Frankly surprising. As pessimistic as I generally am, I at least assumed that the models could handle that prompt. However, worse than I would have thought.
Except for QWEN 3.5, maybe they trained that excessively on HTML clock software? Other models sometimes hit the mark or close enough, but QWEN 3.5 had it nailed.
Funny when I thought maybe it took liberties and made a numberless clock… Then I realize it just missed the clockface so badly that you could just barely make out a 5 in it’s upper left…
Frankly surprising. As pessimistic as I generally am, I at least assumed that the models could handle that prompt. However, worse than I would have thought.
Except for QWEN 3.5, maybe they trained that excessively on HTML clock software? Other models sometimes hit the mark or close enough, but QWEN 3.5 had it nailed.
Funny when I thought maybe it took liberties and made a numberless clock… Then I realize it just missed the clockface so badly that you could just barely make out a 5 in it’s upper left…
I hate AI just as the next one here, but this is pretty funny. They got better than the first time I saw it.