The main purpose is to eventually provide a tailored experience for our users, by learning a recommendation system from historical browsing data. Currently, we only recommend related topics on the bottom of the page by distance of embeddings generated from topic text, which is not enough. In the future, everyone's homepage will be tailored just for itself. For anyone who want to dive deeper in this topic, read the excellent blog posts by Netflix: 1, 2 (behind semi-paywall, but you can easily bypass it by signing-up or use uBlock Origin). I believe, by leveraging both distance between embeddings and causal relationships in browsing history, our users will expect to read more topics that they want to explore, although this will certainly require plenty of work to be done.
Also, in fact this move does not hamper normal browsing at all, since everyone can register an account for free. We are committed to a free Internet and will never charge users directly or indirectly (e.g. by introducing advertisements). Also, by prompting visitors to login, we can get more people engaged in discussing, which is a win-win situation.