Pretty neat, and I'm certainly all for gaining new knowledge, but couldn't the author just have decompiled the .NET binary (using a tool such as JetBrain's dotPeek) to see what it was doing without having to sign the NDA?
I think the authors tack on this was faster than decompiling the binary. Many things are pretty straightforward mucking about in binaries, but I'd bet that USB transactions are not one of them. Combine that with this being an easy-to-sniff over-the-wire protocol, protocol analysis is how I would have done it too.
On the other hand, I thought tessel was an open source project? Why not just hack together your own DFU loader? It basically just needs to read data over USB from the PC, and write it to nonvolatile/flash.
Depends what are you more used to... I usually look at the data transmission for obvious solutions before decompiling stuff. Since he's dealing with hardware, he's probably also likely to be used to making sense of long data streams looking for issues :)
