Well, we're working on a full implementation with IOKit. This means that macOS will handle all WPA2 handshaking, key management, and the rest. This significantly reduces the amount of work that we have to do, and is a massive plus for the security (and I'm generally uncomfortable writing any crypto code, as that is far from my expertise). As well, it should be measurably more performant, due to avoiding a second 802.11 stack. In fact, the majority of wireless kexts (including the realtek USB adapters) implement dual 802.11 stacks, and are often a mess to debug and fiddle with when things go wrong, with homebrewed implementation of WPA. Approaching wireless devices via this method requires either a massive department staffed with cryptographers, wireless experts, and potentially linux kernel developers, or entails a duct taped job where you merely strap IEEE80211 from OpenBSD to your driver, and call it a day. I'm simply not satisfied with the latter, and sadly do not have a department at a billion dollar company, and frankly, I'm sick of having a dongle strapped to my laptop.
It is definitely possible with all of the discoveries that I have made, however, I'm uncertain entirely on whether or not I want to subject myself to the pain and torture that implementing AWDL will cause. It is a possibility.
Unfortunately, due to a massive architectural change (and a lack of development devices), AX-series devices are unsupported. However, all 802.11 AC devices are planned to be supported (316x, 7xxx, 8xxx, 9xxx).
Run the logging script, and create a new issue on the repo with your log file attached.
We're crowd-sourcing testing to the community. We've found it to be significantly effective in determining compatibility with wireless devices, and I would love to continue that trend as much as we can. If you have Xcode installed, and can run a simple shell scripts, please upload your log file to the repo here. There are a significant number of devices which have been unaccounted for, as well as some devices having unknown compatibility. A sample size of n=500 is bounds better then a sample size of n=5.
In addition, if you are technically minded and don't mind reading and writing C/C++, I am ALWAYS accepting of pull requests and contributions. This is not a one-man effort, and my C++ knowledge is only so vast. We have adopted the Google coding style to the entire repo, and request that you format and lint your code before submitting a PR.