As I said in the article, one interesting aspect of Jev is the consistent property. That's really useful for building reliable judges for agentic workflows. I am now working on a little benchmark to test this more broadly and compare it with other decision APIs. More soon.