TomSyn

Methodology

A living title-scored shelf plus registry rows. Scope is schizophrenia publications from 1900 to now. This copy is not that literature. Not a systematic review. Not RoB2. Not a cure. Consensus is not a source. The 718-row census is one function-flavored CT.gov slice. Duration is a column, not a lock named YEAR.

Rules

  • Atom

    One discrete cited fact. Optional quote span when a paper matches. No cite, no new atom. Blank is unknown, not zero.

  • Discrimination

    A paper is not a trial. Recruiting is not a result. Five-week PANSS is not a year. C4A genetics is not enroll-then-randomize.

  • Retraction

    We do not yet poll Crossref retraction. Until we do, do not treat this shelf as cleaned of expressions of concern.

  • Cross-condition

    QTc crossovers and MDD-only adjuncts leave the schizophrenia clean set even if they have an NCT.

Tests this vault must fail

  • YEAR

    Fail unless a completed, function-primary, non-OLE human RCT is ≥48 weeks in this file.

  • C4A-enroll

    Fail unless eligibility or an arm enrolls by C4A (or equivalent) and then randomizes.

  • Paired harm

    Fail unless benefit and injury are scored on the same NCT.

  • Clean set

    Human + DOI/NCT, minus recruiting, QTc, MDD-only, off-topic.

  • Never shrink

    A 232-paper vault cannot become 40. Client storage must not overwrite the server shelf.

Mouths

  • Europe PMC — titles, abstracts, OA snippets — never Consensus as a cite
  • ClinicalTrials.gov v2 — NCT as the unit; census denominator lives here
  • OpenAlex — works search

lock 95 · since 1900 · pulse 1.5h · fetch 6 · practice 3.75m

Packet spine

  • holes.md — three sentences
  • generated.md — PICO and what we will not say
  • ncts.md / trials.json — NCT as the unit
  • clean-set.csv — start a table here
  • params.json — 95 is a lock
  • nulls.md — every claim has an H0

Home · White paper · Legal