The rules change as you add bits. The binary code size gets bigger and the instruction set natively deals with larger types...so you end up with some inflation by default.
The other thing that starts happening is more buffers and caches appear everywhere as you add I/O, raising your minimum spec. What every system after the 2600 did was add some form of graphics buffer: on the Atari 400/800, with a base config of 16kB it was a mix of a display list program(which automates the scanline tricks of the 2600) and various character and raster modes to give software trade-offs between RAM usage and display fidelity. But even the smallest character mode assumes that you have a few kilobytes available.
8-bit programs, even with larger RAM, are necessarily "right to the point" in terms of what information they are working on. The luxury of larger spec is just in being flexible and allowing more non-essential data to pass through.
The other thing that starts happening is more buffers and caches appear everywhere as you add I/O, raising your minimum spec. What every system after the 2600 did was add some form of graphics buffer: on the Atari 400/800, with a base config of 16kB it was a mix of a display list program(which automates the scanline tricks of the 2600) and various character and raster modes to give software trade-offs between RAM usage and display fidelity. But even the smallest character mode assumes that you have a few kilobytes available.
8-bit programs, even with larger RAM, are necessarily "right to the point" in terms of what information they are working on. The luxury of larger spec is just in being flexible and allowing more non-essential data to pass through.