Here's a link to a moderately up-to-date writeup of the idea.
The goal of the project is not to create new technical capacity, it's to make what is already possible in high-income countries practically available to everyone, particularly people in poor countries. We actually expect that high-precision GNSS positioning capacity will arrive in Africa in the next few years regardless of our efforts, but it will likely be in a proprietary, closed fashion that will not be equitably accessible to ordinary citizens. This project may be more about how and when it happens than if it happens (the goals being sooner and in an open, equitable, accessible way).
Hardware improvements are currently outpacing user-facing software utility. The high-precision chips mentioned in the IEEE article you linked to are indeed coming to smartphones, but the software systems to use them are obscure, often proprietary/closed, temperamental, and in some cases frankly broken or even absent. What is available is often optimized for industrial rather than citizen use. The claim "This new chip will give the next generation of smartphones 30-centimeter accuracy as opposed to today’s 5 meters" elides a lot of complexity! There are all kinds of details that make this theoretical capacity very hard to achieve in real life: https://rtklibexplorer.wordpress.com/2019/08/28/a-first-look-at-the-broadcom-bcm47755-dual-frequency-receiver/. These are the kinds of issues we want to address with open software, helping people actually get usable results out of the new high-precision hardware.
We are starting with an external GNSS receiver for both rover and base station use, albeit an inexpensive one in hopes of keeping things accessible from the start. Using an external device as a rover is mostly to keep matters simple initially; there are a lot of hidden gotchas dealing with a GNSS receiver buried inside an Android device (even if it's the same actual chipset as in an external device). Notably most Android devices can't be dissuaded from duty-cycling the GNSS chip to conserve energy, which degrades accuracy and complicates data use. But certainly the medium-term goal is to enable the best possible accuracy with internal GNSS receivers in locally-available phones as well.
The first use-case is indeed land surveying in low-income countries, but we expect that folks in low-income countries will come up with uses we haven't thought of!
Our hopes for impact are pretty ambitious: improving land tenure/property rights for the poorest billion people in the world. For example, about 70% of Africans live in informal settlements without formal land tenure (both urban and rural). In some cases this is due to lack of government capacity to survey smallholders' land, and in other cases it's due to lack of official desire to recognize smallholders. Within the OpenStreetMap community, we've seen some evidence that establishing "facts on the ground" showing where people live makes it more likely for those people to be recognized, acknowledged, and provided with public services (the Map Kibera initiative in a Nairobi slum has been incredibly effective at this).
We are working with neighborhood associations to create mutually agreed-upon cadastral boundaries, which even in the absence of government recognition can facilitate access to credit, ability to sell land, insurance, and help people defend their homes from arbitrary seizure by corrupt officials or resource extraction companies. We're piloting this on a neighborhood scale in Tanzania at the moment using our tricky and complicated surveying system; if the social impact is positive we'll want to be able to scale it, which will definitely require making it a lot less tricky and complicated.
Of the various methods of GNSS augmentation, we want to enable people to independently perform Real-Time Kinematic (RTK) and Post-Processing Kinematic (PPK) correction using low-cost base stations. Basically people can set up cheap ($250) base stations that transmit a stream of correction information by Internet and radio; everyone within a few tens of kilometers of a base station can then make use of that stream to correct their positions (so, for example, the city of Dar es Salaam could be entirely covered by seven or eight base stations enabling high-precision surveying throughout the entire metropolitan area). RTKLib in its current state is probably sufficient for this—from tests we've done in the last few months with our colleagues at the University of Delft in the Netherlands it's only slightly less accurate than top-of-the-line commercial systems for RTK, and essentially equivalent in quality for PPK. We'd also like to enable easy use of more sophisticated Precise Point Positioning algorithms for specific points such as the base stations themselves and key monuments.
We aren't likely to add anything to RTKLib itself, but rather integrate it into a friendly user-facing package that ordinary citizens can use for surveying. We'd certainly want to include Tomoji Takasu, the developer of RTKLib, in the conversation. There's a lot of work to be done just getting RTKLib working with the current GNSS recievers both inside and outside of cellphones, making correction data available, accessible, and useable, enabling people to create their own correction data, and improving user-facing surveying tools to be able to use GNSS data and correction procedures effectively.