[EDIT: I am leaving my original post below untouched, but now I feel that losing "popularity" is an incorrect way of phrasing my views. Rather, the philosophy is straying from what I (for one) felt comfortable with (which can affect popularity if a sufficiant number of users agree with me ). Please keep this in mind as you read on]
I have to say that, in general, I agree with the perception that Cocos2D is losing popularity. I have no source for this, I have seen no stats. I am a hobbyist who has not published a single line of code. If that discredits me, so be it...
But I have to admit that in its latest iteration (since the inclusion of SpriteBuilder) it is becoming less attractive to me. A while ago I posted a request to see if the templates could be simplified. The (sensible) response - and I paraphrase as I am not looking it up - was that everything was tightly integrated: Chipmunk, the Published floders, the CCB files, etc. It wouldn't make much sense to decouple the additions when they don't take up much room. All I need to do is delete the "offending" files (in my case) and carry on. As I said, a perfectly sensible response.
Yet it left a bit of a sour taste in my mouth (figuratively speaking). The thing is, I liked the way that the previous versions of CC2D were just an API provided that you included in a wider project. It was a great, additional, brick in your arsenal of programming LEGOs, with the flexibility to add/remove some wider components (Box2D, Chipmunk, etc). Just add the source in your projects, include references to it in your code. CC2D could be a little, magical back box that you could call upon in order to achieve your sprite needs.
So, is this still possible in the current CC2D. Why, yes... and I don't even have to include the term "strictly speaking". The CC2D base is still there... I have to peel away some layers, but its still there.
However, what I feel has changed is the perception of what CC2D now is. In another thread, I read a discussion about whether the site should be cocos2d-swift, or go back to cocos2d-iphone, or whether to move forward and name it cocos2d-spritebuilder. This discussion represented to me the grey area that the platform has become, as far as perceiving what it serves. No longer is is mainly a graphics API, with additional sound and physics APIs included - it is now a tightly integrated platform.
With the idea of Spritebuilder, the creators have chosen a path - a pipeline if you will. Want to do physics? Well Chipmunk now serves your needs, integrated into a Node (that great, basic unit which is subclassed to whatever your heart desires). However, by endorsing this path (or any path), I feel suddenly dirty if my preference runs toward other things. Suddenly I feel that I am left in the cold if I prefer Box2D to Chipmunk. I liked the clear distinct separation I programmed between the graphics implementation (CC2D) and a game world based on physics. CC2D was mostly a way to achieve graphics and animation. Now, I "perceive" there is a great deal of the platform that does not apply to me, just because it's not how I would approach or structure the effect I want to create. The platform which integrates the basic code I found so impressive and easy to use implicitly endorses doing things in the way that I may not like or need.
The other effect is community. Suddenly the forums and documentation don't have as much information about alternative paths to development - at least applicable to me. I see many threads about CCPhysicsNode... well, suddenly those threads (or many based on other themes) no longer apply to me. I see a solution which details what I would need to do, but using the "endorsed" method and code that I cannot apply. Sure, the solution is still there, but I need to re-work it for my needs. Before, the code could be pretty much copy/pasted into my own programs. It leaves a bit of a sour taste.
Again, the basics of what I love from a pure coding perspective is still there - greatly improved, even, as the version numbers have grown. But the developers have chosen an endorsed way of doing things with spritebuilder that, I feel, may leave a few folks behind, especially from a community perspective. After all, the new focus is on visual programming, with multi-platform possibilities and integrated, tightly coupled features using specific libraries (e.g. Chipmunk).
So, where does that leave me. Well, recently I became stuck on a CC2D project (my vision of what I wanted was unclear). So, in taking some time off from it, I stumbled into Unity - just to try it out. It is a strange sensation. Unity's advantages are many. I was impressed by the way I prototyped something into place with little code in under 2 days. I accept that in giving up coding, I have to submit to the Unity way of doing things. But its hard to mind when Unity does this so well, giving you a lot of flexibility to tweak the sanctioned way of doing things. With Unity, the emphasis is on the tool, not the coding. And I am enjoying its use.
However, I can't help but feel guilty I am not coding in CC2D. With Unity, I miss the nuts-and-bolts programming aspect of creating lower-lever effects by my code. Unity provides pretty much everything. Scripts in Unity merely tie the features already there. With CC2D, I am creating the features...
Well, I was... I am sure I will shake off the "perceived" disfocus (made that up) that I see in the cc2d platform. And I am sure that I'll have no problem going back to the beauty of code after I get bored with the automated stuff in Unity. But the sour taste remains....
Sorry about the lengthy post... I rambled in trying to explain my view. I feel that this discussion gave me room to vent a bit... to be honest, if I am in the minority with this view, then I have no qualms in being ignored