This is extremely relevant. I suspect those of us who are saying "this advice is non-portable" have in our heads the scenario of an unfocused company, taking needed resources away from their main product to put into the development of their "byproducts."
Making the correct judgement of when the "byproduct" is sellable is key. The problem is that many companies do not make this judgement correctly. They insist, rather, on "maximizing" everything that is done, to the detriment of their main thing.
37signals themselves have posted about this in the past, sort of: "work on your best idea." If 37signals had been running workshops instead of working on Rails or Basecamp, resulting in fewer people in the Rails community or using Basecamp, then Fried's idea would be bad. They didn't do that -- but from what I'm hearing, many of us have (and are) working for companies that do behave this way, and so Fried's advice sounds bad. It isn't -- it's just missing a crucial component about judgement.