It's that there's a notion of a device that has so many features that it becomes "too useful." There is only so much time you can devote to so many features. Yet it's clear, for example observing the uptick in sci-fi computer interfaces in movies and such at the time, that crossing a threshold of "enough features" at once was useful at a certain point - having a pocket-sized Internet-capable device with a small-format camera that didn't suck for one thing.
There was also the essay posted recently that argued for a macOS release focused on bugfixes and stability over the disruption of new features with accompanying issues.
I've been wondering for a long time now, at what point will so much innovation have already happened since the 90s-00s that there won't be enough actually useful features to tack on to the next release of X thing except ones that solve problems we didn't have? Has that point already passed? I remember some iDevice releases weren't as notable upgrades as their predecessors for example.
In my opinion, if the AI revolution hadn't happened exactly when it did a couple years ago, this problem of diminishing tech returns would have been much more obvious than it is already. In fact, I think the current LLM rampage of sorts acted as a flood to fill the drought of incremental innovation that would have otherwise occurred.