For instance, I have a working microkernel written in a Lisp dialect for embedded devices. Compiled to native machine code. 100% LLM generated. ~70k loc. In benchmarks it outperforms most other embedded kernel projects by a significant margin. And it only took around ~$1500 in tokens (API costs all included).
Not only was it thinkable, it was common: <https://en.wikipedia.org/w/index.php?title=List_of_self-boot...>
Of course, it would be possible to add new drivers only when necessary, but that would also allow for security problems.
So, in theory it might work, but in practice it would require quite a bit of thought.
Includeos in particular then went towards being a complete "application server" and apparently failed to gain a business as Docker became successful.
I know nothing of hardware, but as far as I know, my keyboard and mouse work everywhere because there is a formal specification on how human interface devices are supposed to operate.
There are probably good (and anti-competitive) reasons for why hardware still needs bespoke drivers, but from the outside, it seems like something we could address. I have no interest in loading your artisanally crafted Wifi driver.
This led to some hilarious implementation shenanigans for the Wii as a transitional console, where the Home button pause screen is not in fact a task switch to some underlying console OS but rather a piece of the SDK that is separately-delivered from each individual game.
For actual real hardware, not really
An alternative approach where it is just one big-ass logical expression is just not better.
Same thing with code, I think - you need some intermediate results like a calling convention, helper subroutines, etc.
A sufficiently powerful AI can do compilation "mentally" - i.e. producing machine code conforming to a specific calling convention. It can also decompile machine code. But you, obviously, don't gain anything doing it this way, if there's one-to-one correspondence between high-level code and machine code. You might as well just write high-level code.
I really hope that software becomes more efficient. But I don't think that it can only be done by generating machine code directly.