I do not believe that swapping to disk is a mortal sin, especially with the fast storage we have today. But I do think that for the prices Apple charges for their hardware, arguably everything except the base model Mac mini and M1 MacBook Air basically, they should be giving users 16GB of RAM and probably 512GB of Storage as well. The hardware is too expensive to excuse the small amount of RAM - whether the users need it or not, they're paying enough for it.
I do believe 8gb is quite pathetic on a machine from 2021 upwards (considering older models are still on the refurb store occasionally), that's not the issue.
The issue really is with not understanding unified memory and how deleting the front side bus (not having to go through a memory pipeline to access your RAM/Disk) is really a significant game changer.
It's not comparable with the way Intel Macs work that still have a Front Side Bus and a North Bridge to go through to get to the RAM chips themselves, and then interrupts (meaning requests to access which blocks up the whole bus in traditional motherboard designs if any of these pathways want access to the CPU) which may be limited by their inherent bus speed vs M1 which deletes this whole issue, and therefore doesn't have the bottle neck of an FSB or North Bridge to deal with.
You would really have to have an IT background to understand what this means in terms of a traditional motherboard vs an SOC.
This is how a traditional computer works, it has all kinds of limitations for each of those pipelines that go to the RAM, PCI-E cards, etc, that have to interact with the north bridge, and traditional south bridge (albeit a lot of modern computers don't have a true south bridge anymore)
Everytime something needs to happen it has to be communicated through one of the pipelines and then to the north or south bridge controller and then the north or south bridge has to interact with the CPU, these are all points where bottlenecks can occur that don't occur with the M1, M2, M3 architecture because all of this now happens on the "CP/GPU" itself.
It's hard to find a good example of a diagram to compare a monolithic (formed of a single chip) CPU dye to but you get the point. But if you can't see why the board design below is slow, then you need the training to understand why.