I think I have stayed on target with your arguments, but I'd like to push back a little or widen the view...
Re WYSIWIG:
Rather than WYSIWIG being bad, I think the introduction of scroll bars destroyed WYSIWIG from the start. It only works on a large monitor where one page can be viewed and worked upon. In the early 1990s it was only at uni or in the workplace that this was possible. Not at home on a 12 inch monitor. It's the same reason that a web page that is larger than the screen is a health nightmare (mentally and for poor mouse wrists).
Re computing using spreadsheets:
This made computing available to people who couldn't program imperatively. It's given more power to the average power user than you can imagine. See Scott Hanselman's interview with Oz du Soleil. https://hanselminutes.com/532/data-literacy-and-the-usefulne...
Re data storage:
Without the ability to program in a declarative way (i.e. SQL) it is going to be difficult to both store and retrieve data in a provably ACID way, yet non-programmers still get it done well enough with files.
Re Smalltalk:
Without OO knowledge, this great tool will not be useful to any desktop computing power user that has not formally learnt programming.
My point is that the tools that have survived and got the job done in the
"computerised office" have not required the understanding of a complete pseudo-math model (aka programming paradigm) in order to get useful work done. People did not need to be able to compile a program without errors. They used tools that can be iteratively hacked upon, and allowed non-purist tech people to become power users in the workplace.
TL;DR
Models of software environments that do not have to be programmable have proved themselves and the people who use them capable of getting the job done. That can't be replaced by your suggested model.
I've listening the episode you linked, and well, I disagree: first no, I'm a computer user and I do not use Excel, nor I use SQL much as well, BUT I do not use Windows as well, and that's the key, for the author there is a deep separation between "the professionals and the users", because Windows is built with that idea and he seems to be a Windows guy, instead I was born on unix (just due to a family friend gift when I was a child, a dismissed SGI O₂), so I've seen such separation BUT also shell scripts, unix CLI design/IPCs and having discovered original desktops after, their model, being switched from zsh+Vim to Emacs, till the WM (EXWM) I've see a far less separated world and it's outcomes.
For the author and I suppose for you as well, programming means create a project, an UI, publish the result as an installable package and so on. Yes, that's not for any computer user. But that's not end-user programming as well.
As I said before my OS is Emacs/EXWM (booted by NixOS), so I have most of my digital life in org-mode, org-roam managed notes, OS config included tangled from org-mode notes. My user programming is for instance a quick script to auto-archives pdfs bill from my mobile and landline, ... in their right place managed as org-attach-ments, creating a relevant org-mode heading with a BeanCount babel block. This means that with not much effort, no specific project, UI design, packaging etc I automate a generic end user task in a way no pre-made tool can do. This is not harder than using Excel or MS Office macros, but the outcome is far, far better. With this model I have a full-text search built-in, simply via ripgrep wrapped in Emacs to get live results as I type, no special setup/indexing needed. All my automation, shell scripts, elisp snippets etc fell in this form, single listings, a file at maximum, something simple enough for a non-programmer (like me, because I have a CS degree but I'm not a programmer). Anything fit the same UI, I can link mails (notmuch/ol-notmuch), transactions (BeanCount), binary files that demand specific GUIs to be read/manipulated (a simple elisp: org-mode link) and so on.
This can be done by ANY computer user, if he/she is trained to this model instead of the Windows one. I know for personal experiments with some friends children. While the Windows model end up in the Android model and in the https://www.theverge.com/22684730/students-file-folder-direc... because that's needed for GAFAM et al. business. The classic model end up in an educated userbase who grow and learn with pleasure for their lifetime. While stuff you learn for Windows or web stuff is rubbish in few years simply because the upstream is changed.
The model I propose is the school model, we start learning, learn a bit for some years, profit from life keeping learning, producing new things. The Windows/GAFAM model is those of the ignorant user, where anyone can start clicking around and arrive to a point of being able to do something, still unable to grow, still not mastering their tools. This models seems to be quick at first, but it's a dripping project that ensure persistent frustration and ignorance, impeding innovation and controlling the users like sheep.
Comments
I think I have stayed on target with your arguments, but I'd like to push back a little or widen the view...
Re WYSIWIG:
Rather than WYSIWIG being bad, I think the introduction of scroll bars destroyed WYSIWIG from the start. It only works on a large monitor where one page can be viewed and worked upon. In the early 1990s it was only at uni or in the workplace that this was possible. Not at home on a 12 inch monitor. It's the same reason that a web page that is larger than the screen is a health nightmare (mentally and for poor mouse wrists).
Re computing using spreadsheets:
This made computing available to people who couldn't program imperatively. It's given more power to the average power user than you can imagine. See Scott Hanselman's interview with Oz du Soleil. https://hanselminutes.com/532/data-literacy-and-the-usefulne...
Re data storage:
Without the ability to program in a declarative way (i.e. SQL) it is going to be difficult to both store and retrieve data in a provably ACID way, yet non-programmers still get it done well enough with files.
Re Smalltalk:
Without OO knowledge, this great tool will not be useful to any desktop computing power user that has not formally learnt programming.
My point is that the tools that have survived and got the job done in the "computerised office" have not required the understanding of a complete pseudo-math model (aka programming paradigm) in order to get useful work done. People did not need to be able to compile a program without errors. They used tools that can be iteratively hacked upon, and allowed non-purist tech people to become power users in the workplace.
TL;DR
Models of software environments that do not have to be programmable have proved themselves and the people who use them capable of getting the job done. That can't be replaced by your suggested model.
I've listening the episode you linked, and well, I disagree: first no, I'm a computer user and I do not use Excel, nor I use SQL much as well, BUT I do not use Windows as well, and that's the key, for the author there is a deep separation between "the professionals and the users", because Windows is built with that idea and he seems to be a Windows guy, instead I was born on unix (just due to a family friend gift when I was a child, a dismissed SGI O₂), so I've seen such separation BUT also shell scripts, unix CLI design/IPCs and having discovered original desktops after, their model, being switched from zsh+Vim to Emacs, till the WM (EXWM) I've see a far less separated world and it's outcomes.
For the author and I suppose for you as well, programming means create a project, an UI, publish the result as an installable package and so on. Yes, that's not for any computer user. But that's not end-user programming as well.
As I said before my OS is Emacs/EXWM (booted by NixOS), so I have most of my digital life in org-mode, org-roam managed notes, OS config included tangled from org-mode notes. My user programming is for instance a quick script to auto-archives pdfs bill from my mobile and landline, ... in their right place managed as org-attach-ments, creating a relevant org-mode heading with a BeanCount babel block. This means that with not much effort, no specific project, UI design, packaging etc I automate a generic end user task in a way no pre-made tool can do. This is not harder than using Excel or MS Office macros, but the outcome is far, far better. With this model I have a full-text search built-in, simply via ripgrep wrapped in Emacs to get live results as I type, no special setup/indexing needed. All my automation, shell scripts, elisp snippets etc fell in this form, single listings, a file at maximum, something simple enough for a non-programmer (like me, because I have a CS degree but I'm not a programmer). Anything fit the same UI, I can link mails (notmuch/ol-notmuch), transactions (BeanCount), binary files that demand specific GUIs to be read/manipulated (a simple elisp: org-mode link) and so on.
This can be done by ANY computer user, if he/she is trained to this model instead of the Windows one. I know for personal experiments with some friends children. While the Windows model end up in the Android model and in the https://www.theverge.com/22684730/students-file-folder-direc... because that's needed for GAFAM et al. business. The classic model end up in an educated userbase who grow and learn with pleasure for their lifetime. While stuff you learn for Windows or web stuff is rubbish in few years simply because the upstream is changed.
The model I propose is the school model, we start learning, learn a bit for some years, profit from life keeping learning, producing new things. The Windows/GAFAM model is those of the ignorant user, where anyone can start clicking around and arrive to a point of being able to do something, still unable to grow, still not mastering their tools. This models seems to be quick at first, but it's a dripping project that ensure persistent frustration and ignorance, impeding innovation and controlling the users like sheep.