Aside from a few minor quibbles (such as the supposed difficulty of telling apart the vulnerable and role functions), I completely agree with Jack's criticisms of Model G. In short, it adds essentially nothing to Model A, is steeped in vague jargon (like "long range" and "short range", the social vs. personal spheres, etc.), and is at odds with how the types actually work (in particular how it conceives of the suggestive function as being somehow "high energy" or more prominent than the mobilizing function — or whatever they're called now).
Model G does, however, have a few selected insights. One is the greater emphasis on the benefit rings, or rather, the bold/cautious dichotomy.
Another is the idea of "energy." For some reason Ben flounders in this video when asked to define energy, while he had previously connected it to Jung's concept of libido (not Freud's), which Jung defined as a kind of "life force". This is a sound idea in itself: Model A addresses information processing (information metabolism) but it does not address the obvious limitations of resources that apply to each function's processing, in particular the strength and boldness traits.
Unfortunately, beyond this very basic outline, the details of Model G seem disconnected with the reality of the types — including which functions are supposedly maximum energy, etc.
That's a very brief take on the semantics — the details are really not that interesting and are addressed in the video. Another interesting idea (which precedes Model G) is the signed elements, but I'm not going to get into that here either. The real point of this post is the structural deficit which I pointed out and Jack later mentioned in the debate. It's very obvious if you look at Andrew's diagrams (and translate the names and numbers accordingly):
As you can see here, some of the standard socionic dichotomies are presented asymmetrically in Model G. Gulenko apparently does not assign any meaning to the left and right sides of the model, he still uses the standard dichotomies such as strong/weak, etc., albeit with different names, as displayed on the right here. Strong/weak becomes "master/slave" and valued/subdued becomes "values/tools". (correction: It is said that Gulenko does consider the left-hand functions to be "better" in the "sense [...] of energy allocation to the function and the degree of freedom of behavior afforded by this." Again, more jargon which does not seem to apply to the suggestive function, or if not actually wrong, at least is not as clear as the existing Model A dichotomies.)
How can we fix this? The obvious thing to do would be to simply switch the suggestive and ignoring functions (or "manipulative" and "control", numbers 6 and 8). Then the left side is strong and the right side is weak. And the valued functions are "outside" and subdued "inside", which is at least as good as it is in Model A. And we still have the benefit rings proceeding horizontally: NeTeSeFe... and TiNiFiSi....
In my opinion this clearly shows that the Model G blocks are defined wrong. The issue is that Gulenko wanted to have a benefit loop of types, but he represented them using the standard Model A ego blocks (as the columns of Model G). So for ILE we have NeTi, then TeSi, then SeFi and FeNi for the ILE and its Process-Extrovert ring. Instead we should express the types entirely in terms of the benefit ring (in analogy with how it is in Model A and the supervision ring) with each represented using a consecutive pair of elements in the ring, ILE being NeTe, LSE being TeSe, etc. Then the types can be thought of as "edges" between the IM elements, and they interact at their shared points. This makes much more sense if Model G is meant to show energy flow, does it not?
Ben made the interesting point (possibly the only one he made in the entire video) that Model G includes not only benefit rings, but also supervision rings if you extend it vertically:
(Written in to the left of the #4 heading.)
Note that this property still holds if we switch the suggestive and ignoring functions. There are literally only two ways to make a grid like this, and Model G does not use the right choice.
Aside from the semantic issues with Model G, this is a very obvious structural flaw. Fixing it might be the first step to salvaging the model.