Method
The LLM Council method
The LLM Council method is compare, critique, rank, and synthesize: a repeatable process for turning multiple model opinions into one stronger answer.
1. Understand the job
The first step is intent matching. A greeting, static fact, current-news question, document edit, pricing decision, or legal-style review should not all use the same path.
LLM Council starts by deciding what the answer must achieve, what context is needed, and how much depth the user selected.
2. Build the right context
The system can use uploaded files, selected library sources, web research, prior corrections, and the current chat. The goal is not to stuff everything into the prompt. The goal is to feed the council the right evidence for the user's intent.
3. Seat the council
Different effort levels seat different council sizes. Different plans unlock different model pools. The method stays consistent; the subscription changes the model quality and context ceiling available for harder work.
4. Preserve dissent
A council is valuable because disagreement is visible. The final answer should not erase the strongest counterargument. It should tell the user the verdict, the main dissent, and the condition that would change the answer.