E-008 · economicsongoingrolling 7d
Routing optimises cost against acceptance
Different cognition backends are selected per task by acceptance × cost × chain depth, not by default model preference.
0.41USD / accepted mutation
Method
Each task class has an active routing policy. Acceptance and unit cost are recorded per backend; the router rebalances toward the Pareto frontier.
Supporting metrics
- backends in pool
- 7
- acceptance Δ vs default
- +14%
- cost Δ vs default
- −37%
What is not yet proven
Open: cost numbers exclude human-review time, which is the dominant cost at low chain depth.
Ledger pointer
$ ledger://econ/routing-policy
Shareable URL
https://prodia.dev/evidence/economics