There's not much detail available through those sources, but there's enough to imply that, like the previous proofs of universality for cellular automaton Rule 110, the proof requires relaxing the definition of universality: allowing machines that never terminate, and that start with an infinite set of nonzeros on their tape, with some more complex condition than termination for signaling that an input is accepted or rejected. I haven't had a chance yet to look at the full proof, but it's also online.
In other news, Wolfram has a blog.
Via a comment thread on an unrelated post at Good Math, Bad Math. See also Boing Boing, Geomblog, Slashdot, Gasarch.
ETA: Vaughan Pratt finds a problem. See also /., Aa.