Just took a look and we're back on the leaderboard, truly a cause for celebration. Let's keep working hard. I just saw a new tweet from @inference_labs. I have a new perspective: the key to zkML deployment is not "proving the entire model," but smartly proving only the necessary parts. Full model verification? Too slow, too expensive, and not scalable—out of the picture. Modular design + distributed proofing is the way to go.