It wasn't me. Nice approach, though. Just intersecting word choices has very little to recommend it for industrial scale author identification but for a small-ish community like HN it might work, and of course it is trivial to implement if you already have the data source lying around.