Unrolling once is still what I'm referring to there.
> a new technique that was only recently developed.
That's simply false. E.g. search for 'covenant' or 'recursive covenant'. It's also described on the mailing list as far back as 2011.
(And, incidentally, the first example of a turing complete machine controlling the release of a transaction on Bitcoin was in 2016 and was vastly more efficient and private than the the approach used by the post-- https://bitcoincore.org/en/2016/02/26/zero-knowledge-conting... )
Whatever I read regarding "covenant" requires a protocol change. Independent of who described the underlying technique first, this doesn't invalidate the proof.
Sure it is. The state machine expressed in the script checks one (or more) steps of the update rule. That's what we mean by unrolling, not the "or more" part but the fact that the script is just running a simple circuit for a fixed operation.
> doesn't invalidate the proof
Proof of what? It's not a proof of script being turing complete. If you're claiming that it's that-- it's invalid on its face.
If you're saying it's a proof that script can implement a static state machine that runs one or more steps at a time checking some transcript computed by an external process and check consistency of state using the outputs-- then sure, that's not news, nor controversial, it's been known almost all of the system's life.
While most of what you write is true (and I believe that the article was written in bad faith), as the article uses state changes in the Turing machine as Bitcoin transactions, it is trivial to statically check the runtime of a state change itself. Whether it's a useful addition to Bitcoin or not is another question though.