REST APIs are deceptively simple on the surface but notoriously inconsistent in practice — different teams implement auth, pagination, and error handling completely differently. AI tools struggle because there's no single 'correct' REST — every API is its own dialect. GraphQL and gRPC have stricter contracts which makes them easier for AI to reason about. The irony is that this inconsistency is also why REST APIs have so many security vulnerabilities — no standard means no standard security either.

What model are you using? I don't think I've encountered it before. But then again, as long as you explicitly instruct an AI to follow REST, it usually will.

REST is underspecified — every codebase does it differently. The model learned that POST-everything works everywhere, so it defaults to that. Adding a concrete example endpoint to your context works better than just saying "use REST."

I have been using VScode with GitHub AI assistant and I asked the AI many times to set up the REST API it did work

What model are you using?