Edit: I replied to the wrong comment. Hermes chose to start with the same engine as Boom - The J85. I'm assuming Boom chose it for similar reasons.
I like their explanation of the choice from a recent tour:
(Youtube transcript)
"So these are out of production, the J 85s. So we didn't work with GE at all. It was all just us working with, we were really working with the maintenance, repair and overhaul shops for them. That's really where the expertise and knowledge lies. These engines were, I think originally designed in the fifties. There's not a lot of electronics on board. There's no firmware we have to work through.
And really, it's a pretty elegant but hydro mechanical system for all the controls. So really it was about understanding the configuration of it and you can kind of chase down all the different tubes and everything to understand how it works. And then there's a suite of documentation out there. So it was really on us to learn how it worked."
J85s are old technology. They're a turobjet, not a turbofan, nor do they have any electronics on board. This makes them highly inefficient engines. They're cheap, there are plenty lying around from old 4th gen fighters, and plenty of old guys who know how to work on them.
I suspect they used these just to get the XB-1 airborne, making progress while they find a better engine.
Where would the LLM get the training data? It's not as if supersonic jet designers commonly post easy-to-ingest design data online.
Besides, most countries couldn't build a modern jet engine even if they had the exact engineering drawings of an existing one available. From fashioning single-crystal turbine blades to establishing a supply chain capable of the quality controls needed, the amount of hours spent is unreal.
As a corollary to AI taking up easy jobs like CRUD apps and spreadsheets, will human work be pushed to making things like silicon chips, nuclear reactors, jet engines and space elevators?
[1] https://www.hermeus.com/