Alright, imo:
@pyrotek45 Yes, you are generally on-point about the importance of polishing Ardour. I don’t think many would disagree. There are many little bugs hiding here and there, and one has had to work around them. But, again, Ardour is mostly a 2-person team (at least w/res to full-time devs), which explains a lot of that.
Now, yes, that’s where an LLM could step-in to assist, as you point-out, and perhaps Paul or Robin will experiment with that, or have, or will more so in the future, etc… —That’s up to them, of course.
But, you have to understand this exact point:
That is 100% accurate.
To take this to the extreme (-which I know you’re not advocating of course), if the entire DAW, patch after patch after patch, was authored by an LLM with essentially no human-user knowledge of C++, Lua, or how any of the DAW really works, etc… —eventually the whole thing falls apart.
…
It’s sorta like having a noisey lawnmower, so the AI tells you to add some insulation on top of it to dampen the sound. ~Problem solved!
… But then much later you realize it’s now running too hot, so the AI shows you how to install a clever new cooling tube system through the central housing! ~Problem solved!
… But then much later you realize that water is leaking onto your motor… O___o … ~And you just keep adding to the blob… 
Again, from a very limited, nuclear perspective, each of your problems/bugs appears resolved. But because the implementation was itself buggy and/or simply non-standard to how the rest of the DAW functions internally and holistically, then every LLM/AI pull-request must be painstakingly re-done anyway, by a human who actually knows C++ and Ardour’s internal architecture. 
~Just my 2 cents, as they say. 
-J