I don't see how s-expressions are inheritently less "typed" than utf-8 text?

Have a compiler target dialect that expects "(int 5)" and "(int x)" have a type-inference step that target such "typed" s-expressions?