This mirrors my opinion about ORM quite well, but it's missing an important angle that might help understanding why ORM made it that far in the first place:
The big selling point, in hindsight, was running document-style persistence on relational. Today we happily dump JSON blobs in there and we accept that we'd have to deal with the consequences if it ever came to querying on some oddball property (chances are consequences won't even be so bad), but back then it was always relational first, and relational everything. All that busywork of spreading things like the change history of the user's favorite ice cream out to glorious relationality, stuff that will never ever be accessed outside its natural tree structure ("the document")? ORM were excellent at that.
Now that we have arrived at something that I'd consider close enough to consensus that some data is fine too keep in document blobs (instead of spreading out), the value proposition of ORM is much, much smaller than it used to be.
The big selling point, in hindsight, was running document-style persistence on relational. Today we happily dump JSON blobs in there and we accept that we'd have to deal with the consequences if it ever came to querying on some oddball property (chances are consequences won't even be so bad), but back then it was always relational first, and relational everything. All that busywork of spreading things like the change history of the user's favorite ice cream out to glorious relationality, stuff that will never ever be accessed outside its natural tree structure ("the document")? ORM were excellent at that.
Now that we have arrived at something that I'd consider close enough to consensus that some data is fine too keep in document blobs (instead of spreading out), the value proposition of ORM is much, much smaller than it used to be.