I came to a similar conclusion as a computer scientist working in industry who ended up transitioning to a community college teaching career. I remember being inspired by the stories of Bell Labs and Xerox PARC as a high school student and undergrad, wanting to follow in the footsteps of researchers like Dennis Ritchie. However, the industrial research landscape has changed in the past 15 years. It's very difficult to find truly curiosity-driven places with long timelines and little pressure, and industrial researchers these days are pressured to work on projects with more immediate productization prospects. I've seen this firsthand at a few companies. The tenure-track at a research university route requires playing the "publish-or-perish" game, which is also a curb on freedom and is also filled with pressure.

Being a tenure-track professor at a California community college is a happy situation for me. I love teaching, and for roughly 8 months of the year I'm dedicated to teaching. Tenure at my community college is entirely based on teaching and service; I'm not required (or even expected) to publish. I also get roughly 4 months of the year off (three months off in the summer, one month off in the winter). I spent much of the past summer in Japan collaborating with a professor on research. The only serious downside is not being able to afford a house within a reasonable commute from work, but I had the same problem in industry; not everyone in industry makes FAANG-level salaries. In fact, my compensation is effectively a raise from my previous job when factoring in going from roughly 3 weeks of PTO per year to 4 months off plus 10 days worth of sick leave; I took a roughly 10% pay cut in exchange for greater freedom and roughly 5-6x the annual time off.

I've learned that being a hobbyist researcher with a stable job that provides summers off is quite a favorable situation, since I don't have to worry about my job security being tied to my publication and fund-raising counts. Most of my computer science research can be done on a mid-range laptop with an Internet connection and access to textbooks and academic databases; I don't need equipment that cost five- or six-figures (though it would be nice to have a GPU cluster....).