Leading questions:
Representative vs Direct Democracy?
Unitary or Federal?
Presidential or Parliamentary?
How much separations of powers should there be? In presidential systems, such as the United States of America, there is often deadlock between the executive and legislature. In parliamentary systems, the head of government is elected by legislature, therefore, there is practically no deadlock as long as theres is majority support of the executive in the legislature (although, there can still be courts to determine constitutionality of policiss). Would you prefer more checks and balances, but can also result in more deadlock, or a government more easily able to enact policies, for better or for worse?
Electoral method? FPTP? Two-Round? Ranked-Choice/Single-Transferable Vote? What about legislature? Should there be local districts? Single or Multi member districts? Proportional-representation based on votes for a party? If so, how should the party-lists be determined?
Should anti-democratic parties be banned? Or should all parties be allowed to compete in elections, regardless of ideology? In Germany, they practice what's called "Defensive Democracy" which bans any political parties (and their successors) that are anti-democratic. Some of banned political parties include the nazi party.
How easy or difficult should the constitution br allowed to be changed? Majority support or some type of supermajority support?
Should we really elect officials, or randomly select them via sortition?
These are just some topics to think about, you don't have to answer all of them.
Edit: Clarified some things
My goal here was to make that irrelevant. Districts emerge naturally rather than being researched and drawn. Yes there's complexity, but not in the surface. On surface it's just a paint-by-numbers of the area around your house.
Pretty much. Optimization algorithms are powerful. By ranking addresses like a topographical map rather than just drawing homogenous districts, you automatically have the data needed to refine the resulting districts. The whole system basically amounts to a matrix of weighted preferences, which is the one main thing current "AI" is actually good at resolving.
They already aren't. Like I said, some neighborhoods will be bigger than others, so some boroughs will have more neighborhoods than others, by the time you get to the county level every district should be almost exactly the same. At the neighborhood level, equal sizes is less important than geographic relevance. By the scale that equal representation matters, you have it.
I can't claim the system is perfect, but it's hard to imagine a better one. Every model has limitations.