Most the leading solutions are just overfit the test cases, and meaningless.

This problem seems weird, I've solved the last test with hacking after using harmonic mean for the others, because I've imagined the speed was constant. However, people who solved without a hack acted as the speed was decreasing. And that doesn't make sense to me (it is not written at the problem description). The problem description IMHO needs to be changed if that's what the author was aiming. If you are reading this before trying to solve it, please, think about exponential decay. It is not linear despite the problem not saying so. Without this previous sentence, one may lose time trying several different solutions (like some people did in the solution set) or just hack it instead of learning something new (just point people at the right direction; they still can try some other technique if they want). Especially since the goal in Cody is to create a small solution.

And I agree that more test cases should be added, besides improving its description. In Cody, there are broken tests. So we can never be sure at times wether our code or the test set is bogus.