How this archive works.

A record is one dated prediction about where a B2B market is heading: a thesis, what follows if it holds, a due date, a binding rule for how it resolves, four to six early signals that should appear along the way, and a few tracking phrases whose coverage decides whether the subject went mainstream.

All of it is fixed on the day the record is sealed. Nothing in it is edited afterwards.

A prediction that is right is also valuable: it names a space before the market does. Publishing it in full on day one would hand that space to everyone it is meant to anticipate — including the competitors of the companies these predictions are made for.

So each record is sealed. Its exact text is written to a file, the file's fingerprint (SHA-256) is published, and the fingerprint is anchored in the Bitcoin blockchain through OpenTimestamps. From that moment the text cannot be changed without the fingerprint changing too. The public sees that the record exists, its market, its sealed and due dates and how many of its early signals have appeared — but not what it says.

Records are not hand-picked. Every week, one record per market is sealed automatically from the most recent analyses the plinio TL signal engine has produced for that market.

A candidate qualifies only if its analysis is complete, it has at least three concrete early signals and its tracking phrases, the assessed opportunity clearly exceeds today's coverage, and every attributed claim in its text is backed by a source. Among qualifying candidates, the one an independent review rates best on grounding and foresight is sealed.

A record is never sealed if it resembles a topic plinio is working on for a client.

Each record carries four to six early signals: concrete, dated events that should happen if the prediction is right — a regulation advancing, a filing that names the issue, a tender that asks for it, research that starts measuring it.

They are checked every day against official registers (EU legislation, the US Federal Register, US company filings, EU public tenders), research databases and news. A signal only counts if it happened after the record was sealed and a quoted source confirms it. The count of appeared signals is public from day one; which signals they are becomes public at reveal.

Every record is decided by the same binding rule, written into it at sealing:

Verified if, by the due date, the subject has gone mainstream: in at least two of three independent counts — news coverage, research papers, company filings — volume at least tripled against the three months before the record was sealed. Partially verified if it has not, but at least half of the sealed early signals appeared by the due date. Falsified otherwise.

The counts are taken for the tracking phrases fixed at sealing. A subject can only count as mainstream once the record is at least 90 days old, so a burst of attention in its first weeks decides nothing.

The due date is set at sealing from the prediction's own time window, never less than 90 days out.

A record is revealed in one of two ways: early, the day its subject goes mainstream; or on its due date, whatever happened. Either way the full text is published — thesis, early signals and their outcome, tracking phrases, sources and the verdict.

The track record counts revealed records only. A sealed record is neither a hit nor a miss yet.

Records come from accounts plinio runs for each market, including the operator's own agency. Before a record is sealed, every company name, product name and web domain in it is replaced by a neutral description, such as "a European B2B consultancy". This happens before sealing, so the revealed text is exactly the sealed text.

Every revealed record links its sealed file and its timestamp proof. Download both and run:

sha256sum <id>.json
ots verify -f <id>.json <id>.ots

The first command must print the fingerprint shown on the record page — the same fingerprint that was published on its sealed date. The second confirms, from the Bitcoin blockchain, that the file existed no later than the block it was anchored in. You need nothing from us for either step; the OpenTimestamps client is free and open source.

Sealed records are never deleted, reworded, re-dated or re-sealed. A record whose text did not match its fingerprint would fail the check above for anyone who tries it — which is the point.