There's always gerrymandering as a concept: If you build everything by algorithm, you are making the problem far smaller, but ultimately the decision of which algorithm to use is never going to be as good as proportional representation. Whether one only looks at population density and geography, or one adds ethnicities or party affiliations, there is no such thing as an objective algorithm that will not lead to biased results. The results will be more fair under any definition than the gerrymandering in, say, North Carolina, but they are still not going to be wonderful.
Given how the US political landscape is, it's not difficult to build a generic system that produces results that will favor the same party in all states. Compactness will, in general, be closer to favoring Republicans, while something that aimed for even population density in each district would favor democrats.
If I had to choose between any algorithm and what we have now, I'd pick the algorithm, but it's still going to be a worse output than someone actively redistricting aiming for maximally competitive elections or a proportional system.