"That property is still useful, and thus in 2017 the AT convention has survived in some interesting places. AT commands have been found to perform control functions on 3G and 4G cellular modems used in smartphones."
But modern modems aren't connected over serial. There's no concept of line speed. This property is entirely irrelevant to modern hardware. A much more plausible explanation is simply that extending the interface made it easier to extend existing codebases to new contexts.
"IoT devices still speak RS-232"
This is actually dangerously untrue! RS-232 used positive voltages for 0 and negative voltages for 1, between 3 and 15 volts each. Attaching RS-232 to a modern IoT device's serial interface is likely to kill the device. What is common between old-school serial and the serial ports on modern devices is stuff that's out of the scope of RS-232 (eg, the 8N1 framing isn't defined by RS-232), and using the RS-232 terminology to describe it is about as accurate as calling it RS-422.
"If you know what UTF-8 is (and you should) every ASCII file is correct UTF-8 as well."
I mean kind of? ASCII is a character set, it doesn't define the on-disk representation. An ASCII file saved with each character packed into 7 bits isn't going to be valid UTF-8 without some manipulation. This is just an odd claim to make given the earlier discussion of varied word lengths and transfer formats that weren't 8-bit clean.
"But in 2005 Linus Torvalds invented git, which would fairly rapidly obsolesce all previous version-control systems"
(cries in Perforce)
Git was a huge improvement over CVS and SVN. Claiming that it rendered everything that came before it obsolescent just suggests massive ignorance of chunks of the software industry.
And yes this is all ridiculously pedantic, but given the entire tone of the article is smugly pedantic it doesn't seem unfair to criticise it on that basis.