This is a really interesting way to look at proxies. I especially liked the idea that a scraper can appear to be “done” while actually returning an incomplete dataset that kind of silent data loss can be much harder to catch than an obvious failure.

It also made me think about the other side of the equation: once all that data exists, how easily can businesses actually be discovered by AI systems? That’s an area I’ve been following through Oglas AI, and the connection between reliable data and AI-powered business discovery is pretty fascinating.