A digital Zen master would say that if everything is starred, nothing is.
We’ve optimized the system for getting things in, but how do we get something good out? How can we make meaningful connections between all of this stuff, and make constellations out of all these stars?
— Frank Chimero
Does collecting things matter if we don’t revisit them?
Frank Chimero's dconstruct 2011 talk would whisper—sometimes yell—from the back of my head for over a decade.
It's what I thought about every time my phone got just a little smarter every year, every time I re-stumbled upon an article or talk that profoundly shaped my worldview, so much so that I forgot the source. And it's all I could think about the first time I met Alice for coffee to discuss her new venture re:collect.
2021 was still early enough that I didn't forsee the generative AI storm brewing but I did know machine learning had to play a significant role in solving the rediscovery problem. I build tools professionally, but I lacked a mental model for how machines learn. Wanting to remedy this, I accepted Alice's challenge and joined the founding team. As the only designer and front-end engineer on the team I owned the web application and browser extension presented below.
The browser extension watches over your shoulder and automatically collects content you pay attention to. Add additional context through notes and highlights to engage with content on a deeper level and leave a trail for your future self.
Technically speaking: this was my first experience writing a browser extension and it was a very positive one. The UI layer is a modern React application that shares a lot of code with the main web client.
The ambition was to deeply index all the content you consume and create, making it effortlessly rediscoverable. It currently supports web articles, PDFs, YouTube transcripts, Tweets, and automatically pulls from sources like RSS, Twitter bookmarks, Readwise, Apple Notes, and Google Drive documents, with more integrations planned.
Technically speaking: due to the small team size I also designed and maintained a lot of the Python API layer the web clients directly interfaced with and was involved in the system architecture decisions as it evolved.
Recall brings back content connected to what you're thinking about as collections of related cards. Pick between tranditional search when looking for something specific or take the scenic route and explore related material, hopefully rediscovering something even more interesting along the way!
Playgrounds allow you to think out loud on an infinite canvas. Keep and make notes on highlights as you Recall what you already know about a particular topic. Spread it out, and rearrange the pieces as you explore the connections. Finally, impose order, hierarchy, and linearity as your ideas clarify.
Technically speaking: there were many fun challenges in building Playgrounds, from the infinite canvas to the card layout engine, rich text editors (Slate.js), multi-level undo system and sync engine to name a few.
I'm grateful for the opportunity to scratch a lot of personal itches and toy with ideas that took up space in my head rent free for over a decade in this role. While the product has since been shut down, the rest of the team has joined SmartNews and continues to innovate in new directions.