Root Fall
A discovery feed for research I didn't know to look for, and a public shelf for what I keep.
I want to learn more, but most of the time I don’t know what to search for. Search needs a question. So I built something that doesn’t.
Root Fall has two parts. The first is a desk that runs on my Mac. Every morning it pulls papers from OpenAlex around themes I care about: chaos, entropy, memory, meditation, micro ecosystems. A small AI model on the same machine writes one plain sentence about each. I mark what’s interesting and skip what isn’t, and the themes shift with me. The strong ones come up more often, the rest fade, and it suggests new ones along the way. This week it offered cryonics and synchronization.
The second part is rootfall.com: the papers I’ve kept and chosen to share, ordered by how highly I rank them and how often I go back to them.
A few things I learned building it:
- The model writes good plain summaries but sometimes gets a detail wrong. Every summary is labelled as machine-written, and every entry links to the paper.
- Research indexes have bad records. The first feed surfaced a text about AI consciousness and cosmic portals, attached to a real conference paper with fourteen thousand citations. There are filters now.
- Most of what felt like a discovery turned out to be known to someone already. That’s fine. It was new to me.
Nothing here costs money to run. It’s early, and the feed is still learning what I like.