How Prism Works — Data Pipeline | Prism
ARCHITECTURE

Seven sources in. Nine layers out.

Prism sits between the chain and the person reading it. Everything below is a real pipeline stage — drag any node to trace how a claim on a Prism is derived from the event that produced it.

DRAG NODES · SCROLL TO ZOOM · SOURCES → REFRACTION CORE → NINE LAYERS
THE PIPELINE

What happens between a block and a sentence.

{{ s.n }}
{{ s.t }}

{{ s.d }}

{{ s.lat }}
EVIDENCE RULE

No source, no sentence.

The summariser can only emit a clause if it can attach a resolvable reference: a block, a print, an attestation or a filing. Unsupported clauses are dropped before rendering rather than hedged in language.

KEEP“Net exchange outflows hit a 5-month high.”
DROP“Sentiment appears broadly constructive.”
HOW EACH LAYER IS COMPUTED
{{ l.k }} {{ l.v }}

See what the pipeline produces.

Open a Prism Read the API