[esjeon]

I’ve been trying this on some social and political science topics, and, tbh, Opus is never going to replace ChatGPT in terms of sharpness of claims. Claims generated by Opus alone are almost always dull and inadequately nuanced. It’s also heavily influenced by popular opinions that include common misunderstanding, preconception, and vague language. The context gets contaminated too easily by external texts, so I have to carefully control and craft the input to Opus.

So my pipeline is mostly stuck with: (1) brainstorm with Gemini (2) plan with GPT (3) augment the plan with Gemini (4) execute the plan with Opus.

Perhaps, given all the materials filtered and prepared, Opus is good for verification.