3 ms·

I guess the answer to this is feeding the LLM generated content to another LLM and telling it to summarise it