About Metaphoria
Metaphoria is an attempt to build something that reliably notices when a piece of real writing is using a mental model, and surfaces it in the moment you’re reading. The long-term goal is for the tool to make itself unnecessary: enough examples pointed out to you, and you start seeing them unaided.
Why build this?
People reason with mental models constantly and mostly don’t notice. Catching one in your own thinking is the first step to thinking better about it — but in the moment, while actually reading, almost nobody does. Training helps less than you’d hope. The gap isn’t knowledge, it’s attention.
A tool can sit in that gap. Not to grade the writing or tell you what to conclude, but to say “this bit is doing sunk cost” often enough and accurately enough that the pattern starts to stick.
What would make it good?
The bar is subtle cases — a model that’s genuinely present but never named. Anything can flag a paragraph that says “this is survivorship bias.” The version worth building catches the paragraph that quietly does it.
That bar only means something if it’s measured, so most of the effort goes somewhere unglamorous: assembling text where the answer is known in advance, and scoring the detector against it. A change ships when the scores say it’s better. The evaluation set matters more than the detector, because a detector can only be as trustworthy as the thing judging it.
What is it not?
- Not a course. No flashcards, no curriculum. The teaching happens on text you were already reading.
- Not content moderation. It surfaces reasoning; it doesn’t judge writing as good or bad, honest or dishonest.
- Not a writing assistant. It tells you what’s there. What to do about it is your call.
Where is it up to?
Early. The detector and its evaluation set are the active work; a browser extension is the intended way to use it. There’s nothing to sign up for and no date worth announcing. The blog is where findings from the build will go.