The thing about Yahoo is that it's massively silo-ed. There's research teams doing fascinating algorithms and machine learning work. There's small hacker cores working on new crazy ideas and technology. There's also a ton of internal politics, NIH across divisions, and meaningless obstacles to getting stuff done.

As an example, I was at Yahoo for 5 years (2003-2008) working in the Search Marketing division. Tons of data, tons of optimization possibilities. Could we use the Hadoop cluster? No, for some foolish internal accounting reasons.