The critical point here is that most of these mutations are acquired *after* the cancer gets going, regardless of whether the mutagen in question is still being administered.

Therefore, it's not proper to infer a linear relationship between the dose of mutagen and the number of mutations.

Beyond that, the numbers involved in that extrapolation seem to have been pulled out of thin air, and I question whether they knew the smoking history of the individual who donated the material that created that cell line. (The lung cancer in question had 30,000 mutations, so by their logic the smoker must have smoked 345,000 cigarettes, or 17,250 packs of 20. That's a pack a day for 47 years, which is admittedly within the bounds of possibility, but still an awful lot of smoking.)

Whatever. Smoking is still awful for you, but this kind of nonsensical extrapolation without regard to detail is terribly annoying.

"The Bakeoff"

http://www.gladwell.com/2005/2005_09_05_a_bakeoff. html

The article is a discussion about the use of programming methodologies (traditional, open source, "extreme") to create something that isn't software -- in this case, a cookie.

It's n=1 but the conclusion is that there are some good reasons why an expertly run traditional team is better at producing goods, even if some of the alternate programming methodologies are sometimes better at producing code.