I hope they continue down the "change the world" path by continuing to look at how their expertise in data analysis and AI can do more (like self-driving cars).
I'm really excited to see what he does.
You seem to be making it more extreme with people losing their jobs on failure. One of the advantages of having a big company is that you get to attract good people that are more risk adverse and to assign good people to projects more easily. If you let go of that I'm not sure what's the difference between what you described and a VC fund.
What might be possible is some type of Skunk Works with independent leadership, which generates new ideas but leaves the fleshing out to the rest of the company. It wouldn't be associated with longterm management of one project though, more as an incubator and consultancy that spins off and checks in on projects they initiated.
Is this a deliberate attempt to troll or are designers nothing but overhead? Poor design, particularly poor interface design, is considered a major factor in many of Google's product failures (Wave, Google TV, Buzz, et al). User interface design and consistency are frequently cited as the largest remaining gap between the Android and iOS experiences.
And in fairness, Wave and Android are hardly perfect examples of engineering. Wave was buggy and never added significant features, Android is still sluggish compared to iOS on similar hardware.
Google tends to lack clear direction/leadership/vision at a product level, not just UI design.
Also, as other comments have pointed out, a lot of Google products just won't survive unless they are backed by the revenue coming from Google Search.
Larry would act as the investor, and anyone inside the company could come ask for money to found a new "company" or for additional funding. Every company would start out with some kind of "shares" to be distributed by the founders (with vesting, etc). If a company fails the people would actually lose their jobs. If they succeed the equity would be worth some proportionally huge amount.
I don't think any company has ever tried to truly simulate startups on a scale like this while maintaining the upside and downside of a real startup.
It's just crazy enough to work and he's just crazy enough to try it.