I have argued for a long time that Permacomputing will be seen as the missing part of the Free Software movement. What use is free software long term if you do not have hardware you can control, maintain and repair easily? This will mean a sacrifice in performance and functionality but gaining control and longevity.
With things like Secureboot, TPM modules and ever increasing demands to lock down systems, there is the risk that even libre software will be snuffed out. While not those technologies explicitly, similar less friendly things may come up in future. And when that happens, being beholden to billion dollar hardware companies won't seem so friendly. A little alarmist, but I didn't think we would be were we are today as it is.
One interesting area is about how to make software that is not hardware locked but easy enough to implement with very little work involved.
This is where projects such as UXN come in. https://100r.co/site/uxn.html
A system spec that is only 32 instructions deep, something that a single person could implement in less than a week. Essentially the hardest part is building the hardware Abstraction Layer. It wouldn't be efficient but it is very portable and thus makes it resilient to any future possible shocks.
This project appears here from time to time and each and every time I am amazed. Thanks for sharing it.
Nils Holm does permacomputing without writting fancy manifestos: https://t3x.org
T3X/0 will assemble binaries for Unix/DOS (maybe Windows) and CP/M.
S9 can do great stuff with very little.
Klong it's a mini APL-like CAS more bound to Statistics than Calculus. No fancy Unicode needed.
Also, Luxferre doing an ultra-minimal numeric VM:
https://codeberg.org/luxferre/mu808
Read the instructions, that mini VM it's surprisingly able.
Finally, Subleq+EForth from https://github.com/howerj/muxleq (muxleq it's just subleq with parallel mux running the exact same intstructions).
From the book you can boostrap EForth from itself with a minimal Subleq DEC file. Enough to run a Sokoban, a calculator (set complex numbers as binomials), and you can implement q+ q- q* q/ to calculate and reduce (lcm/mcm) fractions:
Luxferre's Scoundrel C port can trivially ported to UXN and even maybe mu808. Eforth for sure, with cells and a minimal 'vector/array/' like implementation.