quasi talon
Not to get too negative here, but I just spent ~$30 USD (or more, given that the usage counter comes with a latency) in the web-based API playground for assistants while diagnosing a complicated full-stack development problem, since GPT-4 is too overloaded to run anything sensible right now.
The API Assistant froze mid-session, wouldn't snap back after multiple attempts of trying to add and run an input query; and the session froze from OpenAI's end. All the session data was gone, which would mean restarting more or less from scratch.
Other users have been reporting the same issue.
Now I'm at a point where I'd have to pay the AI bandit another round of money to get to the same point as where I was?
I hope this isn't an integral part of your current profit model -- I don't know what else to say about this ordeal.
Best regards,
🐴