I've been using and supporting Apple software since the 90s and I think this is more an artifact of most bugs not being especially memorable than anything else. People have complained about bugs after every release, a fair fraction of the complaints circulating online will be attribution error or overly-generalizing a niche condition as global, etc. and that seems to have been basically constant since OS X 10.2 or so.
There is one more interesting discussion, however, which I think is probably a legitimate trade-off: the entire industry has been moving to faster releases since it's easy to ship updates over the internet to huge audiences. When that wasn't possible and update cycles were measured in years, teams did spend more time on QA so you probably did have a somewhat lower chance of seeing bugs of the sort which were caught by that kind of process but at the expense of taking much longer to see new features or get fixes for bugs which slipped through either due to complexity or limited trigger conditions. Since the amount and complexity of software has generally been increasing, I think that this is generally the right trend with the addition of telemetry to collect crash reports, but it does seem like this could increase the perceived number of bugs even if the duration of time where you're exposed to a particular bug decreases.
I think there's _less software around_, insofar as there's fewer discrete software titles around, but each title has far, far more software code behind it.
I remember as a sprog, running my cracked copy of Adobe Photoshop 6.0 - that in order to do anything "cool" in PS (especially when you just can't grok how to paint, even with a Wacom) you needed a huge smattering of third-party Photoshop plug-ins: Alienskin, Kai's Power Tools, Flaming Pear, FilterFactory, etc - but nowadays while plug-ins are still a thing I find myself getting-by with only Photoshop's stock functionality. It certainly doesn't help that Apple continues toward completely locking-down their platforms: in-proc and binary plug-ins for Photoshop on iOS will never be a thing.
The other is the fact that server based software has an almost zero time deployment cycle, and even native code for desktop/mobile can be shipped really quickly. This makes software mostly an iterative process. If fixing bugs on a weekly basis in a major mobile app is no biggie, and almost nothing gets blocked because of a bug that can be just fixed next week, you just get a lot more bugs shipped to users.
I feel like we are creating really unpredictable, and hard to implement UIs in the name of "magic". Would a button in the UI been better? I think it would have been, at least for me.
Mostly, my comment was projecting my own experiences with complexity on the projects I have been involved with. You are right, in this case other factors probably dominate.
Bad software is more due to needing to ship stuff ASAP for marketing, the board of directors, and for the investors. This decline in software can really be matched up with the increase of startup culture.
I would also argue that complexity is developers own doing, trying to strap every new fangled library, engine, tool, API ,buzzword of the day, etc. to their program...when there is zero reason to. Nothing about what modern software does is more complex than it was in the 2000s