…but maybe you have.
If so, click over to Science-Based Medicine for Dr. Harriet Hall’s thoughts on them and Staffan Lindeberg’s seminal nutrition study. This is her second recent post on ancestral diets (aka paleo). A snippet:
I am always suspicious of initial reports of unusually healthy or long-lived groups in remote areas, because I have so often seen such reports disconfirmed by subsequent investigations. Lindeberg’s studies were done in the early 90’s and have not been confirmed by other studies in the ensuing two decades. In the Kitava study, the ages of subjects were not objectively verifiable, but were estimated from whether or not they remembered significant historical events. The absence of heart disease and stroke was deduced by asking islanders if they had never known anyone who had the symptoms of either condition. This was reinforced by anecdotal reports from doctors who said that they didn’t see those diseases in islanders. EKGs were done on the Kitavans, but a normal EKG does not rule out atherosclerosis or cardiovascular disease. I’m not convinced that we have enough solid data to rule out the presence of cardiovascular disease or other so-called “diseases of civilization” in that population.
You can guess where this is going.
Dr. Stephan Guyenet chimes in with cogent comments.
Paleo Diabetic 
AND there are so many variables with this sort of thing even if it is verifiable. What do you do with this? Eat what they do? Mimic their physical and social activity? What if its just the water they drink or their evolutionary coronary artery dimensions?
It would not be totally insane for a type 2 diabetic to limit protein
to avoid high fasting blood sugar from gluconeogenesis as well as
limit fat to maximize insulin sensitivity, and do away with grains and
omega-6 oils and also have a little fish or fish oil.
I wonder why no one uses it in order to discuss possible benefits of smoking.
Kitavans don’t seem to get Parkinson’s disease, but no one is going to
be able to prove they are protected from it by smoking even though the
correlation has been statistically recognized.