> The implicit assumption here is that more is better,
No, that is the assumption that you are making and not what i meant. It is an undeniable fact that today we have a far far greater amount of Scientific and Technological Knowledge (sheer Data without making any value judgements) than in any time in the past Human history. Our Society's dependence on Science/Technology has also increased exponentially since the beginning of Industrialization thus forcing us to know/learn more and more. Also see my comment here - https://news.ycombinator.com/item?id=40154158
> The other position is that knowledge is only useful if it lets you do something that you couldn't do without it.
This is a very limiting viewpoint which is what the posted article and my comments are arguing against. The utility of some knowledge may not be apparent at its inception but may turn out to be of utmost importance when other fields have also developed so everything can come together into a greater whole. A good example is how Materials Science revolutionized Civil Engineering in the building of man-made structures.
> If you're gaining knowledge for the pleasure of gaining it, and I see no need to justify why one would do this, then whether you read deeply or read widely doesn't make much of a difference.
The justification has to do with getting exposed to a greater amount of concepts/ideas and not necessarily understanding/assimilating everything. Here reading widely does make a difference.
> But if you want to do something with that knowledge, the approach you take should be informed by what you want to do.
Only in the specific case when the end-goal/knowledge is clear eg. Engineering. Many a time in scientific research one does not even know the utility of something until sometime in the future when things align correctly. A great example is "Number Theory" which G.H.Hardy thought useless as a practical discipline but which has turned out to be indispensable in today's information technology (in cryptography etc.) centric world.