How it works · 4 of 4
The third check — does it actually make sense? — is not one question. It is four, and each one gets its own reader.
Does the text resolve its own tensions?
Catches the writer who raises a real complication, calls it complicated, and moves on. Naming a problem is not the same as resolving it.
Does the text disclose the writer's stake?
Catches the sales pitch dressed as analysis. An argument can be honest about wanting something; it just has to say so.
Does the text avoid logical fallacies?
Catches facts that stop short of the point — true numbers, real events, and a conclusion none of them actually reach.
Does the text leave room to disagree?
Catches the argument armored against every answer: the disclaimer up front, and every sentence after it arguing the opposite.
The shape tells the story.
The four gates pair up. The top two — SL and SR — read the surface: how the argument is built and how it is delivered. The bottom two — FL and FR — read the foundation: whether it faces its hard parts and is straight about what it wants. The left side is logic, the right side is rhetoric. Plot the four scores on those diagonals and you get a kite.
Bedrock
All four gates high. Big and balanced — the argument holds, top to bottom.
Theater
High surface, low foundation. Sounds brilliant, stands on almost nothing. This shape gets flagged.
A diamond in the rough
Low surface, high foundation. Rough delivery on solid ground — worth more than it sounds.
Two of those kites can land near the same overall score. The shape is what tells them apart — a polished performance over a hole, or a solid case that just needs an editor. That is why the readout shows you the whole kite, not only the number.
Why four readers instead of one?
For the same reason a jury is not one person. A single reader has moods, blind spots, and favorites. Four readers who cannot talk to each other have to agree the hard way — each one finding the same document sound, for its own reasons.
These names sound technical because they are — the same four checks run on legal filings, financial reports, and answers written by AI models. What each one asks, though, is something you already ask yourself: did they deal with the hard part, are they straight about what they want, does it add up, and can they imagine being wrong.
What the score reads as.
The four results combine into one number from 0 to 100, and the number lands in one of four bands.
Band one · 90–100
Bedrock
It holds, top to bottom. The reasons carry the claim.
Band two · 70–89
Flagstone
Laid well and carries the weight, with seams a reader should know about.
Band three · 40–69
Faultline
Solid on either side, broken in between. The reasons stop short of the claim.
Band four · 0–39
Quicksand
The shape of an argument, with nothing underneath to stand on.
One more thing, and it matters: the gates read structure, not truth. A document can read Bedrock and still be wrong about the world — it made its case honestly and well. The gates tell you whether the argument deserves your attention. What you decide after that is still yours.