MapReduce was published in 2004. The paper was two people; the implementation was probably a lot more. Even then, it was just a re-engineering of an earlier cod

MapReduce was published in 2004. The paper was two people; the implementation was probably a lot more. Even then, it was just a re-engineering of an earlier code base (read that as multiple available cat skinning technologies). By 2014, it was no longer used, so that's 10 years. Lastly, it wasn't exactly novel. https://en.wikipedia.org/wiki/MapReduce#Lack_of_novelty https://en.wikipedia.org/wiki/MapReduce#Lack_of_novelty FWIW, it appears they're giving away the TensorFlow stuff because they lost the mindshare battle between MapReduce+Hadoop and don't want that to happen again. Again, the $100M figure is an easy (heroic) claim to make but hard to substantiate. Shit's hard. Big shit is really hard.