result.
arXiv
arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Editorial provenance
Version 1. Publication ID: 6d258c7a-7ae8-4eb7-8fd7-638a87f81562.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-17T13:34:46.800Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: 3e43a2226af9f5425cc81e645e81b7d07b38cdca96c6a5144a75545226a62ac3. Source text hash: 75937033d8a5a265b3559666a725053db24490ed666fa34c42de3ddb7851d576.
Revision and source-check history
Version 1 · 2026-11-17T13:34:46.800Z
arXiv cs.AI: When the Mythical Man-Month Meets Agentic Web Systems
Authors Zhiyao Wu, Zi Liang and Haibo Hu argue that adding agents to complex Web tasks does not necessarily improve performance: communication, synchronization, redundant work and coordination overhead can offset the benefits of greater parallelism and task coverage. Their experiments on complex Web-oriented software development tasks find a non-linear scaling pattern: per-agent token consumption initially falls as the agent population grows, then shows diminishing returns. In response, they introduce the Surgical Team Model (STM), which organizes agents through role specialization, constrained inter-agent communication and coordinator-mediated task aggregation. On full-stack Web application development tasks, the authors report that STM achieves competitive task completion quality and comparable token consumption relative to representative baselines, while incurring higher execution . The paper therefore frames agent coordination as a tradeoff, not a case where simply increasing the number of agents guarantees better results. Its abstract does not give quantitative results or details of the representative baselines, so the strength and scope of these comparisons cannot be assessed from this evidence alone.
Original source
cs.AI · Article
Source details
Original source date: Nov 14, 2026 (2026-11-14).
Date basis: Date basis not provided.
Added to Escape Meridian: 2026-11-17T13:34:46.800Z. First observed: 2026-11-17T08:28:29.280Z. Retrieved: 2026-11-17T13:33:52.393Z.
Limits
The abstract provides no quantitative results, so the size of the reported differences in task completion quality, token consumption and execution is unclear. The abstract does not identify the representative baselines or provide details of the full-stack Web application development tasks, limiting assessment of how broadly the comparisons apply.
Open question
How much higher was STM’s execution than the representative baselines on the full-stack Web application development tasks?
Source-stated
The authors say that adding agents can bring parallelism and task coverage, but coordination costs may offset those benefits.
increasing the number of agents does not necessarily improve system performance, as communication, synchronization, redundant work, and coordination overhead may offset the benefits
Source-stated
In the authors’ experiments, per-agent token consumption initially decreased as the agent population grew, then showed diminishing returns.
per-agent token consumption initially decreases as the agent population grows, but eventually exhibits diminishing returns.
Source-stated
The authors introduce STM, a framework using role specialization, constrained inter-agent communication and coordinator-mediated task aggregation.
a structured coordination framework that organizes agents through role specialization, constrained inter-agent communication, and coordinator-mediated task aggregation
Source-stated
For full-stack Web application development tasks, the authors report competitive task completion quality and comparable token consumption against representative baselines, with higher execution latency.
STM achieves competitive task completion quality and comparable token consumption relative to representative baselines, while incurring higher execution latency.
Source and contributor context
Wu, Zhiyao, Liang, Zi, Hu, Haibo · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
No suitable media was found by the supported page extraction.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Agentic AI
AI systems that plan and carry out a series of steps toward a goal, often by using other software, rather than giving one answer.
Latency
The delay between sending a request and receiving the result.
arXiv
arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Editorial provenance
Version 1. Publication ID: 6d258c7a-7ae8-4eb7-8fd7-638a87f81562.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-17T13:34:46.800Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: 3e43a2226af9f5425cc81e645e81b7d07b38cdca96c6a5144a75545226a62ac3. Source text hash: 75937033d8a5a265b3559666a725053db24490ed666fa34c42de3ddb7851d576.
passed · 2026-11-17T13:34:46.800Z · anthropic/claude-opus-5.5
Publication: 6d258c7a-7ae8-4eb7-8fd7-638a87f81562
attribution — pass: The draft correctly credits the paper to Zhiyao Wu, Zi Liang and Haibo Hu, as listed on the arXiv page. It keeps the findings attributed to the authors with phrasing like 'the authors report'. The 'arXiv cs.AI' label matches the feed publisher; the paper's primary category is cs.SE, which is a minor point.
Authors:Zhiyao Wu, Zi Liang, Haibo Hu
summary — pass: The summary leads with the authors' argument, the scaling pattern, the STM framework and the reported tradeoff, including higher latency. It preserves the abstract's limits and caveats. The question asks about an unresolved latency magnitude the abstract leaves open. Terms such as agents, token consumption and the Mythical Man-Month allusion are left to term-library tooltips. That is acceptable, though a brief plain-language note on tokens as units of model text processing (roughly, cost) would aid general readers.
how can agent populations be organized to balance task performance and coordination costs?
date — pass: The originalPublished date of 2026-11-14 matches the v1 submission date on the arXiv page. The feed date of Nov 17 is later and was correctly not used.
[Submitted on 14 Nov 2026]
support — pass: Every claim tracks the abstract nearly verbatim, and each supporting excerpt is exact. The tradeoff framing is supported by the abstract's 'balance task performance and coordination costs.' The draft adds no unsupported benefits or costs. Its limitations correctly note that the abstract gives no quantitative results and does not identify the baselines.
STM achieves competitive task completion quality and comparable token consumption relative to representative baselines, while incurring higher execution latency.
media — pass: The selected media is the verified direct arXiv abstract page for this paper, which is the most direct way to reach 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.
The delay between sending a request and receiving the result.
Definition referencearXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXivThe delay between sending a request and receiving the result.
Definition referenceThe delay between sending a request and receiving the result.
Definition referenceThe delay between sending a request and receiving the result.
Definition referencearXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXivThe delay between sending a request and receiving the result.
Definition referenceThe delay between sending a request and receiving the result.
Definition reference