What it covers
Four sections, one taxonomy — on every page
Digital Money
- Tokenized Deposit & Deposit Token21
- Stablecoin39
- CBDC25
Digital Assets
- Equities9
- Bond (ex. Treasury)13
- Treasury26
- Commodity3
- Funds (incl. MMF)43
- Real Assets & Infra1
- Structured Products1
- Other (e.g., Art, Trade)6
Crypto
- Custody13
- ETF / ETP9
- Lending3
- Other (e.g., DeFi)2
Other
- Infrastructure & other13
Counts overlap where the taxonomy does: a tokenised treasury product appears under both Treasury and Funds, because funds buy into the same assets the direct products tokenise.
How it works
From a claim to a record
- 1
Discover
A systemd timer sweeps 52 institutional news feeds every night. Aggregator lists and feed hits create leads — never records. A lead is a claim that something happened, not evidence that it did.
- 2
Read the source
A record is published only against the institution's own page — newsroom, filing, or regulator notice — and only after the page's printed date has actually been read. If a publisher blocks automated access and no mirror reads, the record waits. A date we cannot read is a date we do not record.
- 3
Grade the evidence
Every source carries a grade: A regulator, B official announcement, C near-primary, D media. Aggregators are not acceptable at any grade. Each classified value — product type, market, network, maturity — links to the source that proves it as an evidence claim.
- 4
Verify mechanically
28 automated invariants run on every change: counts must agree across pages, no source may predate the event it evidences, dates render only to their known precision, renamed records keep their old URLs. One check is allowed to fail visibly — the honest display of what still lacks a source.
- 5
Publish with the trail
A published record shows its developments, each dated and cited. The tracker never reports market size — no AUM, notionals, or volumes. What institutions did, in what role, and what proves it.
Data sources
Every source is graded
A regulator or central bank speaking about its own action.
The institution's own announcement: newsroom, filing, product page.
Near-primary — a named partner or the network's own publication.
Trade or tier-one media. Never the sole basis for a record.
177 of 272 sources cited by published records (65%) are grade A or B — the institution or its regulator, in their own words.
Human vs AI
Verification is labelled, not implied
1,198
AI-verified claims
The model read the source page — including its printed date — and recorded the claim against it. AI review runs on every change.
0
Human-verified claims
A person confirmed the claim against the source. The maker–checker layer is young: this number is honest, and it will grow.
152 / 201
Records human-reviewed
Records a person has signed off end-to-end. Everything else is AI-built, source-linked, and labelled as exactly that.
The records live behind the login
This page is the method. The data — 201 initiatives, 263 dated developments, every source link — is inside.
Sign in