Candidates declare hypotheses, predictions, distinguishing outcomes, dependencies, verifier, utility and cost. The ranker discounts overlap with experiments actually covered by verified receipts. Independent replication has a separate path.
Epistemic meristem
Grow an experiment where a distinction is missing.
When several workers propose different fixes but share the same untested assumption, GenOS needs a new experiment rather than another role.
Deterministic ranking, scoped coverage storage and optional Rhizome growth planning are connected. A proposed assignment alone does not occupy a niche.
Autonomous worker recruitment, full Trinity contract adaptation and empirical calibration remain open.
If three fixes assume that an event is delivered only once, a controlled duplicate delivery can test the shared assumption before allocating another fixer.
Compare against role recruitment and text diversity on hidden-cause incidents, with equal models and budgets. Measure discovered causes, shared errors and cost per useful distinction.
The similarity score uses declared fields; it is not proven semantic diversity. Ranking cannot authorize a worker spawn.
This status describes the slice documented at the reviewed commit. These links open runtime sources, not a live demonstration.