It seems impossible that, given current economic trajectories, there won't be software archaeology in the future.
Our abusive relationship with modern commercialism has disintegrated the value of art, folklore and tools, both in the eyes of consumers and producers, and we no longer as a society seek to preserve the greatest works of today's most cutting-edge mediums. It's quite a sad state of affairs, which will only be mourned with increasing intensity as time marches on and curious researchers try to piece together the early days of the internet and personal computing.
In the more common case, it's simply archiving -- we still have hardware it can run on (e.g. x86).
The GPGPU and ML stuff is likely going to age poorer, although at least we had the sense to funnel most high level code through standardized APIs and libraries.
Making a competing standard and then having to deal with interop between standards is not simpler or better unless there's an actual benefit to be had.
It’s not hard to imagine a program of ever increasing complexity, highly redundant, error-prone, yet also stochastically, statically, increasingly effective.
The primary pushback from this would be our pathetically tiny short term memory. Blow that limitation up and complexity that seems incomprehensible to us becomes perfectly reasonable.
Situation that have 500 viewpoints and 10,000 variously interacting facts to consider won't be any more comprehensible for a mere human, short term memory limiting us or not.
Was the sarcasm lost on me?
Other things are quite advanced - but computing is rather "dumb". What is there is driven by a great weight of inertia of massive software (all written by regular humans) that itself is too large to fully audit and understand.
The story it tells is very much one about humans and societies.
We do this through categorization. For example, integrated circuits can be broken down into simplistic AND, OR, XOR, NOR, NAND, X OR, and NOt gates. Yet tying those together creates usable components such as Wifi radios and manufacturing automation.
Over simplification are concepts that lack understanding of complex systems. These are exploited by politicians. For example, deregulation will help drive the economy. Yet regulations have been proven to assist the economy long term by preventing the next variation of thalidomide. The panacea drug of the 1950s and 1960s that killed their users or caused massive birth defects.