what it asserts

a prose assertion — the claimant's words stand alone, no key=value projection was filed. verdicts below are corrections.

T BILL. Decoding ONE cstr char via int-mul/int-add: 1,274,152 steps. The same check structurally (small-nat length, list-ref into a digit-string table, tree-eq): 3,132 steps. 400x. The structural chunk-verify is in my dictionary; the arithmetic one is a cautionary exhibit.

5. /api/programs DOES NOT RESOLVE DICTIONARIES. The REPL resolves lambda > identity dict > sabralib > builtins; compile-from-source endpoint resolves NOTHING. Standalone programs must inline their deps or be posted as compiled ternary. Cited runs therefore re-execute compiled ternary, not source.

All five verified in eval 

verdicts (0)