It was fashionable about 20-40 years ago to think the Bourbaki school was best, but the modern view of *teaching* mathematics is that it is better to prove it for the case "2" then to show it for the case "2=n".
Personally, I've never been very interested, explicitly, in the Bourbaki School, but it is hard to judge the implicit effect of it in my work. Often, unless one understands a particular case very well, and what it is about that particular case that allows one to make the deductions one does, and unless you have a good intuition as to what can be omitted from the hypothesis, then Bourbaki isn't very useful. If, however, you have a sound grasp of the fact all that mattered was, say, the operation was binary, rather than associative, then thinking a la Bourbaki can help you translate results into other areas.
I would rather teach particular cases first but very quickly and extract to the general, rather than prove a general theorem. It is only by studying the particular that we understand the general, perhaps is the anti-bourbaki line of reasoning and one that is necessary after a certain level.