Jon Pipitone, a grad student here at the University of Toronto, has posted a question on his blog: what would it take to convince you (a software engineer) that a piece of complex climate modeling software was working correctly? It's a hard question --- he'd welcome your suggestions.