A call for harder AI tests than a “pelicancycle”

One post questions the difficulty of a “pelicancycle” test for a model it describes as “1e27.”