really interesting. a lot turns on whether the personalization is actually substantive. if it’s tirz at scale w/personalized layer on top, 503a seems shaky. also curious how the patent side plays out, since 503a doesn’t solve that. very interested to see where this lands.