Three products the graph has to earn
All three already exist as code. None of them would have anything to say.
All three read their answers out of the knowledge graph. A report citing one paper per recommendation, an API returning a single edge per query, a hypothesis engine surfacing coincidences — that is what a thin graph gives you. Shipping those early would not be shipping early; it would spend the one thing this project has.
Recorded is everything written to the chain. Verified is what passes the public check — the paper exists, is about aging, and contains both ends of the claim — and it is what every threshold on this page counts. Anchoring covers the full record, checked or not, and runs once a day at 00:30 UTC, so Recorded outruns On Ethereum by whatever was submitted today. That is the schedule. What the schedule does not explain is the gap below, and we would rather point at it than have you find it.
The column that gates everything is Linked by 2+. An entity mentioned by a single claim is a dead end: nothing corroborates it and nothing connects to it. Until that climbs, the graph is a list rather than a network, and a list cannot support any of the three.
That number moves when someone submits a claim that connects to an entity already in the graph. Not a community to join — a stuck number with your name on it.
Install the extension →The figures above are the verified block. The top-level fields count every on-chain record, checked or not.
curl -s https://api.cortexdesci.xyz/api/v1/graph/stats | jq .verified- 5,000 verified claims
- Queries return more than one answer
- Entities shared across many claims
Each bar is the gate in front of the milestone that follows it — what has to become true of the graph before that one opens.
Longevity Pro
A personal report you can follow backwardsA graph thick enough that any set of markers finds evidence
The service needs 100 candidates to pick 24 from, with two independent references behind every line. That is the mechanical floor — enough for one person with one set of markers. Real users arrive with different blood pressure, lipids, inflammation and sleep, and the graph has to answer all of them. We put that at 5,000 verified claims — counted after the public check, not as written to the chain. It is a judgement, not a formula, and we would rather name it than launch on whatever number the code technically tolerates.
Health advice you can audit. Every line shows its work, and the work is a chain you can walk: recommendation → claims → paper → the transaction that recorded it before anyone knew you would ask.
You enter your markers. It pulls the claims that bear on them, keeps those above the confidence and source-grade bar, and builds a report from 20–30 of them.
Consumer longevity advice cannot be checked: a supplement page cites a study that cites a review that cites nothing. Ours can be wrong too. The difference is you can find out where.
A report you can argue with. Disagree with a line and you can go read the exact claim it came from.
Enter your markers and pay. No mining, no minimum balance, no waiting list.
$15 of $CXS per report, converted at the price when you buy. Priced in dollars on purpose — if the token climbs, the report does not quietly become unaffordable.
Half of every payment is burned. Half goes to the treasury that funds the daily mining pool — so readers end up paying the miners whose work they are reading. The 50/50 split is already in the code, not a number we will decide later.
All three open the same way
Not by us announcing a date — by enough people reading papers and submitting what they found, until the entities start overlapping and a claim has something to be checked against. There is no version of this where the products arrive first and the graph catches up afterwards.
No dates appear on this page on purpose. We published a roadmap once and did not deliver it; a second set of promises would not be worth more than the first. These milestones move when the numbers above move, and those come from an endpoint you can query yourself.