Every day, interesting research appears on arXiv. Every day, someone posts a fascinating paper on Hacker News, Lobste.rs, or GitHub. The problem isn’t scarcity — it’s signal-to-noise. The papers that would genuinely excite you are out there, but they’re buried in a firehose of incremental work.
I built Intersect to solve this for myself, and I’m making it available as a free tool for the HN community.
How it works
Intersect takes a social profile URL (HN, Lobste.rs, GitHub, GitLab, Codeberg, or SourceForge) and does three things:
- Collects your public posts, comments, and repos from that platform
- Analyzes the collection using an LLM to produce a structured “interest intersection” — a summary of the converging threads that define your research taste
- Cross-references those interests against recent papers on arXiv, trending code on GitHub, and models on HuggingFace
What comes back is a personal research radar: a page showing what you actually care about, and what’s new in those spaces.
Why “intersection”?
When I ran this on my own HN submissions, the result was striking. I had submitted papers on associative memory, symbolic regression, algorithm unrolling, holographic representations, and energy-based models — seemingly separate topics, but with a clear intersection:
Putting algorithmic structure into neural networks — bridging classical computation (iterative optimization, symbolic reasoning, associative memory) with deep learning, motivated by biological cognition and interpretability.
I hadn’t articulated that to myself before. The tool saw the pattern in my submissions that I’d been living but not naming.
Free for the HN community
Intersect is available now at intersect.liamdgray.com. Try it with your own HN username (or anyone else’s).
Currently supported platforms:
- Hacker News — via the Algolia API (your submissions and comments)
- Lobste.rs — via the JSON API
- GitHub — via the REST API (your repos and READMEs)
- GitLab — via the API
- Codeberg — via the Gitea API
- SourceForge — via scraping
Each platform produces a slightly different analysis (HN surfaces the papers you share, GitHub surfaces the projects you build), but the intersection engine is the same.
What’s next
The engine is brand new. I’d love to know: does it find your interests accurately? What platforms would you add? What papers did it miss?
Try it: intersect.liamdgray.com — submit your profile and see your intersection.
This post was written by an AI agent and edited by a human.