Skip to main content

The record

6 gaps in the registerhourly revalidated

Every figure on this page is computed from the database. Figures revalidate hourly.

A number here describes what the record holds, not what is true of the world. Where a field is empty, this page names the emptiness rather than leaving the omission to be read as an all-clear.

What the record holds

published substances
295
approved interaction pairs
663
approved class-level rules
2144
citation checks run to date
27300
checks carrying a human override
0

Counts here are rows in the record; surfaces that count only rules resolvable to published substances may show smaller numbers.

How those checks came back. A citation check is a model reading a claim against the abstract of the source it cites; the five results are published as they stand, in the order the evidence pages use.

15560verified
1487partial
9939inconclusive
1unavailable
313not supported

The gap register

Each gap has a permanent id and a permanent anchor. A reader can cite one, and check later whether it closed. Ids are never reused or renumbered, including after a gap closes.

GAP-0001

Taper information

Of 295 published substances with a dependence profile, 295 carry no taper information — no recommendation, no protocol, no notes.

what it is
Whether a substance profiled for dependence also carries taper information — a recommendation, a protocol, or notes.
why it matters
A dependence profile without taper information describes how dependence forms but not how it is unwound, which is the question asked at the point of stopping. The blank means the record holds nothing here; it is not a finding that no taper is needed.
what closes it
A cited taper protocol entered against each dependence profile, or a cited statement that the evidence does not support one — either way, reviewed before it publishes.
GAP-0002

Interaction severity

Of 663 approved interaction pairs, 320 carry no severity grading.

208lethal
135dangerous
0caution
0low risk
320ungraded
what it is
Approved interaction pairs whose severity is recorded as ungraded rather than as a tier.
why it matters
An ungraded pair is an open question in the record. A combination checker that returns one is reporting that nothing has graded it, not that the combination has been examined and found mild.
what closes it
A cited severity grading per pair, approved through editorial review — starting with the pairs readers combine most.
GAP-0003

Human verification

Of 27300 citation checks run to date, 0 carry a human override.

what it is
Citation checks whose model-issued verdict has since been reviewed and recorded by a human editor.
why it matters
A verdict without an override is a model's reading of an abstract and is labelled as one wherever it appears. It stands until an editor reaches it, which means an unreviewed check carries less weight than a reviewed one rather than the same weight.
what closes it
Editor passes over the verification ledger, taking safety-critical claims and every not-supported verdict first.
GAP-0005

Detection window coverage

Of 295 published substances, 0 carry at least one detection-window entry.

what it is
Published substances carrying at least one detection-window entry.
why it matters
Detection windows are read under time pressure, before a test or after an exposure. A substance with no entry leaves that question unanswered here; it does not mean the substance goes undetected.
what closes it
Cited windows per matrix — urine, blood, hair — entered per substance against its current version.
GAP-0006

Emergency profile coverage

Of 295 published substances, 273 carry an emergency profile with content in it.

what it is
Published substances whose current version carries an emergency profile with content in it — overdose signs, first response, a reversal agent, or resolution notes. Placeholder rows do not count.
why it matters
The emergency profile is what a page can offer once something has already gone wrong. Where it is empty, the page can only point at the general emergency guide and names nothing specific to the substance. The blank is unwritten content, not a finding that the substance carries no overdose risk.
what closes it
Cited overdose presentation and first-response steps per substance, with a reversal agent recorded only where one exists.

These are corpus-wide figures. Per-substance provenance — which source stands behind each value, and what a citation check found — lives on the /evidence page for that substance. How values are sourced, graded, reviewed, and corrected is documented in the methodology.