Where did this measurement actually come from?

Every scientific and industrial data pipeline eventually writes down where its data came from. Almost always in a field like meta.source — a string, inside the file, that the file asserts about itself. It can be edited afterwards with no trace, carries no independent timestamp, cannot be checked by anyone outside the project, and is not bound to the contents it describes. It is a claim about provenance, not evidence of it.

This proof-of-concept runs three instrument channels, each owning its own Zenon account chain. Every sixty seconds a channel produces a data frame and publishes a 56-byte commitment: the SHA3-256 digest of that frame and its declared domain tag, bound together in one block signed by the instrument's own key and timestamped by momentum height.

The frame itself never goes on chain. Only the proof of it does. Change so much as the source label afterwards and the digest no longer matches — use the verifier at the bottom of this page to watch that happen.

Keys, frame hashing, block construction, and Ed25519 signing all happen in this page. Chain reads and the final submission go through a public mainnet node operated by the hypercore.one community (my.hc1node.com:35997). No fees, no backend, no wallet needed.

A worked example: solar data standing in for tokamak data.

Fusion control systems need magnetic fluctuation data long before a reactor exists to produce it, so real solar magnetogram data gets used to exercise the signal-processing layer. It is a sound engineering practice, and it carries one specific hazard: solar-derived results being read later as tokamak-relevant. The two domains differ by seven to nine orders of magnitude in Alfvén time and roughly three in field strength, so a number that means one thing in the first means nothing in the second.

The usual mitigation is to record the data source in each frame's metadata. That is precisely the claim-not-evidence problem above. Committing the digest and the domain tag together, under the instrument's own key, makes the relabelling detectable by anyone — a reviewer, a funder, a partner lab — without trusting the project that produced the frame.

To be explicit about scope: this is an out-of-band audit record, not a control path. Nothing here belongs in a real-time control loop, and nothing about it needs to be. It is the validation record that outlives the run.

Cheap gas is not the same as feeless.

A 64-channel B-dot array committing once per second is 5.53 million transactions per day. At a tenth of one cent each — cheaper than any production chain ships today — that is $5,530 per day, roughly $2.0M per year, just to keep a provenance record.

Which is exactly why per-frame attestation gets dropped: someone reasons that committing every frame is unaffordable, so the pipeline commits a batch, or a daily summary, or nothing. Provenance is not usually lost to malice. It is lost to a line item.

Daily operating cost for Zenon Plasma: $0