Usually in cellular (e.g. LTE, 5G) both the mobile phone and the cell towers implement all of their PHY using a mixture of custom SoCs, FPGAs and DSPs, and all the code behind of this is as closed as it gets (even the drivers).

What they are doing is basically implementing as much as possible of their cellular network with code that is open. Literally getting the signal as close as possible from the antennas, sampling, and processing everything with code that can be studied and changed.

This openess makes it easier for researchers to develop new ideas for wireless networks (just reprogram your CPU, instead of having to beg to a manufacturer to provide SDKs for their super secretive equipment).

Could it be used as part of the stack to create city wide non-cellular wireless network for Internet[1], Are there any such current projects anywhere?

[1]https://needgap.com/problems/51-non-cellular-network-mobile-...

Single Indian Telecom (Airtel) paid ~ 2 Billion USD(current price) in 2016 for 173.8 Mhz spectrum across 1800/2100/2300 Mhz bands. Where as small YC startup(Link in my parent comment) has created Bangalore city wide WiFi Internet(2.4 Ghz/ 5 GHz has been made unregulated and free).

So, definitely telecom spectrum cannot be called same as WiFi as one requires Billion dollar oligarchy to enter the space and the latter can be entered with 4 member startup team.

I don't deny the value of such project as this modules cost some big money (and licenses), but unless they really do a very good job it will introduce too much complexity IMO. You can also call someone at $vendor and cry in their shoulder.

It's already tiresome to work with different vendor quirks.