How the Content Engine Routes
One engine, three input types, three different governance paths. The routing decision — not the drafting — is what makes this safe to run automatically.
Every piece of raw input — a paper, a user's own forecast result, a platform metric — is classified by what kind of claim it makes before any content is generated. That classification decides how much human review it needs, if any. This is the part that makes full automation safe: the gate is chosen by the content's risk, not applied uniformly.
"Half-life compression in claims processing" — a working paper from a verified academic contributor
A researcher's paper or finding. This is a new claim about the world — it can move an index, and it carries someone's professional name. Highest stakes.
A subscriber's own forecast result changed. This isn't a new public claim— it's reporting someone their own numbers back to them. Lowest stakes.
A hook post or proof post for social. It's aimed at strangers who have no context yet — needs a glance, not full expert sign-off.
Most content-automation systems apply one review standard to everything, which means either: (a) trivial personal reports get stuck in a review queue that never scales, or (b) expert claims that could move a market skip review because "it's just content." CVM's engine decides the gate from the classification, automatically — so a researcher's paper never bypasses expert review, and a subscriber's own weekly update never waits on a human who has nothing to check.