I mean, it sounds like it's only ideal for cases with significant system prompt overhead. I don't think Jev was built to have a large well described prompt setup. To me its more like a happy go lucky small label classification tool with important decisions left to stronger agentic models or yk humans.
