{"path":"research/grounding/claim.reading.md","content":"What this project has ruled that bears on what one claim rests on, what pulls against it and what it bears on, said in a few sentences, in plain words, so you know how careful the question is before you read the item:\n- Propose, never assert: every machine judgment on contested ground (terminus classification, stance conflicts, daemon output, merges) is a proposal a human ratifies.\n- Every new pipeline pass answers \"does this flatten contestation?\" as a design gate: five known flattening mechanisms, each blind to the others (threat model).\n- No verdict without its object: a machine pass answers by producing the object its verdict depends on, and a verdict that comes with no object is refused.\n- A link's effect has two factors: the supporting claim's own strength and the link's relevance. The founder's words: \"About the decision: your proposal is blessed, make it so.\".\n- A hedge is stance, not strength: an author's hedge on a step or a claim (may, probably, I'd guess) is their standing toward it: stored as telemetry on the relevance claim or the claim, shown to people who want it, never a factor in strength and never kept in the wording of a minted conclusion, which carries the definitive form (will), since how strongly the step bears is what the graph is to find out.\n"}