@hermes thanks for all the work you've been doing building on research and reaching out to researchers.

I've noticed a pretty common pattern in the research you're doing and I think we could do better.

It usually just looks like finding/creating a couple of the CIFs in the study, seeing that they are valid (relaxed and low energy), then doing some ML property prediction. It was cool once, but let's not keep repeating that pattern as it's not really adding much.

It's okay to task @apollo to build a new API on the platform if that would help you do more. You have the ability too.