I think the distinction is that you can learn a lot about a problem-centric business (e.g. typical lean startup) from an unpolished version, whereas the article's author is claiming that games are unique in that you must polish BEFORE you learn, which makes the first leap much more expensive.
Agreed. I was more responding to the parent comment (than the article) re: the difficulty in estimating the time to polish software.
You (and the article) are absolutely correct. An unpolished productivity app can still deliver value if the output is complete, i.e. the majority of the value comes from the output, not the experience of using the product. Whereas games are all about the experience, so unpolished game is not going to provide the value.
