The model choice note that turns into a routing rule

A thread-map prompt for turning repeated Claude/GPT comparisons into reusable routing rules.

If three people keep arguing about which assistant is better, the useful output might be a routing rule, not another comparison thread. Example: - use one assistant for long reasoning drafts - use another for tool-heavy local edits - use a cheaper one for first-pass summaries - revisit the rule after pricing or context limits change The important part is to keep the reason attached. “Use X for coding” is too vague. “Use X for small patches when it reads local context and keeps the diff narrow” is something a team can actually test.