I agree with the 2 above and would add that when I started with vvvv, i was in the same position and maybe differently to a lot of other software I found that just looking at and messing with existing patches didn’t get me very far. There is sometimes some quite complex logic and usually patches that people have done are optimized and cleaned up to the maximum, so it might be hard to follow sometimes, especially in the beginning.
Understanding basic data types, spreads and a couple of basic nodes will be enough for you to start throwing some stuff together. Looking at the help patches of nodes you don’t know also helps (press F1 after clicking on a node), not just to understand what they do, but simply to know THAT they exist for use at some point.
For me it also took a little while to fully understand what it means that everything is evaluated every frame. So if you want something to only happen/be triggered once you have to make sure it only does it ONCE.
It can be hard to think of stuff to just do without a clear goal. The best is always a real-life project to get you started. Maybe start with something simple like making an mp3 file play (hint: filestream). Then you go on to use FFT to analyse it. Then you could make a simple visualizer using that data. Maybe start with a simple thing in a GDI renderer and then move up to something 2D or 3D in an EX9 renderer.
Nodes i use in almost every project: LFO, TogEdge, Counter, Switch, i, GetSlice and the basic math operations (+ - * /) and boolean (AND OR …).