Skip to content
OpenSampling

About OpenSampling

OpenSampling is the AI for laboratory measurement evidence — what to measure, how to collect it, where it can be measured, when to measure it, and how it relates to continuous physiological signals.

It is built and operated by Humans Nexus, the infrastructure for remote microsampling and monitoring in Europe: devices and kits, custody, logistics, accredited laboratories, structured results. The evidence and the execution are one system, which is why an answer here can end in a programme there.

OpenSampling is a Humans Model, as defined by Alex Diaconu, the inventor of the Humans Model.

How an answer is built

  1. Retrieval is deterministic. A question resolves to analytes, devices and specimens by exact name, alias or code; the claims the corpus holds about them are retrieved as rows — analyte, device, matrix, value, unit, n, design, condition, page, table — never as passages.
  2. The composer may only cite what was retrieved. It writes the answer from those claims and nothing else.
  3. The verifier outranks the composer. Every citation must name a retrieved claim; every number in the answer must resolve to a cited claim’s own fields; every sentence carrying a number must carry a citation; and no answer may cite one side of a disagreement while leaving the other unsaid. Any failure kills the whole answer.
  4. A refusal is an answer. A question the corpus cannot support ends in a refusal that names the nearest evidence it does hold.

What “open” means

Open means open evidence access and transparent provenance — every number cited to its source, every contradiction shown two-sided, every refusal explained. Not open data, not open source, never unrestricted access to proprietary evidence.

The engines

OpenSampling's evidence engines are our retrieval, our claims corpus, our verifier and our council, built on frontier language models we do not train. We publish their benchmark.

Guthrie
seconds. The cited answer with counts and disagreements.
Tietz
minutes. Feasibility objects, dossiers and living reviews.
Astrup
research preview. Paired signal + sample real-world evidence, for partners.

Neutrality

Evidence is never pay-to-play. A commercial relationship may affect what a customer can publish, keep private or execute; it never affects retrieval, grading, ranking or refusal. Humans Nexus distributes collection devices; that position is disclosed wherever a device is compared and never enters the ranking.

Posture