using machine learning to block distractions

During the summer of 2020, I tried to fight the problem of online distraction by building an intelligent, content-blocking chrome extension.

I got the idea for this project while in grad school. Many of my grad school lectures were hosted on YouTube. I was lucky if I could make it through 10 minutes of lectures before getting distracted by Key and Peele videos. I needed a way to consume content related to my classes on YouTube and Reddit, without wasting time on distracting content on those same sites.

I tried almost every content blocker out there (freedom.to, ColdTurkey etc) and none of them cut it. Most of them simply blacklisted ‘distracting’ domains. None of them were intelligent enough to allow me to view YouTube videos about Computer Science, while stopping me from watching Chapelle’s Show.

The extension I created took a different approach. Rather than enter a list of distracting domains, the user would input a list of topics they needed to focus on. Then, as the user browsed the internet, each page they viewed would get converted into topic space. If the topic vector of a page was too far from the topics that the user was supposed to be focusing on, the page would get blocked.

While the project has been discontinued, you can watch a video demonstration here (the real meat starts at around 5:25)