Interesting that all four models converge on such similar designs, for such short prompts.
They were trained pretty much on the same data.
alt Hacker News
They were trained pretty much on the same data.