Moving past the silly headline, there is actually a pretty substantial achievement here. Using formal verification tools, the team proved the following properties for the library:
* Memory safety (no buffer overruns etc)
* Type safety (all compiler-visible interfaces/abstractions used as per spec)
* Functional correctness (all the crypto implementations are faithful to their algorithms)
* Side-channel resistance (all crypto is constant time)
This has been confirmed by fallible tools and is checked against human-made models which are also fallible, but this code is still likely to be about as close to bug-free as currently possible.
I think Barghavan's work in this area are the future of cryptographic coding in the medium term and likely all security-sensitive code long term.