I have very little experience in programming. My observation though is that software like Firefox takes way to much memory these days. When it does the CPU usage is also very high. Again I have very little experience but should there not exist a memory compression option that implements compression automatically when compiling software? With this I would not need compression like zram on boxes that doesn’t have Linux installed. If the compression method has many parameters maybe there is a possibility to use machine learning to get the right balance.
Another idea would be. Inspired by error-correcting code memory (ECC memory). When data corruption is not acceptable at all. For instance in financial computing. Why not use RAM compression where you use a checksum and data repair. Compressing data would also allow smaller memory devices that might not be so sensitive giving errors.