> Many if not most of the special-purpose objects around us are going to be replaced by apps running on tablets.
I respectfully disagree.
The trend for specialised vs. generalised devices seems to go in cycles over a period of a few years, in a similar way to the classic thick vs. thin client cycle. Consider games consoles vs. gaming on PCs, the iPod vs. mobile phones with media storage, etc. Neither extreme is ever going to take over entirely, and the bias moves as technology evolves.
I think this is mostly driven by trying to balance convenience and power. When new tools come along that are generic enough to make a certain broad class of jobs easier, we tend to jump on them. Many jobs get moved to those devices, and specialist devices that used to perform those jobs become obsolete. On the other hand, if you get too generic, you start to introduce waste and therefore inefficiency, which pushes things back the other way. Also, if your generic device is OK at doing lots of things but not particularly good at any of them, there is still a market for specialised devices that do a particular job better because their priorities are more appropriate.
We used to write software that ran on desktop PCs, but it turned out that a lot of practically useful software is essentially a simple user interface to a simple database. Native applications had common pain points in this field that could be overcome by hosting the code and data centrally, in areas like installation/updating/backup. Thus Web apps were born.
However, today, we're seeing major players in the industry trying to turn just about everything into such an application, and they are failing. It turns out that while Web apps are great for presenting relatively simple database UIs, they are relatively weak at performing most other tasks. Cloud computing is a pretty direct extension of the same argument.
I suspect things will go the same way with phones/tablets/mobile devices. A generic mobile device with a bunch of common built-in peripherals and sensors will solve a wide variety of real world problems, and thus various kinds of mobile app have been born. No doubt many more variations will follow over the next few years, as these devices support new functionality that was not previously available and ideas will spring up to take advantage of that functionality. The devices will be good enough for these purposes and will be widely adopted as a result.
On the other hand, Swiss army phones could easily start to suffer from both overspecification in breadth of features and underspecification in performance of individual features. For example, the suggestion in the article to replace reading glasses with a smart phone seems unrealistic and oversimplified to me: it sounds great initially, given that we have cameras and screens on these devices, but then you consider the vast range of different reasons that people are prescribed glasses, the consequent individuality of each prescription, and the fact that glasses do not generally require holding in your hand to use them.
In short, I'm afraid I don't buy pg's argument here at all. A certain class of applications, some of which already exist and some of which will be developed, will probably move to handheld multipurpose devices. However, specialised tools aren't going away any time soon, because any generic device is always going to be either a poor replacement for a good tool or too highly specified to be efficient for a broad market, even if the technology exists to combine high-quality implementations of all the required features within the required space and cost constraints in the first place.