It is my hope that ATProto, with the PDS / portable data / app & algo choice, means that this "Cognitive Moat" discussed in the post, is less of a moat. There are some people working towards building an API that would facilitate (and lower the barrier) to having good recommendations for a new user based on what they bring from the last app (really the user's data, so bringing is not exactly right)
It's still a big lift to make good recommenders and we should lower that barrier by (1) taking back our data from their silos and (2) building the tooling to make things easier for app builders
Then there is the way ATProto breaks down the system and allows you to plug-n-play algos into apps. Separating the algos from the apps means you can just sell algos (recommendations, feeds, moderation) in an open ecosystem
It is my hope that ATProto, with the PDS / portable data / app & algo choice, means that this "Cognitive Moat" discussed in the post, is less of a moat. There are some people working towards building an API that would facilitate (and lower the barrier) to having good recommendations for a new user based on what they bring from the last app (really the user's data, so bringing is not exactly right)
It's still a big lift to make good recommenders and we should lower that barrier by (1) taking back our data from their silos and (2) building the tooling to make things easier for app builders
Then there is the way ATProto breaks down the system and allows you to plug-n-play algos into apps. Separating the algos from the apps means you can just sell algos (recommendations, feeds, moderation) in an open ecosystem
https://atproto.com/articles/atproto-for-distsys-engineers