GLM 5.1: Pelican Test
simonwillison.net> Update: On Bluesky @charles.capps.me suggested a "NORTH VIRGINIA OPOSSUM ON AN E-SCOOTER" and...
This is the perfect detraction for all the "omg, this benchmark will be trained on". Well, yeah, but then you can change it. And ask for a noun verbing on noun. And ... it still works!
I've been trying that prompt agains other leading models and honestly GLM-5.1's is by far the best.