I'm wondering if he could just adjust the API to give an error message that an LLM (or a person) could understand and then correct. It might be worth putting so

I'm wondering if he could just adjust the API to give an error message that an LLM (or a person) could understand and then correct. It might be worth putting some effort into beefing that up, to cut down the support requests.