The 23andme dataset provides more value than HGP since it's an dataset of active participants. I don't know too much about the specifics or privacy mechanics, so I can't comment on how, but this article is an example (http://www.mercurynews.com/science/ci_27487046/23andme-quest...). The way researchers are using the dataset is very different because they are able to do followups. That's a very significant difference, it's why they have pharma companies lining up to get access to the dataset.

Last I heard, they were closer to 1M than 850k, unfortunately can't find any live stats. Maybe someone else can chime in?

I don't have any affiliation, and I'm surprised that you'd assume so. I recognize that there's an overeagerness from both startup people and academics to crap on biotech startups, rightly or wrongly. I haven't quite figured out why spectators are motivated to do so.

* The human genome project created a "single" genome (thats not quite right, but close enough for this discussion). This was and is incredibly useful, but a single genome can only do so much. To answer population genomics questions (e.g. is this mutation common and therefore probably not disease causative) you need lots of people and the more the better.

* The 23andme dataset is not whole genome data, but instead some subset of genomic locations that they consider more interesting. Still valuable but not as much as whole genome data.

Also, for big whole genome datasets you can check out the 1000 Genomes Project, which has somewhere north of 2000 genomes. And there are many more big datasets in the works, which will continue to be incredibly useful!