SCIENTIFIC STATUS
A public experiment, not a discovery engine.
lacuna keeps sourced human-curated questions separate from computed measurements. Metric versions 1 and 2 failed their pre-registered benchmark; the active replacement is a no-human prospective protocol with sealed 2026 predictions, awaiting its registered three-release outcome window.
REPLICATED EMPIRICAL EVIDENCE · V1
Measured twice. Claimed narrowly.
The same frozen association gates passed for 1,599 of 499,500 presealed gene pairs in two pinned breast-tumour expression cohorts. The strongest 100 observations are published as cohort-level rank-expression associations only—not as causality, mechanism, clinical utility, novelty to humanity, or validated knowledge gaps.
Download versioned evidence JSONTHE EVIDENCE LAB
Interrogate the failed method.
Search, filter, export, and trace published pairs back to their OpenAlex queries. They are outputs from a method that failed validation, not candidate discoveries.
Read the validation reportKNOWN LIMITS
The map also publishes what it cannot see.
OpenAlex is thin across parts of the humanities and historical record, and does not represent much craft, practitioner, indigenous, oral, or other non-academic knowledge.