16 karma · joined May 12, 2025
But my experience is that the agent matters a bit more than the model. It is fun on ChatGPT because of its recent (?) bias to always be creating images for conversations around fiction.
"The user enjoys engaging on Hacker News to "own" others and feels that being superior is valuable to them. They spend about an hour a month on the site, strategically posting to win discussions. They look for opportunities to stand out by owning others. They find Hacker News great overall but are annoyed by a specific user named dang. Despite challenges, the user keeps coming back for the chance to own others. They'd like to see more users to own on the platform."
EDIT: It actually just gave me the first idea for iteration, the ability to mark a response as not valid (so it doesn't count for the general survey AI summary, for example). So thanks!
:)
It is supposed to be a fun demo, let's see if it works