The Capture Prompt

Researchers found that two humans achieved more than 99% accuracy, reading these words versus 80 or lower for the software. The guillotines in this one just cut up a nice big broad worst for everyone to enjoy. So then the capture researchers said, let's remove the broad worst. And I said, no, but they removed it because capture researchers did preliminary tests where they took just the hard to decode words from texts that was faded or messy or scanned funny.