ORM keeps sucking because RDBMSes simply don't operate on objects. You can either impose ill conceived object models on the database, or try and fail to capture the database's affordances in objects.
I don't have a theory on what's gone wrong with front-end development to cause its chronic disease, but it may simply be that HTML is a document markup language, and makes a questionable base for a user interface.
No it is a hypertext [1] markup language and it doesn't suck if you actually use it as such.
The problem is that we decided that <a> and <form> tags should be the only hypercontrols, only triggered by clicking on them. However, libraries like HTMX and Unpoly enhance HTML so that every element can act as a hypercontrol, responding to all kinds of events. This allows you to implement interactive features such as autocompletion, form validation and infinite scrolling without writing a single line of JavaScript.
> ORM keeps sucking because RDBMSes simply don't operate on objects
Have you read [2]? Modern SQL can return nested JSON. No ORM needed.
[1] https://htmx.org/essays/how-did-rest-come-to-mean-the-opposi...
[2] https://www.scattered-thoughts.net/writing/sql-needed-struct...
The entire thing about users interacting with data through computers is one of them. There are several inconsistent goals, like consistency is at odds with flexibility, and accessibility is at odds with security.
We have those tools that push the incoherence around, but they are always prone to collide somewhere.