That's not how LLMs work. These systems take a stream of tokens, do some linear algebra to find a stream of tokens within the parameters of their vector similarity, and spit out the result.