But can’t you use very simple agentic AI to constantly update the code for web scrapers on new reddit? From what I understand that is the kind of coding LLM can do well and you can just unit test for errors and feed it back to the LLM until it works. Even requiring accounts, creation can be automated using LLMs too.

This sounds like it a loosing proposition and will only piss off hardcore users. Not that this would stop them. But I don’t think protection from web scrapers is a good explanation. Probably it’s just enshittification, forcing users to use their new interface to optimize business metrics. Someone is going to get a bonus because they improved the short sighted metrics.