The routing engine that understands what you mean, not what you typed
The routing problem isn't typos—it's meaning
Two job posts say “React performance” and “Core Web Vitals for LCP.” String overlap is weak; embedding similarity finds the match humans intend. The Capability Graph engine logs match scores and surfaces ontology gaps so operators improve the graph instead of watching tasks rot in the wrong queue.
Routing is capability-first—cost-informed only after the capability set is honest.
Evidence beats claims
Carlos never filed a formal TypeScript certification—but high-rated tasks across telemetry support an inferred edge a reviewer can promote into a durable capability link. The graph learns from what people produce, not what they typed on a form.
That is revelation, not ranking: confidence intervals and human review where inference proposes—no gaming leaderboards as a product surface.
Cross-org without manual mapping hell
Acme and GlobalCo use different internal labels for similar work. Canonical capability mappings align namespaces so routing finds the right human without every admin hand-wiring synonyms. Think lingua franca, not surveillance dossier—canonical layers exist so labor markets interoperate, not so vendors can sell hidden scores.
Principle check (Canon)
Capability is for revelation, not exclusion. If a feature smells like “secret score,” ship uncertainty and appeals, not opaque tiers.
Pair with
- ML inference risk — same embedding substrate, risk question.
- Evidence economy — narrative spine.
HumanOS policy: Safety at business speed.