First, let me say before i get critical about Knuth's work, that he did an incredible amount of systematic, meticulous, highly accurate and important work. The problems that i am going to talk about relate to the format he used to encode his work. The bottom line is that Knuth was like having a Shakespeare who decided to write in pig latin.
Knuth came up with the idea of literate programming as a sequel to structured programming. Unfortunately his approach was completely wrong. I am not the only one who has pointed this out (http://akkartik.name/post/literate-programming). I am a nobody compared to Knuth, so maybe stating the emperor has no clothes on may seem offensive. I don't mean any personal disrespect to Knuth, but his TeX product has left me cold from day 1. Having your code comments presented in nicer typography is well and good, but honestly, do you consider Knuth's TeX system any good? It is a disaster in my book, a bizarre, complex curiosity that he squandered a big chunk of his career on. Knuth was so ridiculous that he refused to use TrueType for encoding fonts. He invented his own font format, Metafont. He was so famous and influential nobody pushed back on him, but who else on earth would refuse to use one of the commercial type formats? Its like someone building a motorcycle and deciding that neither English units or Metric is good enough, and that you need to use an incompatible set of nuts and bolts. The history of computers is full of non-agreement on basic standards. ASCII vs. EBCDIC, Mac vs. PC, iOS vs. Android, etc., but why when 99.9% of the world is one of two camps, do you invent your own 3rd form which has no substantial advantages.
The problem with Wolfram Mathematica's notebook approach is that you are limited to that single way of working with Mathematica. There is only one modality you can use. It makes it very hard to enter the world of software interchangeable parts. The invention of interchangeable parts goes back to the 1800's, and in America the great inventor Eli Whitney was one of the pioneers. It later was used to devastating results in the American Civil War, which was the first time armaments had been mass-produced using that technology. I want to see software enter this era, and Mathematica's notebook approach did not facilitate this, so I consider it an impediment. You are entirely correct that interchangeable parts means small chunks of code that snap together like lego.
Knuth's choice of MIX was never fine at any time. At any moment from 1965 onward a single company has owned the lions share of the CPU market. It was IBM, then DEC, and for at least 36 years since the IBM PC, the Intel instruction set has 99% of the desktop market. Nowadays the ARM instruction set has 99% of the mobile market, but server and desktop are 99% intel architecture, and if Knuth had for example picked the intel instruction set, not only would most of the code still run fine, because the intel architecture has been phenomenally backwards compatible, but there are many commercial cross-assembler tools that efficiently convert intel instruction set to other chips like Motorola 68000, MIPS, ARM, etc.. By using MIX he doomed his unbelievably meticulous work to be basically unused today. What company can make a living selling and maintaining a MIX cross assembler, when only one human in history ever used MIX? I argue that Knuth was being perversely manufacturer neutral. I can't tell you how many programmers i have seen have his books on their shelf, but never actually used them.