This is a low memory-usage median algorithm. It does not copy the data and then perform a quicksort or quickselect, but uses the actual data to calculate the median (without modifying the actual data). It works by starting with an estimated partition value at the mean value and calculate the number of values below and above that partition value. Further iterations recalculate a partitional value from the difference between the number of below and above counts, until this difference is small enough that single steps lead to the actual median value. The algorithm is coded in Rust, and tested with rustc version 1.23.0 and cargo version 0.24.0. It is provided as a Rust library project, with an executable for reading simple data files (with one plain number per line). A C implementation is provided as well. This work is licensed under the OpenBSD License.