Reading this article confirms my tremendous respect for Google for:

1) Creating a culture where writing unit tests, fixing bugs and the like are not viewed dirty work or an overhead, something that scientists (or even engineers) are above

2) At the mean time, being a place where creation of new technology (even if not strictly related to the primary product i.e., not just information retrieval algorithms) is not viewed as "academic" or a risk to be avoided at all costs.

There are companies which do (1) well and companies that do (2) well; they have created wealth and made products I enjoy and use. Only a handful managed to do both: they changed far more than their industry.

The danger with integrating research as part of day-to-day development is that you might not end up with groundbreaking or disruptive innovation. The problem is that forces which are controlling day-to-day work are essentially controlled by value network of the company (social and technical resources within and between businesses) and constrained by resources, values, and processes of the company. So, there is a danger of missing inventions which will replace the current (profitable) way do doing things (i.e., start Groupon kind of business model, or Twitter, or Facebook).

However, I also 100% agree that having pure "research labs" is not effective: in many cases some great idea from research arm are just forgotten.

Clayton Christensen wrote a few books on this topic (and I am a very big fan of Clayton Christensen's work).

I would like to also mention that Google is not the only company which has this approach: it seems to me that majority of companies in Silicon Valley operate like this (I don't think Facebook has "research labs", but they do invent things).

Two disconnected thoughts that came to mind reading this...

1. I agree with Google's approach. Microsoft spends $10 billion a year on R&D yet I can't think of a revolutionary product they've come up with (Kinect was bought from the outside)

2. I notice the word "patent" is never mentioned in this piece. A big reason companies like Microsoft and Intel have research labs is to patent protect themselves in the future (wasn't there an article on HN just a few days ago that mentioned Google's anemic patent portfolio?)

MS's approach to research is fundamentally different than Google's. Google's approach is about creating better Google products. Microsoft's approach (which is really Bill Gates's dream) is to advance science.

Both are good goals. But if you ask a researcher from MS how important it is to get in a future version of Windows, they may well shrug their shoulders. Their feeling is if they can advance technology then all boats rise, which is good for Microsoft.

I must say when I'm actually doing research, and not just dev work for product Version Next, I don't want to do unit tests and a whole bunch of other cruft. My code isn't meant to be shipped to customers. Its meant to test a hypothesis.

I read repeatedly that much of Kinect's software was derived from their in-house R&D. Is that not true?

I agree mostly, though your assessment of Microsoft Research is not completely true. Don't forget they came up with the Songsmith :D

Isn't what Google practicing a kind of applied research? Where MS Research is doing basic research which advances the science.

As were Google earth, gmail, google docs, wave etc