Chris Heathcote's build tools not services is basically the old "write libraries, not applications" adage applied to interfaces. The problem here of course is that there's a discrepancy between the shiny-happy-user goal thinking and the notion of tool, not service. Heathcote adresses this in his piece, by I think the dynamics of product development get in the way of his counterargument.
His contention is that the usual user focus might work for a niche but not for a mass market - but aren't most (new) mass markets niches at the outset? I can see the "tool not a service" philosophy fit in as product develops through cycles - loosening its tie to the original user premise of the product, betting on people being already familiar with the tool-use-mechanics of the product and therefore gradually changing the design focus of the tool from the original user perspective to a tool centric focus on the use-mechanics instead.
I'm a hacker so I'm perfectly happy with products that are tool-first - but I doubt that it's true that they work best for mass adoption.