Routing optimises cost against verified outcomes
Backend selection is evaluated against cost and verified-outcome efficiency, not by default model preference.
Backend selection is evaluated against cost and verified-outcome efficiency. Detailed routing policies, weighting, backend metrics and rebalancing logic are proprietary and reserved for controlled review.
- acceptance vs default
- materially higher
- cost vs default
- materially lower
Open: cost numbers exclude human-review time, which is the dominant cost on shorter task chains.
Public summary & integrity check
Each public case study ships a summary dossier and a SHA-256 digest so you can confirm the bytes served here are the bytes you hash locally. This is a byte-integrity check, not a cryptographic signature. Full reproduction materials (protocol, controls, expected outputs) are access-controlled and released only to approved researchers through the Research Console.
# Recompute the public SHA-256 digest shasum -a 256 economics.reprod.json # expected: … # Or verify against the published manifest shasum -a 256 -c economics.SHA256SUMS
The SHA-256 digest above provides byte-integrity verification of the public summary dossier. Detached Ed25519 signing (over signed research bundles) and external notary co-signing are tracked separately in the Foundation evidence ledger. Full reproduction kits are not published here — request access via the Research Console.