« Bits vs. Paper | Main | Fame Comes to Hub Kittle »

February 20, 2007

"Google 101" class at UW

Interesting article in the paper about a Google developer who created a programming class at the University of Washington. He evidently did it in his 10 percent time (isn't it 20 percent? Was that cut down at some point?). It was a 5-week course for 15 students.

The article promises that "The class is aimed at creating programming prodigies and revamping the way colleges teach computer science". But on reading more, it sounds like they are just teaching the students how to distribute computing across a bunch of machines. Certainly a worthwhile addition to the curriculum, but not earth-shattering. To their credit, Google appears to be doing this purely for the students' benefit, not as a recruiting tactic, at least not a direct one.

Posted by AdamBa at February 20, 2007 11:20 PM

Trackback Pings

TrackBack URL for this entry:


No, it's still 20% time -- that's a weird typo to make twice, though I suspect I know the reason. Google bins its projects into 70%, 20% and 10% buckets: 70% is the core business, 20% are related exploratory projects with a visible route towards becoming core business, and 10% are "out there" projects which might be tangential or wholly unrelated to the core. I guess this project would be in the 10% bin.

Posted by: Moishe Lettvin at February 21, 2007 06:42 AM

OK, that makes sense. Actually I had wondered how Google decided if a 20% project was acceptable (or if there was any approval process); having the 20% focused on something that looked productizable and 10% on anything you wanted makes more sense.

Could you write a book in the 20% time? What about the 10%?


- adam

Posted by: Adam Barr at February 21, 2007 07:00 AM

Sorry, I was unclear.

Engineer's time is split 80/20 -- your 20% time is completely discretionary (within reasonable bounds), so you could use that time to start a new project, teach a class, mentor, etc. My guess is that writing a book would be fine.

Projects are bucketed 70/20/10 -- this is orthogonal to the engineer's 80/20 time split. A project which grows out of 20% time could be placed in any of those 3 buckets.

Posted by: Moishe Lettvin at February 21, 2007 09:23 AM

I read the downloadable materials on my last plane flight. It looked interesting -- and we desperately need to start thinking more about what we can do with large distriubted systems that are full of faults.

I teach at the UW (physics) and looking at the course materials I would not have given this course too much in the way of credits. Perhaps the exercises were tough (I'm afraid I didn't look at them carefully). But this mostly looked like a seminar and a way to be introduced to map-reduce. Teachings on map/reduce, the theory, and GFS are great -- we need more of that. But courses aren't meant to be vocational; ideally you'd like to give the students enough background that when they get out there they can invent the next version of these algorithms.

At any rate, the last 10 years have seen a huge amount of research go on at various companies that academics haven't been part of. It is great to see a path to feed that information back in. Being a member of the university community, however, I'm always nervious when any company sponsors something so directly.

Posted by: Gordon Watts at February 24, 2007 02:53 PM