1) Many languages have eval. So what he's saying, that Lisp is somehow unique because of this feature, still doesn't make sense under that assumption.
2) Wouldn't his C vs. Lisp be more of a comparison between compiled vs. interpreted languages? The notion of eval in a compiled language doesn't really make sense. Similarly, the lack of eval in an interpreted language doesn't make sense either.
If he is referring to eval, how would eval provide an open future whereas C would be locked by strange limitations?
Minsky seems to be dumbing things down in this interview. I believe when he says "Lisp programs that write other Lisp programs" he's referring to something more complex. Maybe extensible runtime. Maybe the Lisp he's referring to is different from modern Lisp implementations.