You can play the Zeno game with this.
The first half of the items is actually an article of n/2 items - so you think of n/4 in the first 5 minutes.
So you never actually think of an item.
Sci-fi quantum physics takes care of that. Eventually, two correlated ideas randomly pop into existence, and then it's just a race to see how many more ideas you can build on them within 5 minutes.
It could be (cumulative) exponential. Ideas would then be independent and random. That doesn't feel right, because ideas are very much related, but a better model doesn't immediately spring to mind.
Had he said half, I would never consider the distribution of ideas over time. And I bet pg considered the distribution while writing the essay, and he probably also realized it wasn't really all that interesting, and consequently left it out.