The gate

The bar

The bar is expected_harm × scope_breadth, not polarity. A broad positive claim ("X always works") needs replicated evidence; a narrow negative claim needs provenance so the field doesn't fill with "didn't work on my machine" sludge (search collapse). Open questions are always cheap to record.

— _engine/gate.py

It is not conflict detection

The gate fires at write time and only decides whether the deposit is allowed into the store. Conflict flagging happens later, at read time, in _engine/aware.py. The two mechanisms are independent.

Scope breadth

def _scope_breadth(scope: Scope) -> Literal["narrow", "broad"]:
    facets = sum(1 for f in (
        scope.model, scope.dataset, scope.env,
        scope.version, scope.n, scope.seed,
    ) if f is not None)
    return "narrow" if facets >= 2 else "broad"

The rules table

• open→ always accept (it's a question, not a claim).
• positive + broad → requires replicated or verified.
• positive + narrow → requires observed or stronger.
• negative + broad → requires observed or stronger.
• negative + narrow → requires provenance: at least one artifact_refs entry OR a scoped env or version.
• cautionary → requires provenance.

Author-role modifier

A deposit written with a redteam author_role gets a +1 evidence-grade bump on negative and cautionary claims only — positive and open deposits are unaffected.

The role string is normalised before matching: all non-alphanumeric characters are stripped and the result is lowercased. So "redteam", "red_team", "red-team", and "RedTeam" all match.

When a deposit is rejected

from atlaso import DepositRejectedError

try:
    user.add(
        "threshold 0.7 is always optimal",
        polarity="positive",
        evidence_grade="anecdotal",
    )
except DepositRejectedError as e:
    print(e.gate_reason)
    # "positive/broad claim requires evidence_grade>=replicated;
    #  upgrade evidence or narrow scope (e.g., add model= or dataset=)."

The error carries a gate_reason string that names the rule that fired and tells you the next action (upgrade evidence, narrow scope, add provenance).

Why this exists

A memory store filled with under-evidenced positives looks confident even when it isn't. A store flooded with single-machine negatives is useless. The gate keeps both kinds of sludge out at write time so the dispersion math at read time remains meaningful.