I think this is, from a quick skim through, the beginnings of a very good idea. An LMS that does almost nothing. Quoting directly:

"What would this LMS look like? In my view, it would have three things:

1) a course roster with stellar SIS integration

2) a gradebook

3) a rock-star LTI and API

That’s it! Oh, except it would also be open source, students would control their own data, including publishing any of their work or evaluations to the block chain, and you could host it locally, distributed, or in the cloud. Never mind the pesky privacy laws (or lack thereof) in the country hosting your server, because the LMS is back on campus. Not connected to the internet? That’s okay too, because there is a killer app that syncs like a boss (like Evernote. Has Evernote ever given you a sync error? No, I didn’t think so.)

Who wins with the new LMS? Students because they own and control their data and it costs less to buy and run. Instructors because they have a solid core with the option to plug any LTI into a class hub. Institutions because costs are lower and the system more secure.
Who loses? The EdTech companies. Or do they? Without standard wiki features and discussion portals, startups and the old standard barriers can invest their R&D and venture funds in really great tools."

The principle is a little like that of Elgg, that consists of a very small core, with everything else coming from plugins that use the API.

It seems to me that, though this concept allows its users (teachers, students, admins alike) to do what they like with tools and data, it is still firmly based around the assumption of a traditional classroom model, and seems, as much as the traditional LMS, to reinforce that view. It's still a course and grading management system, not a learning management system. It needs something that goes beyond the classroom, even in a traditional institutional setting. It needs much more flexible groupings, networks and sets.

With that in mind, this lightweight LMS still seems heavier than needed. A SIS might well provide course information, so that might be redundant. If not, a plugin or service could be written, rather than including it in the core. I am not at all sure that an integral gradebook is needed either, for much the same reason. It might, instead, benefit from a standards-based open source learning record store using xAPI (TinCan), like Learning Locker.  Or, perhaps, an integration of an OpenBadges backpack. Either, along with APIs that allow integration with things like SISs that could make badges look like grades or that could identify relevant learning records, could serve the necessary functions and allow a great deal more openness. Perhaps integrated support for some kinds of grouping and networking would help satisfy the needs of those that want to build institutional courses. All that is really needed is that rockstar API to pull it all together. This begins to sound a lot more like Elgg, and something that could, in principle, be implemented within it.

The blockchain idea is a good one: being able to free data from a central machine is much to be wished for. But it bothers me that privacy laws are seen as pesky and that they should be circumvented. They are pesky, for sure, but with good reason. We cannot force students to part with private data where laws do not protect them (I do have at least one course that does this, but it's one of the conditions of enrolling because we are actually studying such things). What people do of their own accord is, of course, just fine, but the tacit assumption that this LMS-lite continues to reinforce is that learning happens in courses that lead directly to accreditation. That's not about people doing things of their own accord.

With that in mind I can foresee a few interesting issues with authorization too, whatever path is taken. The mechanisms for deciding who allows what to be seen by whom might turn out to be quite complex because of the tension between hierarchical roles implied by this system and individual access authority implied by the freedom to use anything from anywhere, especially given the balkanization of social media space that currently exists and that is likely to form a good part of the basis of actual learning activities. Anything that is not public is going to have to interface with this in some quite tricky ways.

For all its embedded assumptions, I like the idea. Building an Elgg-like system with integral LTI, especially if it could support more learner-centric technologies like xAPI, OpenBadges and so on, seems like a sensible way to go