FOSS Projects: E-Mail needs a replacement. Start building one. Encryption and anonymity as core of the specs. Build Branding, marketing, professional UX and proper Clients for all Plattforms. Yes, including Apple. Lets get going with this overdue problem.
We need a feasible distributed Facebook Killer. Diaspora is Meh, with shitty branding and UX and others are even worse.
These are the wrong kinds of projects for FSF to sponsor. There are armies of hipstercoders with well-trimmed beards and MacBooks in cafes attempting to tackle this problem to improve their Personal Brand. By this I mean, they are getting high attention in the contemporary world, and don't need attention-boosting by the long-sighted greybeards of FSF. They also have a low multiplier, in terms of code written and software freedom gained, since they're meant to be used by regular people and not programmers. They get more absolute use than the proposals below, but are not FSF work, just independently important work.
git is an example of a project FSF might have sponsored. People getting paid to write software who want to upstream their changes have lower barrier-to-entry through not just the convenience but the process that grows around git, so it's created a culture of upstreaming among corporate Linux users that used to fork and forget. Code review tools to replace the adhoc code review happening on the sub-mailinglists where most Linux kernel code is traded could be helpful, especially if it fits the processes of the companies actually writing the code, sign-off of ownership or whatever else they need, though they may already be happy. This would be more interesting to work on in the 90s, but is more solved now, maybe.
Another might be, a lighter-weight less-finished version of Android for use in televisions, printers, "internet of shit" devices, that communicate through passive-matrix LCDs like refrigerators or overlays like televisions. If it were rigidly-open enough that any company using it would have to give the user control of the device, and compelling enough that many companies wanted to use it, then this could increase software freedom significantly. It could perhaps even contribute if the manufacturers insisted on putting in some proprietary blobs, but I would prefer something more revolutionary. Even when included by "mere aggregation," on an embedded device binary blobs are infectious because once you allow one in, it makes tracking top-of-tree really hard. The OS could, for example, have a sandbox for blobs, like NaCl in Chrome, to reduce this infectious effect.
Another might be, an open CPU and compiler for 0.5GHz single-core applications. There was some research paper about a legacy-free low-cost CPU that was performance competitive with ARM using half the gates.
Another might be, an open toolchain for NVidia, targeting CUDA applications not graphics. I guess there is one in TensorFlow that Google just released? I don't totally understand it. but if NVidia could be convinced to use gcc to do their JIT, even if they continued releasing binary drivers, they may contribute back to a free software foundation-layer driver that's good enough to write scientific applications. This would give people a lot more computing freedom because the applications for these GPUs are broadening.