[Wheatman]

Well, apparently it has ended up being true.

The orion model underperformed expectations by a lot, and even fell beh8nd chatgpt-4 in programing tasks.