One thing to consider for visual programming is whether you are describing data flow or control flow. Most of the examples where visual programming is working

One thing to consider for visual programming is whether you are describing data flow or control flow. Most of the examples where visual programming is working well are either for data flow: shader/material graphs in game engines, compositing and other node based editors in various video,3d,vfx software. Other examples of case where visual programming is used audio software and industrial automation logic - it's similar although here it describes more of continuously evaluated process that all happens at the same time, with some parts potentially being evaluated at different rates instead of simply evaluating whole thing in a loop. All that is more similar to functional programming. Describing control flow, implies imperative programming. You have to ask why are imperative programming languages more popular than functional ones? I don't have a good answer to this part. The same reason might also be at fault why visual programming isn't more popular, assuming visual programming is better for functional programming. But why is visual programming bad for describing control flow? It's not like we don't have standard way for doing that. Flow chart style diagrams, is one of the basic tools used when teaching programming. I guess one of the factors might be goto vs structured programing. ~50 years ago programmers realized that arbitrary jumps across the code makes a mess, and it is beneficial to restrict the control flow to few specific patterns. Flow chart style visual programming is more or less goto spaghetti. A disciplined programmer could restrict themselves to the patterns that match with structured programming, but the same could be said about using goto in text based programming languages. Reality shows that unless they are forced, relying on individuals being always disciplined (or knowing what the best practices) doesn't go very well. It's more effective to have a first class support for the preferred structures, and remove or strongly discourage unstructured tools (even though in theory they are more powerful). With that said structured visual programming isn't impossible. Scratch is somewhat that, but it also shows why at that point it becomes somewhat pointless. You more or less get your usual imperative text program but with additional blocks drawn around. Closer to structured code editing(completely separate concept from structured programming) of text based language than visual programming. There is still bit of hope. All the examples for DRAKON language I have seen, looked mostly structured. Not sure how it's enforced. It does have first class loop structures, but supposedly branching based loops like in the flowcharts are also possible. Most recommendations for best practices will encourage programmers to limit the control flow complexity within a function, as having too many nested loops and branching will still make the code hard to understand. This makes me wonder about cause and effect. Do we limit the control flow complexity because text based languages are bad at expressing it, or do we use text based languages as a tool to restrict complexity (which is always bad). Or is it just a coincidence that we can use text based languages, because the amount of local complexity that we can comfortably work with fits the limits of text based languages. One more thought is that the text based programming isn't exactly purely text based. We don't perceive the code as stream of characters, like the computer does.Remove all whitespaces including newlines and it will become nearly impossibly to work with code beyond few expressions. Elements like line indentation and syntax highlighting rely on human visual pattern recognition. So maybe there is a way for more visual manipulation even for languages that strongly resemble our current text based programming languages.