I think being self-taught in programming works relatively well compared to being an autodidact in theoretical physics becaude programming is a trade or applied science, not a pure theoretical science. Being self-taught in theoretical computer science and doing great work in that field is more difficult than being a self-taught programmer and writing great code. (Obviously, learning cs theory can help you be a much better programmer too, and learning cs theory yourself to the degree that you can gainfully employ some learnings in practice to your code is well withing the realm of self-teaching, but this again is different than becoming a theoretical computer scientist or breaking ground in the field, as discussed in this article vis a vis physics.)

I imagine this is true for many trades. One can be an effective autodidact carpenter. An effective autodidact musician. An effective autodidact writer.

It's interesting to consider which trades being an autodidact is seemingly most difficult in. Being an autodidact doctor or surgeon seems a major stretch, regulatory issues aside. Being an autodidact civil architect or even mechanical engineer seems difficult as well, though no doubt possible for some. Perhaps the worse the consequences of mistakes (doctor, surgeon, architect) and the higher the bar of knowledge to be effective (mechanical engineer in advanced spaced, doctor, etc) the less likely autodidacts are to be effective. This reconciles well with programming being a feasible autodidact field -- consequences are neither as dire as being a surgeon nor failures as clear to see and measure as being an architect, and it doesn't take a great deal of knowledge to get started and be productive in programming.