The idea is that parallel thinking is easier and faster than to align everything in a time series
So the user inputs two prompts running on parallel threads on the same ollama ai server where the ai can pick up smart results from one thread and combine it to the other. For brute force type of reasoning where you have “target prompt + any prompt” as input.
Opt in feature: HuggingChat could become much smarter the more active users it had if it combined results from different running prompts.