
Capability
C³ Assurance Modes
Three cost/rigor envelopes: light, verified, audit
C3 Assurance Modes provide three distinct processing tiers that allow users and applications to trade off speed and cost against evidence rigour depending on the stakes of the question being asked. This tiered approach recognises that not every query requires the same level of verification, and forcing maximum rigour on every request would make the platform prohibitively expensive and slow for everyday use.
Light mode is designed for fast, exploratory lookups where speed matters more than exhaustive evidence chains. When a user wants a quick answer to check their understanding, explore a topic they are unfamiliar with, or get a directional sense of what the evidence says, light mode delivers answers in seconds with basic source attribution. Light mode uses the fast embedding lane for retrieval, applies a simplified ranking strategy, and generates lightweight receipts that record the retrieval decision without full synthesis rationale. It is the appropriate choice for early-stage research, brainstorming, and general knowledge queries.
Verified mode is the workhorse tier for production answers that will be shared, cited, or used in professional decision-making. Verified mode runs the full hybrid retrieval pipeline with standard or premium embedding lanes, applies multi-source corroboration to ensure claims are supported by independent evidence, and generates complete CROWN receipt chains with full synthesis rationale. The resulting answers carry confidence bands and coverage labels that indicate the quality of the underlying evidence. Verified mode balances rigour with reasonable cost and latency, making it suitable for the majority of professional use cases.
Audit mode provides the highest level of evidence rigour available on the platform. In addition to everything that Verified mode provides, Audit mode runs deterministic replay to verify that the same query would produce the same answer across multiple executions. It applies stress testing to measure answer fragility and identify the minimum evidence set supporting each claim. It performs multi-source corroboration with stricter diversity requirements, and it generates compliance-grade evidence packs that can be exported for regulatory submissions or legal proceedings. Audit mode is more expensive and slower than the other tiers, but it produces answers with the level of documentation and verification that regulated industries require.
The assurance mode is specified per query, either explicitly by the user in the Ask Surface or programmatically through the API. ApiCrux enforces the credit cost associated with each mode, and the mode is recorded in the CROWN receipt to ensure that anyone reviewing the answer knows what level of verification was applied. This prevents misuse where an answer generated in Light mode might be mistaken for a fully audited conclusion.
For the CueCrux platform, C3 Assurance Modes make transparency scalable. They allow the platform to serve both casual exploration and regulatory-grade verification from the same infrastructure, with clear, honest communication about the rigour applied to each answer.





