OP here, thanks for trying out. I've added a tuning slider at the front-page so yo can go set the pace, going a bit conservative like those human folks or go wild :)
Hi, thanks for trying out. OpenAI just limit the whole prompt to 4096 token (that's around 2,000 words), so I have to limit each answer. I tried hacking by adding summarization for the first part of the convo, then chat along as normal, but the result wasn't that great. Will tweak more.
OP here, thanks for your feedback. In your case it seems to be too aggressive. When it's aggressive, it tries to come to conclusion what courses/blogs/tutes you should learn/follow.
I set temperature like a bell curve with peak temperature (randomness) at question 10th. Maybe I'll tune it down a bit, get it in the mood first.
On Linux VM, if a program is given root privilege and it just execute whatever command ChatGPT spit out, then you'll got hell a problem solver/havoc wreaker.