Very true. One distinction- my article is about training AI to be able to retrieve Twitter Data, not training on Twitter Data. Very different problems. Grok seems to be oriented towards the latter