The sorry state of UIs is more of a marketing thing than a computer science failure. There is a trade-off between gloss and performance. Gloss is the easier s

The sorry state of UIs is more of a marketing thing than a computer science failure. There is a trade-off between gloss and performance. Gloss is the easier sell. Let a little of the web doodaddery in, and users will not let up until you bring the rest of it. At that point, we just took the long trip back to Electron. Also, learning from terminals to make better UIs--especially if the big selling points are speed and composability--is developer Stockholm syndrome. The speed is only there because the user-facing I/O is very limited, and rich media is out-of-the-picture. The composability is only there because we have 40+ years of command flags to format and reparse every single ad-hoc format into something the next command can understand. Interfaces like Jupyter and Mathematica did a decent job integrating rich media. Symbolics Genera--and probably a few other language-machines--had a better story on composability (by having a common structured data format). Amen to "We should stop babying our users" though. Engelbart called that one from the outset--with something to the effect that if we're going to be using these things all our productive lives, most learning curves are going to be worth it--as long as the feature isn't garbage.