This post is using regression to build a reward model. The reward model will then be used (in a future post) to build the overall RL system.

Here's the relevant text from the article:

>In this post we’ll discuss how to build a reward model that can predict the upvote count that a specific HN story will get. And in follow-up posts in this series, we’ll use that reward model along with reinforcement learning to create a model that can write high-value HN stories!