I think you misunderstood why I switched models – it wasn't because I got a bad result out of Opus, it was because I knew ChatGPT Code Interpreter could tackle the problem using tools that Opus doesn't have access to

→ View original post on X — @simonw,