Undominated.ai

Independência

Undominated.ai does not take a cut of inference. There is no affiliate revenue, no paid placement, and no gateway. Revenue is none in v1.

none cut of inference. We do not sell inference, take a percentage, or route traffic.
none affiliate, placement, or gateway. No monetised outbound links, no fee for a row, no token reselling.
none revenue in v1. A paid data feed is the only line contemplated later, and it would sell accuracy, not ranking.
not built the Blind Rank Test. It is specified. The hashes are not here because the test is not here.

1 · What we do not take

We are not a router. The moment we earn a margin on a recommendation, the recommendation is worth less. So we do not sell inference, take a percentage, or send your traffic through us.

We are not an affiliate directory. There are no monetised outbound links in the product. No provider can pay for inclusion, ordering, a badge, or a mention.

We do not take placement fees, provider sponsorship, or payment to review a ranking before it is published. We do not accept an API credit or gift conditioned on coverage. We do not resell tokens. We are not a gateway. Becoming one is the obvious way to make money here, and it would cost us the only durable advantage we have: being right when others are not.

2 · What we do take

Nothing, in v1. There is no revenue stream.

The only line contemplated later is a paid data feed: history, provenance, and change alerts as data rather than as a website. That customer would pay for accuracy. They would not pay more when we recommend one model over another.

If we ever take vendor money, this page changes in the same commit as the money — before the money arrives, not after.

3 · What you can hold us to

  • No cut of inference. We do not sell inference, take a percentage, or route traffic.
  • No affiliate revenue. No monetised outbound links in the product.
  • No paid placement. No fee for inclusion, ordering, a badge, or a mention.
  • No gateway. We do not resell tokens and we do not sit on the request path.
  • No pre-publication review. No provider sees a ranking before the public does.
  • No gift conditioned on coverage. An API credit offered in exchange for a listing is a listing we do not take.
  • No revenue that pays more when we recommend one model over another. That is the whole firewall. A later paid feed, if it exists, sells the data, not the order.

4 · The Blind Rank Test is specified and not yet built

We will not invent hashes for a test that does not exist

The specified test computes the ranking twice: once normally, and once with every monetisation field replaced by a zeroed stub. The two rank-relevant hashes must match, or the build fails.

That hash would prove one narrow thing: no monetisation field is readable by the ranking code path. It would not prove independence. A human who picked the lens weights after seeing which providers pay still passes a byte-identical test. The hash covers the machine. A public changelog would cover the human. Presenting the narrow proof as a broad one would itself be the failure the mechanism exists to catch.

When the test is built, the hashes and the reproduction command will appear here. Until then this page says the test is specified and not built, rather than displaying a figure we cannot trace.

5 · If this changes

If any commitment on this page changes, the change is a public commit, and this page is updated first.

How ranking works, and where the data is weak, is on Metodologia.

Len van der Hof · . English is the source language of these commitments. A translation that drops a hedge is a mistranslation. The English original is authoritative.

Markdown for LLMs