Hypothesis
Beyond a team size that depends on the task’s dependency structure, adding agents increases coordination overhead faster than throughput, and an orchestrated topology moves that point later than peer-to-peer messaging.
Proposed methodology
- Vary team size and topology (orchestrated vs. peer) on tasks with controlled dependency structure.
- Measure completion time, cost and message volume.
- Fit and compare coordination-cost models (linear, Brooks-style pairwise).
Required data
- Message and delegation graphs per run.
- Task dependency graphs.
- Per-run cost and completion outcomes.
Evaluation metrics
- Speed-up vs. team size.
- Coordination share of total compute.
- Out-of-sample error of the fitted cost model.
Current status
Planned. No data has been collected and no experiment has been registered for this track.
Known uncertainties
- Results may not transfer between agent frameworks.
- Synthetic benchmark tasks may not represent real projects.
Source references
- Brooks, F. P. (1975). The Mythical Man-Month. Addison-Wesley.
- Amdahl, G. M. (1967). Validity of the single processor approach to achieving large scale computing capabilities. AFIPS Conference Proceedings 30.