how many programmer hours has gone into your algorithms ?

For google, according to Udi Mander on their blog, its supposedly 1000+ man years invested in the page rank algorithm : "By some estimate, more than one thousand programmer/scientist years have gone directly into their development, and the rate of innovation has not slowed down." On the flip side, you have google's other half YouTube constantly rebuilding. From Venublog's coverage of the case studies in the recent : Notes from Scaling MySQL - Up or Out, Youtube's rebuilding was quoted as - "Pretty continual, 2-3 times (replication, sharding, federation.)"

So I wonder what counts as starting again, and resetting the "coding hours" invested in a platform/website/algorithm to zero ? And what counts as competitive advantage in so far as new entrants would need to match/beat the amount of time invested. Sometimes also your technology becomes a legacy putting you at minus zero. I'm pondering such a question today as a constant investor question is "How long would it take for a competitor to replicate what you have built ?" (or the alternate : "If you were to build this again, using a 2008 framework, how long would it take you and what would you need?") Aaaagh life's questions.. if only the answer was "Give me the money, and we'll find out what the answer is...."

Cautionary Build it and They Will Come Tale of Meetro over on Techcrunch : "I can still feel the magic of when I was on layover in the Denver Airport, I met one of our users, and we grabbed a beer. This is what I dreamed Meetro would be about all the time, but those moments were too few and far between. We did fix this in the end but it was too little too late."