I've been following TPS for almost a year, and it's gone from what looked like a Hyper Rails update to something much more. The project has delayed slightly with the introduction of unplanned features as well as expanding current ones. Pantera does need solid community support before doing something though, which is why things have the potential to change and improve even as we go along. For example, the coaster building system was originally planned to be a re-hash of the one in Hyper Rails, but following some comments and feedback, they are looking to b-splines or some other more advanced system.
The community is primarily from RCT, but I have to say that TPS is no real RCT replacement. It's all about park and ride simulation, there is no money/management/game features planned yet. (This may change in the future.) Basically, TPS will be for the RCT builder and modder enthusiast who was very disappointed at what couldn't be done.
While TPS and NL2 have a lot of overlapping features, I still don't see them as being in competition with each other. TPS will feature a full custom flat ride builder that's looking to be more robust than scripting, and have fully assemble-able custom trains. You don't even need to import any custom 3D models; you can mix and match parts that are built-in. Check out the videos on their Youtube channel. Also, the terrain editing system is shaping up to be pretty robust so far, and just like in Hyper Rails, the building space will be huge. Also, I trust the walk mode to transition in and out properly, given how the one in Hyper Rails was pretty well done. (At least by 2002 standards; the controls may have been poor, but the flow of going through a queue, boarding, riding, and unboarding a ride was done perfectly.)
Lastly, I suspect the game engine will be better optimized for parks. You can't really build more than 2 or 3 rides in an NL2 scene without unusable lag, although things have gotten slightly better after a patch. However, there's too many things we don't know yet, although what we do know is that TPS will ship with a decent base of pre-made assets: textures, models, rides, etc. instead of leaving it completely up to the community to build and share these items. The library won't be as large as in RCT3, but there will be enough to build something nice.
I do hope the track shaping tools improve, or at least that the track format is done well so that 3rd party programs can support it easily. The way I see it right now is that some projects are better suited to NL2, and some will be better suited to TPS. NL2 is still primarily focused on the single coaster, and the new abilities really just add to the scene without really shifting its focus. The engine can't handle enough rides yet to build a flourishing park environment. On the other hand, TPS will feature a robust pathbuilding system, guests, shops, etc. making it easier to truly complete the park environment. The compromise is detail and realism, but that's something I can live if my goal is to create a lively park scene focused on large environments.
Then again, I also had $100 to spend on both. Professional license and $100 tier on TPS's Kickstarter. I support the next gen of coaster/park simulation because let's be honest, it's been way too long since anything like NL2 or TPS have happened. Having both is even better, when SM/HR/RCT and so many others have died out. I disagree with the idea that TPS and NL2 compete fiercely; they're different programs meeting different needs.