This is exactly me.
In 1984, when the Mac came out, I was a starving artist in Denver. I had a lot of college at that point, but no degree. I had foolishly wanted to learn everything interesting, so I had a year of art school and three years of coursework in biology, physics, and mathematics, but I was working in labs and retail by day, and by night as part of a songwriting partnership that ultimately went nowhere.
I was fascinated by MacPaint and wanted to learn how to do that. I had some FORTRAN and some BASIC from coursework, and a little FORTH from a friend, but they hadn't captured my imagination. MacPaint did.
I couldn't afford a Mac. They were 2500 dollars. In 1984 that was half of my annual income. (Yes, I realize that's unbelievably low. Starving artist.)
With help from the woman who became my first wife, I got hold of a Commodore 128. I started by typing in BASIC programs from Compute magazine. One of them was an assembler, and that's where things got really interesting.
I did end up writing an absolutely terrible paint program, but what really obsessed me for months was writing variations of Conway's life. The one I spent a lot of time on was a version that ran multiple sets of rules on the same field at the same time.
The color of a live cell was determined by the set of rules that governed it. Live cells of different colors could see each other for the purposes of liveness analysis, which is how the rulesets interacted. They were otherwise independent of each other. I spent countless hours poring over whatever references I could find about how to make things faster and more comprehensible, and tweaking the rulesets to see what happened. I would leave a program running overnight to see what it looked like in the morning. Often it was a boring disappointment. Sometimes it was spectacular. The rush of the spectacular ones kept me going.
I wrote other programs, too, trying to understand how window and graphics systems and compilers and interpreters worked. I worked in bookstores and read everything I could get my hands on about programming. In those days, retail bookstores paid crap wages, but allowed the staff to borrow all the books they wanted. It was a good deal, from my point of view.
In the fall of 1987, my mother sent me an ad that Apple Computer ran in the Mensa Bulletin, looking for hires. I applied and, to my utter astonishment, Apple hired me, and that started my profession as a programmer and a technical writer.
The way I got the job prepared me well to take advantage of Apple for learning more. I came into the company ravenous to learn all I could about programming. The company had three institutions perfectly designed for an autodidact like me.
First was the corporate technical library, a huge collection of technical books available for employees to borrow. Their policy was that if you wanted to borrow a technical book and they didn't have it, they would order two copies: one to lend, and one for you to keep. I got a lot of expensive technical books for free.
Second was the software library. They wouldn't give you a copy to keep, but they would order software that you wanted to check out, and they would let you renew the checkout for as long as you needed. They also maintained copies of basically every piece of software Apple had ever developed, released or not. My first hands-on exposure to Smalltalk was a check-out copy of Apple's implementation of Smalltalk-80.
The final pillar of my self-education was Apple's competitive intelligence lab. It was essentially a hardware library with instances of more or less every interesting hardware architecture you could imagine. They had a Cray Y-MP. I learned Lisp and emacs on Unicos running on the Cray. I used to sign up for time on the Xerox machines to play with Smalltalk and Interlisp and the Star Office system.
I agree with those who say that teaching yourself leaves holes. I've been identifying and filling in those holes ever since 1987. On the other hand, I can't say I regret how my natural curiosity ended up leading me into what became my profession. Undoubtedly, I would have been better prepared if I had taken a disciplined approach to learning programming from the beginning.
On the other hand, if I'd had that kind of discipline in the 1980s, then I probably never would have discovered that I liked programming in the first place.