Bloat by size and efficiency is very different to a bloated feature set
In the argument around optimal set of software features, add adding "everything" in the Kano model provides a strong way to manage the coverage of features for a population. Enables one to strike a balance between "all the features for everyone" and "enough features to make enough people very happy"
Size and efficiency
Assuming software is equally well implemented, then the point Joel is arguing is that taking space to add features is well worthwhile.
This never holds for poorly implemented software that impacts UX directly.
In many cases that is true today given huge PC capacity and all the applications delivered on-line. However what Joel in 2001 did not predict is that many of users in 2015 would have on 16GB of storage on their primary computing devices, the smart phone. So big apps are unwelcome. Also these devices have limited navigation UI, and focussed purpose applications, so lot's of features is an anathema.
In that article bloat referred to size of software download, install and RAM usage.
Assumption of Monolithic Software
A key problem of the 2001 article is it was assumed, as was prevalent Microsoft model, that all features of an application had to be contained within that application.
Prime example Joel used in article is Mozilla. Today's Chrome and Firefox browsers ship with fewer features, but can run hundreds of times more functionality thanks to extensions and plug-ins. User can pick and choose what they need.
Likewise can make comparison Microsoft Word(tm) to Emacs or Sublime text.