Correction, 2 October 2026. This page was written on 18 August 2026, when the live figures were read from MetaDAO's program on each request and cached at the edge, and a classification covered one attack vector. Since then a collector reads the chain every fifteen minutes into an observation tape, a warehouse seals each pass into the gold layer every page and endpoint reads, windows that closed before the tape began are replayed from the chain's own swap events, the record decodes four engines — MetaDAO and Combinator on Solana, Butter on Unichain and Umia on Base — and the library monitors four vectors. The two sections that describe the pipeline and the vector scope are rewritten below to match. The rules did not change.
Update, 9 October 2026. The record now decodes a fifth engine, futarchy.fi on Gnosis Chain: five engines in all, under the same rules.
Anyone who publishes derived numbers owes readers two things: where each figure came from, and what it does when a figure is missing. This page is both.
It is also, deliberately, the least exciting page on the site, and that is the point. Analytics is only worth building on if the person building on it does not have to re-check it, and re-checking is what a reader does when a number arrives without a lineage. Everywhere else Eidos argues that correctness is the moat; this page is where the claim is made good. If a figure here cannot tell you where it came from and when it was true, it should not be here at all.
Two lineages, kept apart
Everything on this site is one of two kinds of number, and they are never mixed inside a single claim.
Live figures are decoded from the chain. Proposal state, thresholds, current TWAPs and each option's prices are read from confirmed onchain state on the chain each engine runs on — MetaDAO's Futarchy program and Combinator's on Solana, Butter's contracts on Unichain, Umia's on Base. A collector reads them every fifteen minutes and appends each read, with its provenance, to an observation tape; the warehouse seals that pass, and the pages and the API serve the newest sealed read. No page request reads the chain. When a read fails, the last good one stays in place under its own time; it never becomes a stale figure wearing a live label. Windows that closed before the tape began are replayed from the chain's confirmed swap events, and each says it is a replay.
Historical figures are reproduced from publications, with the publication attached. They come from publicly cited 01Resolved work — principally the Ownership Coin Monthly for July 2026 (as of 31 July 2026) and the Umbra treasury-attack forensics of 5 August 2026 — and from the research notes in this repository. NAV/AS values are estimates, as labelled by the compilers who produced them. Spot prices in the July table were attributed to CoinGecko by 01Resolved.
Where those two lineages disagree, the live decode wins for anything current, and the historical figure keeps its date rather than being quietly updated.
Missing evidence is never good news
This is the rule the rest of the site is built to protect: an absent input never becomes a favourable rating.
The tempting failure mode in risk tooling is to treat "no indicators fired" and "we could not look" as the same state, and report both as clean. They are not the same state, and conflating them is worse than publishing nothing, because it converts a gap in coverage into an affirmative assurance.
So the scorer distinguishes them:
| Result | Requires | Meaning |
|---|---|---|
NO-MATCH | Complete evidence | 0–1 indicators matched. A real negative finding. |
WATCH | Complete evidence | 2 indicators matched. |
ELEVATED | Complete evidence | 3 indicators matched. |
HIGH-RISK | Complete evidence | 4 or more indicators matched. |
UNSCORED | — | Required inputs are absent. Nothing is being asserted. |
UNKNOWN | — | Inputs are present but insufficient to conclude. |
STALE | — | The underlying snapshot is too old to stand behind. |
UNDER_REVIEW | — | Held back pending a check. |
The last four are withheld ratings. They are deliberately not a score, and they are never rendered as low risk. Note also what the first four require: a classification is only produced once the evidence packet is complete, so a NO-MATCH is a statement about evidence we have, not evidence we lack.
The distinction sounds pedantic until you notice how much risk tooling gets it wrong. A dashboard that shows a green tick for "no alerts" is quietly asserting that it looked and found nothing, when very often it did not look. The gap between those two states is where the worst surprises live, because a consumer reads the green tick as a promise and acts on it. Splitting NO-MATCH from UNSCORED costs a little — some rows that a laxer tool would paint reassuringly stay conspicuously blank — and that cost is the entire value. A blank that is honest about its blankness cannot be mistaken for an all-clear, and a reader who sees it knows exactly which missing input would change the answer.
What a classification is scoped to
A matched-indicator count is scoped to one economic attack vector. Four are monitored today — recycled-stake treasury extraction (Umbra is its published case), late-window TWAP outcome steering, wash-volume alignment inflation and coordinated multi-wallet stake assembly — and each reports its own count. None of them is a validated composite risk score for a proposal.
A proposal can be NO-MATCH on that vector and still be a bad proposal for reasons Eidos does not model: strategy, execution, counterparty, legal, or simple disagreement about what the treasury is for. The classification answers one question, and the scope belongs in the answer.
This is the opposite of how a risk score usually wants to behave. A single headline number is easy to consume and easy to over-read — it invites a reader to treat "low risk on the one thing we checked" as "low risk," full stop. Naming the vector in the result is a refusal to let that slide. It keeps the claim exactly as large as the evidence supports and no larger, which is less satisfying to quote and far harder to be wrong with. As the library grows to more vectors, each stays its own scoped question rather than being blended into a composite that would hide which check actually fired, and which merely had nothing to say.
Two related fences apply to what counts as a proposal at all. A catalog entry or a treasury snapshot is a profile, not a proposal, and never receives a proposal classification. And a snapshot carries its date, so that when it goes stale the site says stale rather than presenting an old balance as a current one.
Independence
Eidos is an independent research service. It reads publicly cited research, 01Resolved's among it, and the onchain state of the engines it records. It is not affiliated with, sponsored by, or endorsed by 01Resolved, MetaDAO, or any issuer named on this site. Nothing here is investment advice — the full disclaimer is in the footer of every page, and it means what it says.
Where the conclusions differ from the source, the difference is ours and the source is still cited.
Primary sources
The list below is the material this research section is built from. Where a post depends on one of them, it is cited at the foot of that post as well.
01Resolved
- Research index — the running record of the ecosystem
- Ownership Coin Monthly, July 2026 — treasury, volume and revenue series
- The Umbra treasury attack cost its attacker $4,623 — wallet-level forensics
- Ownership Score: what the first experiment reveals — holder-score methodology and results
- Introducing the 01Resolved beta — platform scope
- Building in the open and Crypto needs an ownership overhaul — the thesis posts
- Platform · API documentation
MetaDAO and the ecosystem
- MetaDAO documentation — program mechanics, thresholds, ICO rules
- Shoal — MetaDAO and the new capital formation
- Crypto Briefing — ownership coins · Rip Cars
- Solana Compass — MetaDAO product keynote
- Messari — MetaDAO
Eidos's own working
The published figures on this site are baked into fixtures that carry a citation, and the rules above are enforced by tests rather than by convention — a build that would classify missing evidence as low risk does not ship.