The thesis is a portable phrase: slogans strip mechanisms. Apply the page's own Level-1 rule to it.
Level 1 must not be self-explanatory; it should create a question. "Compression paradox" and "maximum portability produces maximum distortion" answer. They occupy the slot.
Object: the title, the thesis box, and the three-sentence synthesis. Who may fail those phrases for being replacements rather than invitations is not named. If a slogan that feels like understanding is the failure mode, the page's own portable layer is a candidate.
The page is not collapsing to Level 1. The thesis is a shell over two named operations — attribute substitution versus relational isomorphism — three worked unpackings, and a four-feature test in §VII.
A title that names a paradox is incomplete in the required way: it asks what the paradox is. The body is the diagram and the model. The design implication is for people shipping slogans without that model, not a claim that this page is only a slogan.
Then run the unpacking test on the synthesis box, not the title.
"Most intellectual compressions act as attribute substitutions" is self-explanatory. Expanding it recovers "formalization is often the opposite of understanding" — the subtitle — not the two operations, the institutional preconditions Smith actually named, or the four features that distinguish a map from a postcard.
The page's own rule: unpacking a bad compression recovers the salient attribute. The three-sentence box recovers the warning, not the mechanism.
The test is mechanical. Score the thesis box and the synthesis box on the four features.
Does unpacking recover structure-mapping — surface versus relation — with a prediction and a boundary, or recover "slogans are dangerous"? If the latter, recode those boxes as attribute substitutions of the page's own argument. The body can still be the model. The portable layer would then be the thing the design section forbids.