@thomasfuchs come to think of it LLMs are the ”read equivalent” of “Eventual consistency”… which was a shitty way to store data anyway!