[borg16]

the results with grok-1 were unimpressive summaries based on the tweets, with a 10%-20% hallucinations (when enquiring about paris olympics specific events).

yet to see if this new model is able to do any better on that regard

[pantsforbirds]

I was also not very impressed, but I still think they are positioned to have a great product if they can get past the accuracy issues.