As far as actual learning goes, educational institutions are great in cases where the knowledge is hard to access, expensive or dangerous. If you want to be a doctor, you can't just go to and get a cadaver or an MRI machine to practice on, these are expensive. If you want to learn how to run a nuclear power plant, you can't go buy radioactive material to learn with, there are national security concerns with that. Learning how to fly a plane is dangerous and planes are expensive. But educational institutions can address all of these issues.
Using the term bubble is perhaps an effort to sensationalize the issue, I'm not sure. In any case, I think it is certainly useful to look at fields where access to knowledge is easy, cheap and safe and ask if an institution is required for learning or if that's wrong thinking. It happens that writing computer programs is one of those cases. This is probably why all of these articles talk about software developers (them being posted on HN by programmings probably has something to do with it too :)).
Saying that a system produces useful results is not that same as saying that a system is optimal. Cars in the 1950s could get you places a lot faster than walking, but that doesn't mean they got the best possible gas mileage. Computers in the 1980s did amazing things, things that would be at least impractical without them, but 1980s computers could've been a lot better, orders of magnitude better. So, stating that educational institutions have useful results is not necessarily a counterargument to the deficiencies in educational institutions. In fact, I don't recall reading any article in this recent string of education articles that claims that the problem with the higher education system is that it doesn't produce useful results. In fact such a statement is easily and, probably, objectively, verifiably falsifiable. So this isn't a question of useful vs non-useful. This is a question of optimal vs non-optimal.