The principle problem for me is that Mathematica V12.1 does not work on the Microsoft 7 OS, although WRI claims that both V12.0 and V12.1 work on that platform. (Version 12.0 does work.) The problem is that notebook sizes and font display are altered so greatly (like the Option Inspector display is altered so much that it stretches over more than two 25" monitors!) that it is simply unusable. WRI did attempt to fix it, first by giving me suggestions for making system changes, none of which worked, and then with an hour and half Zoom session which was not able to fix the basic problem.
I first used Mathematica in 1996, just before Version 3 appeared. I liked the idea of a notebook interface. After some time it seemed to me that Mathematica notebooks were a fantastic new medium for development, communication and publication. You can hand a reader not only static information but also active calculation and dynamic displays. Such documents have much higher integrity than static printed documents. It is still possible to make mistakes, but not nearly as easy. Calculating or plotting from an expression is a kind of "proofreading". No error occurs from copying a result to a printed document. It is already in the document.
Unfortunately, this has not worked out nor made much of an impact. Exchange of information by users through the notebook medium is entirely through a few Mathematica aficionados. There is no detectable publication through this medium at all, despite attributes that should make it superior to pdf files or printed papers by orders of magnitude. Why doesn't the Mathematica notebook medium dominate technical publishing?
One reason is that this is a new medium with new attributes. (30 or 40 years is nothing in adapting to such a medium!) With every single approach that has style, smoothness and elegance, there are a hundred ways to produce cutesy drivel that descends to "computer junk" as Edward Tufte might say. It takes time, experience, usage, critical evaluation and communication. One important consideration is how much effort is required from the writer and from the reader to use a feature.
References and numbered equations are a feature of printed papers that must go back at least 200 years. Wolfram dutifully incorporated them into notebooks. But they copy the original and fail to take advantage of the medium. Try to use Find to locate an automatically numbered equation. When accessing a reference or equation, don't transfer the reader to another part of the document and then (maybe) provide some way to get back. Rather, bring the information to the reader at the place he is and where he wants to use it. So we might for convenience still provide a list of references at the end of a notebook but references in the text would not be numbers or links but Openers or Tooltips or Buttons opening a window with the reference.
Accessory information may include many things besides numbered equations. How about definitions, theorems, graphics, dynamic displays, sets of axioms? These might be in an Association in a crib sheet palette. The crib sheet might contain paste buttons for common routines used in the notebook and links to their documentation and also a dropdown Keys menu for the Association. Mathematica has thousands of routines but if there are a dozen rather specialized routines used in a notebook then this is a great convenience to the reader. It's building on intrinsic attributes and assembling selected resources for a restricted application.
A major shortcoming of Mathematica is Wolfram's penchant for supplying routines whose specification is top-down without providing bottom-up alternatives using basic primitives. Many of these routines involve graphical displays, to which WRI adds cutesy features with hidden code. They may provide varied forms but often not the form one wants and then the writer is stuck. If "everything in Mathematica is an expression" (not true) then why not "everything in graphics is a graphic primitive"? Then WRI could provide common routines with top-level specifications without the writer being trapped in them.
Another problem with Mathematica is the proliferation of Styles and Fonts that go with various constructs. Many of these Styles are deep in the Core stylesheet and may be difficult to find. I especially object to the graying out of various types of information. For example unit Quantities are displayed in Gray. Why? If these are the result of a year's experimental work and the chief result why are they in Gray when everything leading up to them is in black? Does the Physical Review print units in Gray? If one goes to the Core style sheet (and has one's own style sheet) one can change it but how many writers will know how to do that? Similarly for the pinking out of an Inactive expression. The Inactive and Inactivate routines are very nice. They give an additional way to present step-by-step evaluations and derivations, which I like to do. But they should look like regular expressions and not be pinked out. It has a Tooltip that indicates an Inactive expression; it doesn't need a second indication.
The entire FrontEnd gives the appearance of being a succession of ad hoc additions held together with rubber bands and scotch-tape.
The documentation facilities for application writers are just barely usable. The documentation for it is skimpy beyond compare. It still uses the old Workbench 3.0. I've worked with two other applications authors who tried to do documentation but gave up because they couldn't make it work. What is needed for writers is not necessarily in sync with WRI's documentation needs. Wolfram takes no interest or notice of what would be convenient for them. Workbench might be useful for WRI development but why couldn't a separate, cleaner, more transparent and better documented application be provided for those who just want to do documentation of their own applications?
Applications, as opposed to a simple package, are a powerful form of the new medium but most users are barely aware of them.
Mathematica has many nice features as well as many undesirable ones but for me the dream of a new medium sits like a neglected step-child in some dusty corner while Mathematica itself lies in shambles on the floor.