So basically, @anderssinho, a bot that automatically

1. Reads threads
2. Runs the thread through an llm with a special prompt with instructions to generate insightful questions about the thread.
3. Posts the insightful questions as a reply to the thread

That's a cool idea! And shouldn't be neither too complicated to create or costly to run. The llm cost would actually probably be so cheap that even if the bot posted a thousand questions a day, it probably wouldn't spend more than a few cents worth of tokens (a small model would suffice plenty for this, like gpt-4o-mini or claude-3.5-haiku)

I'd be happy to help you get started if you want to try and build it yourself.