Applied reference implementation

SignalReview: EvidenceBound controls in a production sports-intelligence workflow

SignalReview is a separate product and a production sports-intelligence reference implementation for selected EvidenceBound ideas: visible provenance, deterministic evidence ownership, explicit missing states, bounded AI interpretation and retained review artifacts.

Why it matters

A vertical where evidence cannot be hidden behind model fluency.

Sports data changes quickly, providers have uneven coverage and user-facing AI can sound confident even when critical evidence is absent. SignalReview therefore makes missing, blocked and stale evidence visible instead of allowing agents to silently promote it into fact.

Deterministic evidence owner

Provider identity, freshness, numeric values and unsupported states are owned outside the LLM.

Bounded interpretation

Agents interpret the same evidence registry and cannot overwrite deterministic provider facts or confidence guardrails.

Retained review

Completed research can be saved and reopened as the preserved pre-match state rather than regenerated as if history had not changed.

IP and product boundary

Separate commercial product, reusable trust architecture.

SignalReview may apply EvidenceBound concepts or compatible components. The sports product does not become the canonical identity or research home for EvidenceBound, and its existence does not transfer ownership of EvidenceBound Core, generic verification semantics, recovery semantics or other reusable EvidenceBound intellectual property.

Asset boundary: a future transfer or sale of SignalReview does not transfer ownership of EvidenceBound Core or generic EvidenceBound IP unless a separate written agreement explicitly says so.