A dynamic website categorization algorithm feature where given a category such as sports (can be augmented with a list of example websites or keywords, perhaps), the website filter is able to dynamically determine if the website fits that category based on the page contents, even if the domain isn't traditionally sports (e.g. a general online newspaper outlet featuring a sports article).

Similarly, sometimes general forum websites like Reddit can be used productively so a dynamic website filter could better determine if that subreddit or specific post fits a category or not.

On the backend, it would be a machine learning classification algorithm, of course. The user could even help provide labeled data over time by providing feedback on the Block page if the website was correctly or incorrectly blocked. In the latter case, Tech Lockdown wouldn't immediately give access to that page, of course, and would need account for bad-faith label verification, perhaps by aggregating feedback across different users and/or getting other users to help verify the labels other users provided.

In cases where the user can access a page that should have been blocked, there can be a page in the Tech Lockdown dashboard where you can retroactively label visited pages correctly (when the user is their more enlightened self) so it learns those examples. Perhaps even a recurring email or notification pop-up that prompts the user to help classify previously visited page and/or some gamification element to encourage users to help contribute to labeling or confirming model-applied labels (also relevant to cross-verifying labels to account for bad-faith / mislabeling from other users).

Obviously more of a long-term feature but I think this would be closer to the ideal end state for an online / device distraction-blocking service.

See potentially useful links on website categorization algorithms: