Confidence: medium. Evidence: benchmark, fast-decaying. Last substantive change: 2026-07.
This subsystem owns which model does which work, and how spend is allocated and bounded. It is where model-specific facts age fastest and where routing principles age slowest.
The conclusion
Route by task and by the evidence a task requires, and do not bind the factory's identity to a single frontier model. Budget is a safety and control variable, not merely a cost cap: a runaway loop is both a spend problem and a blast-radius problem. Model rankings perish quickly; routing principles are durable.
How the thinking got here
One-model agents gave way to model specialization by role, then to dynamic routing, and then to the recognition that the harness and the infrastructure can move outcomes as much as a model upgrade does. A factory tied to one model inherits that model's failure modes and its pricing.
Credible alternatives, and when each is right
| Approach | Right when |
|---|---|
| One general model | simple, uniform work |
| Fixed role-based models | roles have stable, distinct needs |
| Cascades | escalate only when a cheap model fails |
| Competitive ensembles | quality matters more than cost |
| Learned routing | enough history to train a router |
| Local or open models | bounded tasks, privacy or cost constraints |
Where it fails and what we still don't know
Failures include provider-correlated outages, ungraceful degradation when a model changes underneath the factory, and budget overruns that become incidents. Evidence is moderate and decays fast. Open questions include routing on accepted-outcome rather than benchmark score, uncertainty calibration, model retirement, and preventing budget overrun before it happens.
What would change our mind
Durable, provider-independent evidence that a specific routing policy maximizes accepted outcomes per unit cost would turn routing from art into a measured discipline.