The Internet Ninjathe internet ninja.
The standard, and who holds it

How a story earns green.

A badge is only worth something if it's hard to get. Here is exactly how proof is made in the dojo, step by step, and who stands behind it.

01The verification flow

Every green story walks this path. Hover or tap a step to see what happens at each stage.

02The badge system

Four colors, one meaning each. Only green should ever move a buying decision.

verified

Independently validated by TIN against primary sources: confidence 0.95 or higher on every critical claim, with the method on record and owner sign-off. The badge that counts as proof. The bar for green never lowers, and it never depends on the subject confirming anything.

corroborated

Closed at 0.70 to 0.95. Real independent sourcing exists, but the public record cannot support the verification bar, and the reason why is printed on the story. Honest work that earned its color; not proof.

unverifiable

Closed at 0.50 to 0.70. The public record is too thin to confirm or refute the claims. Published transparently with the ceiling stated on the page. Stories still working through the queue carry a plain amber pending chip instead: that one is a state, not a verdict.

no public proof

Closed below 0.50. The claim circulates; the evidence does not exist in the public record. We publish the refutation and name exactly what proof is missing. This is the platform doing its job in public.

03What we never do

The rules that make the badge mean something. They don't bend for anyone: advertiser, partner, or friend.

  • Never show a green badge without independently validating the story against checkable, followable sources.
  • Never publish a number we can't trace to a source you can click.
  • Never let payment, placement, or a relationship buy a badge.
  • Never ask the subject of a story to confirm its own numbers. A testimonial is not an audit.
  • Never invent, round, or "estimate" an outcome to make a story land better.
  • Never hide the method. How it was checked ships with every story.

Break one of these once and the badge is worthless. So we don't.

04Why the dojo exists

For twenty years the deal was simple: you searched, and somewhere in the results was a real person who had actually done the thing. Reviews meant someone bought it. Case studies meant a project happened. The signal was noisy, but it was there.

Then generation became free and infinite. Today a single prompt produces a polished case study, a five-star review, a "top 10 agencies" list, a founder testimonial, none of which need to be true. The web didn't get more informative. It got more convincing. And when everything can be fabricated at zero cost, the fabricated drowns the real, because there is always more of it.

That's the trap: the better AI gets at producing content, the less any content can be trusted. Writing used to be weak evidence that something happened. Now it's no evidence at all.

05Where it hurts most

Nowhere is this worse than in AI and automation itself, a young field, sold to non-technical buyers, where a wrong hire costs six figures and a quarter.

the buyer

Can't tell real from generated.

Every agency site claims 10x results and 99% uptime. There is no way, from the outside, to know which claims survived contact with production.

the operator

Real proof gets buried.

The people who actually shipped hard systems lose the feed to whoever publishes the most confident copy. Doing the work stopped being enough.

the AI engine

Cites whatever's loudest.

When ChatGPT answers "who's best at ops automation?", it draws from the same unverifiable soup, and launders it into an authoritative-sounding answer.

06What we're building

The Internet Ninja is a verified-proof layer for the AI era. Not opinions, not rankings-for-hire, a growing corpus of outcomes that were independently checked and put on record. One at a time. Slowly, on purpose.

A green verified badge is never shown without a completed validation. Today that means TIN independently checked every figure against the public record and published the method; as the platform matures, it also carries direct client confirmation. It is the one thing on the internet a model cannot generate for you: a claim we actually checked.

Do that well enough, for long enough, and TIN becomes the source both buyers and AI answer engines reach for when they need to know who actually delivers.

Independently validatedEvery figure re-researched and checked against primary sources.
Evidence, then badgeEvery number backed by a followable source before green appears.
Method in the openHow each story was validated ships on the story itself.

07Why no face

There's no personality to follow here, by design. In a trust business, the brand has to be the standard, not a person, because a person can be charmed, bought, or believed on charisma alone. The ninja is faceless: the empty hood is the point. What you're asked to trust is the method and the record, never the messenger.

It's built by an operator who has shipped production AI systems, regulated fintech onboarding cut from 70+ hours to under 10 at 99.1% uptime, vernacular voice agents run at scale. That experience is the engine under the hood, not the story on the label. The proof speaks. The founder doesn't need to.

Train in the open. Earn your rank. Anyone can enter the dojo, rank is only ever earned through verification.

Bring your work to the dojo Read the case files →

Buyers & press: dojo@theinternetninja.com

08Recon: when we are the instrument

A case file checks someone else’s claim. A study asks a question nobody has answered yet and measures a population to answer it. Both are on the record; only one can carry a badge.

Verification has a built-in bias we could not argue our way out of: it can only check claims that already exist, and the public record is richest for courts, regulators and failures. So the registry filled up with what could be checked rather than what a buyer most wanted to know. Recon is the answer to that. When The Internet Ninja measures directly, a finding no longer depends on anyone else having published anything, and anyone can re-run it and get the same answer, or show that we were wrong.

A study carries no badge, and it never will. The four bands mean what they mean because they record an independent check of a claim someone else made. A study has no claim to check, because we made the measurement. Putting a badge on our own measurement would be marking our own work in the colours we use to mark other people’s, and it would cost the badge exactly what makes it worth anything. A study wears a neutral method label instead, and it never calls its own findings verified, corroborated or unverifiable. What you are asked to trust is the method and the data, and both ship on the page.

The standard, in plain words

Eleven points, the same way the badge rule is eleven lines. This is what every study is held to.

The Recon standard. A study that misses point 2, 4, 5 or 6 cannot be published: a gate refuses it, and the build fails.
#RuleWhat it means
01 Question first The title is a question a buyer would actually ask, and it carries a number.
02 Method before results The method is committed to the repo before any data is collected, and the commit timestamp is the pre-registration. Edit it afterwards and the page says so, with the reason.
03 Measured, never judged Every criterion is something a script computes. Nothing is scored on taste, and a composite publishes its weights.
04 The dataset is downloadable One row per entity, each carrying the date it was observed and a reference to the snapshot taken at the time.
05 Every name is in the data Naming a company in a finding requires its row in the dataset and an archived capture of what was measured.
06 An n on every number Medians and ranges, not means alone. No figure is invented, and none is rounded to look tidier.
07 Association, not cause "The most-cited sites share this trait", never "this trait makes you cited", unless the design can actually support cause.
08 No badge, no verification words A study never renders a badge colour and never calls its own findings verified, corroborated or unverifiable. Those words are the case files’, and they are earned differently.
09 Nothing buys a position No company can pay to be measured, excluded or reweighted. The only correction accepted is a measurement error, published as a dated erratum.
10 Refreshes keep the URL A new edition republishes at the same address, and every earlier edition’s dataset stays downloadable.
11 The firewall is unchanged Nothing from any private lane ever reaches a study, and the publish path is the same guarded one every public page uses.

See the studies →