It would be nice if someone would point out where these companies go to manufacturer their products after getting a prototype up
Good idea...and is one of the biggest challenges in this space. It's great that things like Arduino have made it easier to build something physical and connected, but it does not translate to something that can be manufactured at scale, nor does redoing a design for scale manufacture suffice (also need to get the commit of a manufacturer that can and will work with you). Then you have supply chain, quality control, time, inventory...lots of challenges. Still, all that stuff is getting easier too. Would be a great follow-up post sometime.
Manufacturing is still non-trivial. It's not the kind of things two guys in an apartment can figure out like launching a web app. There are a lot of stakeholders and expertise involved, no matter where you actually go to get it done.
Tell that to the numerous kickstarters, some of which are two guys in a garage figuratively speaking. I might be assuming but it sure comes off that way in quite a number of cases. Not discounting the need of having background knowledge but still
Things like Arduino and 3D-printing are making it easier to prototype actual devices, and imho the effect of these is disruptive as it greatly increases the number of people who can develop ideas and it greatly increases the ability to iterate. Net-net is more innovation. So even if the rest is still hard (design for manufacturing and manufacturing at scale) there's still a disruption going on relative to the way things were a few years ago. Remember that 15 or 16 years ago, if you wanted to sell something online, you needed to buy servers, configure software, scan film and upload image files, etc. It's gotten easier and so will hardware manufacturing, because the demand is there.
This post goes great with your other one on Raspberry Pi. Hardware is definitely becoming cheaper and for folks with time and curiosity, it is a whole new fronteer to hack on. Time is a huge constraint right now (imho), since it's a lot easier to make pictures filtering app than it is to hook up your Roomba to the Pi, but in a few years we will definitely see an abundance of hardware innovations that don't require VC funding.