He takes a while to get there, and doesn't quite drive the point home, but this is actually what André Staltz is proposing. Mesh-first devices. If we start at the bottom of his article, here's basically the plan:
1. Work on polishing IPFS, CJDNS, SSB[1], Beaker Bowser and some others into a "mesh-first" setup.
2. Sell, subsidize, or outright gift mobile devices to offline users in Africa (I'm going to assume these are android smartphones) setup to automatically join a mesh network, with these apps installed.
3. Users will transport these phones around, adding and creating content. Their phones will automatically sync content with other devices as they come in range. In general, only content that the user is interested in (that they follow) will be synchronized, though some may act as hubs syncing all manner of content for later redistribution.
4. As usage grows, there will be various regional and national mesh networks that build up.
I think a good comparison is to Cuba's weekly sneakernet[2] distribution, just with wireless mesh.
[1]: Secure Scuttlebut is one of the author's creations: https://www.scuttlebutt.nz/
[2]: https://www.wired.com/2017/07/inside-cubas-diy-internet-revo...