Research Intelligence
AGI Research Intelligence Platform
A team of AI agents that finds new research papers on general AI and scores how much each one matters.
01The problem
Hundreds of AI papers are published every week. Finding the few that move the field towards general intelligence, meaning AI that can handle new problems it was not trained for, takes weeks of reading, and even then you only cover part of what is out there.
02What it does
You give it a research question. The system plans a search, finds the relevant papers on arXiv (the main public archive for AI research), removes duplicates, and scores each paper. It ends with a ranked report of what is worth reading.
03How it works
Each job has its own agent. A planner decides what to search for, a discovery agent finds and cleans up the papers, and an evaluation agent scores them. A supervisor keeps them in order. Every paper is judged against the same ten questions, such as “does this help AI solve problems it has never seen?” or “can it learn from only a few examples?”, so the scores can be compared fairly.
04What came out of it
A literature review in hours instead of weeks, covering several times as many papers as a manual search would.
Built with
Want the technical detail?
The code and the full technical write up are on GitHub.