THis is very similar to what Oxide.computer has done. They wrote their own tooling to go from Rust API to spec to Rust and TS client. Very specific to their use case, but I think its the right approach. If something is wrong in the client, you probably generated a bad spec, and therefore something in the API is rotten and should be adjusted. Using the spec as an intermediate to know if upstream is bad, can be a great benchmark for developers. My co-founder and i have a long term goal of building an end to end tooling chain for OpenAPI to generate spec and clients/tests/docs under one roof. Its definitely far off in development, but think its where devs are moving