Given that the next release will not be a TX1 64-bit OS you could sill make life a bit easier for your new customers by doing the following. Configure the kernel to support ppdev, and J26 peripheral loadable modules such as SPI_DEV. This would be trivial for you and allow anyone a chance to do something useful without hours of research. It would be great if you could make it easier to find the available GPIO on J26. I know of 18, not including the SPI and UART signals that can be used as outputs. I haven’t yet confirmed which ones can be used as inputs though I suspect the answer is all of them. You could also provide up to date information about how to work around bugs that prevent even using the current tools successfully.
The marketing people a nVidia can’t possibly believe that anyone would plunk down $600+ bucks to see the CUDA demos run ( once they can figure out how to get around the provided bugs… ). As impressive as that is one could see a u-tube video for free for that experience. At least pay someone on staff to install a new host OS and use the tools you provide to find the bugs and make solutions available to your customers.