What we sell
The signature is the product
Every figure here comes with a receipt anyone can check, and checking it does not involve us.
Nobody holds the signing key. Not even us.
It is split across 34 independent nodes on a decentralized network (Internet Computer). No single node holds the whole key and nothing assembles it. A signature happens only when enough of those nodes agree on that exact set of bytes.
That is what separates these receipts from a screenshot, a PDF statement, a CSV export or an auditor's letter. Each of those is a claim by whoever handed it to you. A signature is an event 34 independent nodes took part in, and you can verify it without asking us anything.
Three more things sit outside our hands. The key is tied to our identity, so an imitation of our service gets a different key and your check fails. The timestamp comes from those same nodes before our software sees it, so a reading cannot be backdated. And the receipt names the software that produced it, supplied by the platform rather than by us, so you can rebuild that software and confirm it matches.
A signed reading is never re-signed.
It is signed the moment it is taken. Where other providers quietly correct last month's figures, our record stands as it was taken, which is exactly what makes it worth showing to someone who has to defend it later.
Then you check it yourself.
Every figure ships with a small program. Your auditor runs it, it re-reads the blockchain, redoes the arithmetic from the signed readings, verifies the signature, and prints PASS, FAIL, or could not check. It takes every input from the receipt itself. For anyone who would rather not run our code at all, a second verifier shares none of it and needs only a standard Python install.
One limit worth knowing up front: a signature proves each reading is genuine, not that we showed you every reading. The feed publishes its own gaps so you can see the coverage you are getting.
That is the whole product. The rest of this page is what we point it at.
1.A daily record of what a rate actually was
Every day at a fixed time we read a protocol's own rate straight from its contract and sign it. Not our estimate of the rate. The protocol's own number, dated.
A web address per series returning the whole history as JSON or spreadsheet ready CSV, every row signed individually, with any gaps in coverage listed rather than smoothed over.
Every commercial data vendor restates history, and free sources are unsigned and undated at source. We cannot revise: a row is signed the moment it is measured or it does not exist. If you have to defend a figure to a regulator or an investor months later, that is the property that decides whether you can.
An index provider or benchmark administrator builds a published rate from inputs it has to defend in a methodology review. A research desk backtests against a history nobody can accuse them of curating.
2.What a position actually earned
A fund holds a position in a DeFi protocol. At month or quarter end it has to report what that position earned. Today that means a dashboard screenshot or a block explorer export, and nobody can check either one. We sell a signed receipt for the return between two dates that a stranger can verify without asking us anything.
A link, one per position per period, showing the return, the exact dates it covers, and the two readings it was calculated from. Plus a file for the audit folder containing those readings, each signed, the arithmetic, and the verifier program.
At month end the accountant opens the link and takes the figure. It goes into the NAV or the factsheet, the file goes into the audit file, and the auditor runs the program and gets PASS. What used to be we believe the manager becomes we ran the check.
3.A wallet's holdings over time
A dated series of what a wallet held, each entry signed, with one fingerprint covering the whole series so nothing can be quietly inserted or removed later.
This is the raw input to a net asset value calculation. The accountant needs holdings at each valuation date and today rebuilds that by hand from explorer exports. Here it arrives already dated and already signed.
Quantities, not valuations, so your own pricing policy stays yours. You value the position the way you always have; we settle what was held.
4.What a liquidity pool position really earned
For a pool position, the answer to a question a single blended rate destroys: how much did I earn, as opposed to how much did the prices move?
Three numbers, reported separately: trading fees earned, and the change in each of the two tokens held.
Report the fee income as income and the token movements as market movement, which is what accounting rules require.
Because a single blended figure would need a price feed, and a price feed is one more thing you would have to trust. Three separate numbers keep the evidence to what the chain itself records.
5.What a wallet held at a past moment
Proof that a specific wallet held a specific amount of a specific token at a specific past moment, read through three independent providers that had to agree, pinned to one block, and signed.
This is the building block under the two series above rather than a product on its own, and it is worth saying so. A balance at a block is public and stays public, so what we add is the reading taken through providers that had to agree, the block pinned, and a program that re-derives it for your auditor. Most of its value shows up as the input to a dated series.
Attach it to a year end pack, an investor report, or a due diligence answer. The recipient runs the program instead of believing a screenshot.
6.What the market expected, before anyone knew
Not our prediction. The market's own price for a future yield, recorded on the day it was quoted. Certain markets price a yield all the way to a fixed maturity date, which is a number embedding what traders actually believe. We read five of those markets once a day and sign what we read.
Anybody can say afterwards that they called it. A prediction only counts as evidence if it was written down before the answer was known, by something incapable of editing it later. A signature at the moment of reading is exactly that, and it is the one thing here that cannot be reconstructed from the chain afterwards by anyone.
A research desk or a manager publishes a rate view. When the maturity date arrives, the signed record shows what the market said and when, so a track record can be audited rather than asserted. It works the other way too: if a counterparty told you in August that a given return was achievable, the August reading is on the record.
The record starts when the measuring starts, and ours began in September 2026. Nothing has reached maturity yet, so there is no scored track record to show you. That is a matter of calendar time and the only way to fix it is to have started, which is why it is running now.
Why any of this is credible
Anyone can rebuild our software from source and confirm it is byte for byte the code that produced their receipt. The controller list and the full history of deployed versions are published too, so the software itself is as checkable as the figures it signs.
We also cross-check ourselves. Where a public source publishes a rate for something we measure, the two figures, the window and the exact difference between them can be recorded in one signed record, and the record deliberately does not say who is right. Usually both are: ours is a realized return between two signed readings and a public feed is typically an average of daily quotes. That is a capability we run against our own numbers rather than something we sell, and no production record has been published yet. See how it is verified.
See a receipt produced and checked, step by step, or read the journey for your own role.