The description should make clear that the final chunk is implicitly padded with zeros if it's length is less than the chunk_size. This can be figured out from the sample test cases, but it is difficult to see because there is also the zeroing behavior from the all-ones index. (I initially thought that the final chunk's actual size is used for it's chunk_size, since that leads to the same behavior in certain cases)

In the Python translation, the random test cases require you to handle chunk_size > 15, but the description and the C++ cases say that should be rejected by returning false. This should either be removed from the Python description (keeping past solutions valid) or the test cases should change to be the same as the C++ cases.