It helps having this article next to (nearby) the HN article discussing the 80K static coding violations in Toyota throttle position sensing code, or whatever it was exactly.

THIS article didn't get into it very much, but "obviously" you'll have four times the bugs in a highly abstracted 40 part piece of code, than in a much simpler, less capable 10 part piece of code.

Reality is bug count never scales at a rate as low as linear with code length of course.

At any rate the two HN articles are both better if read at the same time.