Hey HN, Our small team has been working on a side-project for the past few months. We wanted to see how far we could push the capabilities of LLMs with a pretty heavy task - analysing, grouping, and summarising the content of browser tabs or bookmarks. So we built a browser extension called TabCrunch. I've always had a problem with browser tabs and I know many others who do too. The problem is, once you start retaining reading material for later, it quickly piles up. People often have hundreds of tabs open across their various devices that cover many different topics. And it becomes impossible to organise them or even get through all of them. As the context capabilities of LLMs grew, we thought we could finally solve this problem. So we decided to give it a shot and see how far we can get. It turns out that it’s super demanding for today’s models, but with some optimisations and experimentation, we made it work pretty well. The link shares our methodology as well as a lot of what we have learned along the way. I hope it helps others experimenting with LLMs and I hope you find TabCrunch useful too! Because this is a soft launch, if you do register I would encourage you to use this code: TECHDEEPDIVEHNAPRIL24 to get 20 analysis of up to 500 links each. Just use it when you register, you can get the browser extension on: or your browser's extension store. Looking forward to hearing your feedback. |