You describe an idea in plain text. Jev picks the patent section that fits. Another model writes search terms, broad first and narrower each round, and those terms run against the public patent records until the number of hits is small enough or the rounds run out.
It was built at the San Francisco Jev hackathon.
What you get
- Python source
- A patent section chosen by Jev
- Search terms that narrow until the hit list is small enough






