Assume ℎ:(𝑃⇒⊥)⇒⊥. To apply LTT-Classical, add 𝑘:𝑃⇒⊥. Then ℎ,𝑘:⊥ by LTT-⇒-E; LTT-Classical discharges 𝑘 and yields 𝑃. Finally LTT-⇒-I discharges ℎ. This is the unique classical step. Every line is an entailment judgment Γ;Δ⊢𝑄; no term judgment Γ⊢𝑡:𝐴 is produced, so the derivation adds no inhabitant of an object type.
The number-quantified proposition is small and hence analytic. It may occur in comprehension and in both induction schemes. The set-of-naturals quantified proposition is analytic but not small; it is forbidden in comprehension and 𝖫𝖳𝖳0 induction, but permitted in 𝖫𝖳𝖳∗0 induction. The set-of-sets quantified proposition is neither; all three uses reject it. The matrix being a tautology or ⊥ does not change these classifications because the offending binder remains.
Let 𝑝(𝑛) be the small code for ∃𝑘:𝖭.𝑛=𝑘⋅𝑘 and put 𝑆:={𝑛:𝖭∣𝑝(𝑛)}. Then LTT-Set-𝛽 gives Γ;∅⊢𝑛∈𝑆⇔𝖵(𝑝(𝑛))≡∃𝑘:𝖭.𝑛=𝑘⋅𝑘. The left-to-right direction is biconditional elimination followed by its first implication; the right-to-left direction uses the second implication. Existential introduction at 𝑆 proves the translated comprehension instance.
In type-free second-order arithmetic the translation is ∀𝑋∃𝑛(𝑛∈𝑋), where 𝑋 is a set variable and 𝑛 is a number variable. Removing type annotations does not change the source-level smallness judgment. The outer quantifier ∀𝑋:𝖲𝖾𝗍(𝖭) does not range over a decoded small type. It is therefore the precise obstruction to a small proposition code and hence to comprehension.
For arithmetical 𝑃(𝑛), translate comprehension by {𝑛:𝖭∣|𝑃(𝑛)|} and use LTT-Set-𝛽; translate number induction by LTT-Nat-Ind0, since |𝑃| is small. These are the two nonlogical cases of the forward rule induction.
The imported reverse chain is 𝖫𝖳𝖳0⟶𝑇𝜔𝑈⟶𝑇𝜔⟶𝑇2⟶𝖠𝖢𝖠0. Definitions 5.30–5.31 provide satisfaction for the bounded signatures; Theorems 5.32–5.33 give soundness and completeness, and Corollary 5.33.1 therefore lowers a 𝐵𝑛+1 derivation of an already-𝐵𝑛 judgment to 𝐵𝑛. Corollaries 5.33.2–5.33.3 complete the displayed chain. This is a semantic existence argument using interpretations and satisfaction, not a recursive transformation on derivation syntax, so the imported theorem does not supply an effective proof translator.