Highlights for me: MSR has implemented post-quantum stuff (and it works), Spectre/Meltdown of course, google's TLS team talking about trying to test TLS 1.3 and everything that broke as a result, mozilla using F* to verify their crypto/TLS libraries and learning that "encrypted" aircraft (ACARS) communications use a monoalphabetic cipher with one of around a dozen hard-coded keys.
Academic = mathiness
real world = no mathiness
Academic = comes up with new ideas
Industry = tweaks and implements those ideas for practical applications
A better way to capture the differences you're looking for might be the following categorization:
1. Theoretical cryptography, which is primarily concerned with the mathematics and computational complexity of cryptosystems that are very new, not widely deployed or currently open research problems. This includes post-quantum cryptography, pairings-based cryptography, multiparty computation, indistinguishability obfuscation, homomorphic encryption, cryptanalytic attacks, etc.
2. Practical (real world) cryptography, which is primarily concerned with the mathematics and computational complexity of cryptosystems that can be practically used or which are widely deployed. Implementation, performance, peripheral security infrastructure and side channel attacks also fall under the purview of practical cryptography. The safety and development of cryptographic libraries and protocols falls into this realm. A lot of cryptanalysis also falls into this area.
There are many academic cryptographers in both domains, which is why I say that "academic" is not a useful signal as to which area someone is working in. As an example I have off the top of my head, Phil Rogaway (who won Real World Cryptography's Levchin Prize in 2016) is an academic cryptographer who specializes in practical cryptography. He invented PMAC and OCB, which are both practical primitives for use in authenticated encryption cryptosystems; nevertheless, research like Rogaway's requires significant understanding of relevant mathematics and complexity theory. On the other side of the spectrum, Craig Gentry is an industry cryptographer working at IBM as a research scientist; he won a MacArthur Fellowship for his PhD thesis, which was the first feasible realization of homomorphic encryption. He was also part of the research team that made the first significant progress towards multilinear maps in pairings-based cryptography using a graded encoding scheme.
Having implemented Monocypher, I can only concur: merely implementing the algorithm from spec required a solid understanding of some of the mathematical concepts involved. For instance:
Yes, yes, descriptive vs prescriptive, but I think it's dumb to destroy a perfectly good prefix just due to the fact that cryptocurrency is currently at the peak of the hype cycle.
I was going to say people might be interested in going, but it's already basically over.