[infogulch]

There's still no direct way to detect which pin blocks it from opening. Maybe you could determine if the failed pin is the same as a previous attempt by listening with a stethoscope, or very finely measuring the turning angle, but you can't directly feel out which pin. So there may still be a way to reduce the search space in theory, but that attack still seems very difficult to pull off, and for the complexity it seems vastly better than previous locks.

[IIAOPSW]

Well, it didn't advertise itself as a "very difficult to pull off lock", or a "vastly better than previous lock". It advertised itself as an "unpickable lock". That's a very strong claim to be making. I wouldn't be satisfied with anything less than a proof that it is impossible to reduce the search space down to sub-exponential.

Building a lock which does not leak any information about what's happening inside is equivalent to building a mechanical, room temperature quantum computer. For if that information isn't leaking to the environment in some way, there is no mechanism to decohere a superposition state. Hence in principle a mechanical lock which is secure in the information theoretic sense is impossible. It is still theoretically possible to make a computationally secure lock (eg a mechanical implementation of a hash function). But there's currently no real proof that one-way functions are actually one-way. The security of such a lock is subject to a foundational guess in cryptography.