I find the WebCriteria approach interesting but I’m still suspicious of it. The objective measures like load times are useful, but I’m not convinced an agent can replace usability testing, though those aren’t perfect either.

It’ll be interesting to see if in the future they spawn other agents in addition to “Max”, each modeled after particular types of users. Whereas Max tries to be Everyman and Everywoman, it might be better to have a stable of agents representing different people. Or a customizable agent you can plug demographics into.