Can we get a proof of work spam protection to email?
It’s a neat idea in theory. But in practice, e-mail filtering works better as an exercise in content and sender categorization. The best systems use a multitude of techniques to generate features and then use machine learning to determine the high-dimensional space that contains spam messages. This space can vary for each recipient.
Google’s lead antispam ops guy told me at a conference a few years ago that moving to this approach dramatically improved their spam filtering accuracy. But some amount of manual feature engineering is necessary, such as looking at the sender’s domain name and assessing things like domain age, registrar reputation, etc.