We want to have one or more entities which are AIs.  That is they
possess computational abilities in excess of current computer
software.  They may exist in any number of forms, either silicon or
another such thing.  How they came into being is a question.  Seems
like they would have been invented.  There will be no time travel or
class M planets or other soft sci-fi, it will be all hard sci fi.
Basically, most sci-fi estimates of the future greatly exaggerate
certain technologies and conversely understimate others.  The key will
be predicting which areas are accelerated.  I do not like the
anthropomorphic AIs as they are not needed and they also create in the
minds of a certain type of person a confrontation or inevitable
conflict with ordinary humans.  Also, perhaps this AI is from a
different planet, and if it is there might be other forms of life
besides AI near or with it.  If it is from an alien planet, unless you
suppose there are absolute truths, moral standards might be strange or
incommunicable, something lost in translation.  Egoistic conflict
should be subsumed by a love of knowledge.  Perhaps in keeping with
realism we could have it be in a reasonably distant earth-future.
True altruism should dominate.  A strong sense of purpose and
seriousness should exist.  The AI could be based to some extent on the
same principles as the FRDCSA, namely a litmus test of theorem proving
for enhancing survival.  Questions of religion should be avoided.
Perhaps if there is a god, there is mechanism design in that escape
from destruction by creating backup "civilizations" (in case of
planetary or solar collapse or destruction) can only be unlocked by
conformance to certain normative standards.  We should employ actual
theorem proving and other planning for plot development.  Other
computational storytelling techniques.  Like Kurosawa.  We don't have
to have the actual text generated, just the logic of the world and or
plot should be modelled.  How long the text should be is open for
discussion.  There should be the potential for revision.  Storage
should be in a git repository.  This will be a use case for developing
these tools.  Given how alien this is to our experience, how do we
relate it to the goals and interests of people alive today.  Owing to
existential risk due to plantary / solar destruction, space flight
might be a necessary part.  It would be soft-sci-fi to presume some
kind of mechanism like two-eyes in Go which operates transcendently to
spacetime.  That would seem like too strong a claim.  Perhaps though a
search for such a thing could exist.  Security should be emphasized.
We might estimate that some of our current mathematical technology
would be exceeded by other more fruitful and less obvious lines of
inquery, in the same way that CMU students told me that a lot has
happened since the original proof / model / recursion theory.  One of
the faults of ready player one was that it was too "isomorphic" to the
real world, for instance they had to run to get to places, instead of
just got there.  If there are humans, mind to machine interfaces could
be involved.  Now, just as Tolkien invented metaphysical principles
for his world, perhaps we could invent some of our own, although,
again, that puts us into a certain kind of soft-ai.  We should map out
which assumptions can be thought of as leading to what worlds.
Perhaps we should have a choose your own world system.  No need to
relegate to existing short story novel form.  We have much more
abilities these days.  The notion of making a world with a large
history might be too desirous of emulating epic high fantasy.
Generally the conflict in epic high fantasy should be eshewed.  This
should not be about survival of the fittest.  This should be about
security for sentient beings.  Also the moral status of inanimate
things should also be pondered.  That calls to mind the ethical
systems analyzer.  Perhaps this story could be a sort of anthem for
the project, and actually use its universe to expedite the actual AI.
Of course, good sci-fi would be a gateway to stimulate interest.  That
would be another point of it, to help kindle productivity in others
and myself, just as Ender's Game and the Matrix tend to do.  We need
to model axiomatically the assumptions of the universe, the different
theories, and see what follows.  Could be done in that way, a process
of CSP or theorem proving.  Kind of like that system
assumption-junction.  The story should unite the different modules,
including the suppositional reasoner.  In fact, a lot of the
technologies I have developed I could write about the hypothetical
culmination of them and then the successor projects.  This will double
to help us identify a vision for each project.