In the comments, Bob Murphy writes:

When Bryan says the first-best solution is to
tax education, is he just making a point that there are negative
externalities? In other words, does Bryan also think a government tax
on pollution is the optimal thing?

“Optimal” in the sense of first-best Kaldor-Hicks efficient, yes.  Of course there are many background assumptions here, too: That government knows enough to pick the efficient tax rate, that the deadweight costs of collection exceed the deadweight costs of bargaining, etc.

I realize he brings up public choice issues, but the standard
libertarian anarchist objection to taxation (even in cases of negative
externalities) goes beyond mere public choice arguments.

As I see it, the standard libertarian anarchist objection to taxation is moral: Even if a tax is first-best efficient, we shouldn’t do it.  I discuss my full views on Pigovian taxation here.

So I’m asking, before I get all worked up and perhaps write an
article over at explaining what’s wrong with Bryan’s view
here, can someone point me to a discussion where he spells out his
views on when it’s a good idea to tax something?

Again, there’s “good idea” in the sense of economically more efficient, and “good idea” in the sense of morally right.  My approach to economic efficiency is straight out of the textbook (or better yet, Landsburg), though I naturally think that most economists have a big statist bias when they apply the concept.  Morally, I don’t buy libertarian anarchist absolutism, but the real-world efficiency case in favor of taxation isn’t strong enough to overcome the presumption against taking people’s property without their consent.