We adopted GraphQL from the get-go with our startup [0] and I won't say that we're worse off per se, but I will say that I don't think there's a huge benefit for our core application from it.
We ended up going offline-first (or rather we're working on being offline-first) so the application roughly looks like: download all the data, store it locally, operate on it locally, sync changes back and forth. For that, GraphQL is not buying us a whole lot. In a few places it even gets in our way (for example, there are currently not union types for inputs so we end up having a more verbose than necessary schema). We've also battled the famous N+1 problems and have had some performance problems with the client side cache performance (we use Apollo). There are a few niceties when it comes to automatically checking the validity of requests and some code generation we've been able to do with Typescript that's pretty nice.
Most importantly, I still like it for our 3rd party API - we're an issue tracker/product management tool and GraphQL fits quite nicely for exploring our data and for third parties to just fetch what they need.
