I don’t see a problem with talks that show code (to support whatever the talk is about) without giving it away.
Read the whole thing in context. He's apologizing for jumping the gun - for giving a talk about a library he intended to release, before it was ready. The implication is that it was probably a "hey try this library!" kind of talk, that has little value to the audience without the code, so he's saying he should have waited.
Giving talks about closed code that will never be released is separate from all that, and clearly not what he's talking about. Hence the exception, "case studies and such".