Launch HN: Browser Buddy (YC W24) – A recommendation system for Internet writing

Hey HN! We're Arnav and Jeremy and we're building Browser Buddy, a recommendation system you can chat with to find quality Internet writing based on your interests and aspirations.

As we've grown up, this writing was a source of inspiration that helped us discover ideas and opportunities we didn't even know we were looking for (ex. applying to YC because of PG's essays).

But despite so many new creators and websites coming online, the best of the open Internet remains hard-to-find, scattered across personal sites (https://www.paulgraham.com/, https://www.eugenewei.com/), niche publications (https://www.noemamag.com/, https://worksinprogress.co/), and various independent publishing platforms (https://bearblog.dev/, https://substack.com/, https://medium.com/). Outside of "social" media platforms, there's very high friction to get into a new subject or stumble upon credible people who write about your interests.

We feel there should be an easier (and mobile-friendly) way to find fantastic media and curate this intentional, interesting information diet for yourself.

Browser Buddy is an iOS app that curates this interesting, thought-provoking writing for you from across the Internet. It's particularly good to explore topics like programming, startups, math, philosophy, machine learning, and design.

Here's a demo: https://www.youtube.com/watch?v=VmEeo4mjc7U

Here are some example recommendations:

"I'm trying to grow my early stage consumer internet company": https://cdixon.org/2015/01/31/come-for-the-tool-stay-for-the... https://www.lennysnewsletter.com/p/how-the-biggest-consumer-...

"I want to learn how to build beautiful web interfaces": https://frankchimero.com/blog/2015/the-webs-grain/ https://alistapart.com/article/responsive-web-design/

"Teach me the history of the Internet": https://www.mic.com/impact/how-geocities-webrings-made-the-9... https://computerhistory.org/blog/history-of-the-future-octob...

"I'm trying to read more about games and game theory": https://franklantz.substack.com/p/playing-balatro https://joecarlsmith.com/2022/03/16/on-expected-utility-part...

"I've been getting into network science and network theory": https://www.ribbonfarm.com/2010/10/27/warrens-plazas-and-the... https://www.strangeloopcanon.com/p/seeing-like-a-network

We trained a language model to recommend webpages how people do through hyperlinks. Hyperlinks can be an expressive way to describe a webpage (https://www.ribbonfarm.com/2009/07/01/the-rhetoric-of-the-hy...), but there's a lot of spam and low-quality linking online that would serve as bad training examples. We found the structure of the link graph to still be a fantastic way to understand what content is salient, and used it heavily to filter and build our dataset. The resulting model is best for expressive, exploratory queries where you describe what you are looking for (like a prompt to an LLM) rather than entering in keywords (like a search on Google). This model is the main "curation" step in our system that picks from our index of ~150 million (and growing) webpages.

We built Browser Buddy to try and recreate the feeling of getting a thoughtful recommendation from a smart friend. Our early users have described it as a "refreshing stream of timely and timeless writing", "serendipitous discovery", "rabbit holes that feel joyfully unfunneled". We are iterating on the concept and how it's presented, but we really value the HN community and would love to hear what you all think:

https://apps.apple.com/us/app/browser-buddy/id6752281959

Thank you for your time and being a part of the Internet we love!

Jeremy ([email protected]) and Arnav ([email protected])