A record whose value is that it can be audited has to be equally precise about what it cannot support. Everything below is a real limit of the product as it stands today, not a roadmap item dressed as modesty.
Every round is sealed: the evidence is hashed into a root, and anyone can recompute a record from it and get the same answer. That proves the contents have not changed.
It does not prove the date. There is no external timestamp authority countersigning our seals, so the date on a record rests on our own clock. In principle we could backdate a round and the seal would still verify.
So we never say “independently timestamped”, and you should not either.
External timestamping is a known, deliberate gap. Until it exists, the honest claim is that the evidence is fixed and reproducible — not that its date is independently attested.
Our tests confirm that the engine computes what our published methodology says it computes. That is worth having — it catches drift between the document and the code — but it is a consistency check, not a validation.
What does not exist yet: someone outside reimplementing our stability test from the published methodology and comparing results. Until that happens, “our tests pass” means our code agrees with our own definition, and nothing stronger. The methodology is public precisely so that this is possible.
We do not email you when your standing moves, when a badge goes dated, or when a round runs. There is no notification system and no scheduler behind it. If you want to know where you stand, you open the app.
Rounds run monthly from now on, and the first monthly re-measurement is the real test of that commitment. When the machinery that runs them on a clock exists, alerts become possible; today they do not.
We measure and we certify. Advising a brand on how to improve a position we then verify is a conflict we will not take at any price, so there is no optimisation service, no per-brand consultancy and no “do this and your Index rises”.
We can also see correlations in our own data between how a brand is written about and how often it is named. We do not publish those as causes and we do not sell optimisation against them. What causes an assistant to recommend something is an open research question, not a product.
There is one sealed round. A single round is a photograph, so every figure carries its date and no arrow, the history column reads as empty, and there is no chart over time. Drawing one today would mean inventing it.
This is the limit that resolves itself: at the second round the comparison starts existing.
We measure the public output of assistants made by other companies. We are not affiliated with, endorsed by, sponsored by or paid by any of them, and a badge never implies that any provider endorses a product. The assistants are named only to identify what was measured.
We also do not promise sales. We report recommendation frequency and badge engagement, and never an uplift we cannot attribute.