hey @extenebrislucet! thanks for the feedback. The metric we used for validation was citation count ("breakthrough" = within 24 months an article in top 15% by citation count is published within a given radius epsilon of the prediction previously made). There are definitely some surprising predictions that don't belong to steadily advancing fields. The most impressive is that we predicted the importance of llm code generation back in 2023, when Cursor was just starting out. There are several videos about this prediction at youtube.com/@innovationlens There is a live free demo at innovationlens.org, with restricted access. For a small fee you can access fresher data too.
Give it a try, I'd love to hear your feedback!