Turning AI-assisted findings into trustworthy science.

A provenance layer for scientific pipelines — so every finding is traceable back to its source.

The problem

Discovery is accelerating. The infrastructure for trust isn't.

The chain from raw data to discovery is the dark matter of science: it exists, it matters, nobody can see it. AI is producing discoveries at scale, making it even harder to know which ones are real. Mareforma tells you which findings you can trust.

What mareforma does

Three primitives. One traceable pipeline.

@transform

Record every step automatically

Wrap any processing function and mareforma records what ran, from which data, with which code, and when — without changing how you work.

ctx.claim()

Know where every finding came from

Every claim is anchored to the run that produced it — the data, the code, the confidence level. Not just a result, a traceable conclusion.

mareforma status

Know when a finding has nothing behind it

When a step produces no evidence, mareforma flags it. The difference between a real result and a model guess becomes visible before it costs you.