Given that Google Trends doesn't show that bump, I'd assume it has to do with how the data was collected. Maybe all stories with < X votes/comments older than 2015 are not included, or deleted from whatever index you used?

In case he doesn't, you might as well email him about it. He's a very responsive guy and might find it interesting.

I think text vs. link used to be XOR, but isn’t any longer.

It’s still outside the hn mainstream to use both in the same submission, so that might be biasing the model in strange ways.

From the post:

> But to simplify, instead I’ll just limit to stories that have only text bodies, instead of links.

This line implies that pre- and post- 2016 stories are text only, so this change should not affect the data so much.