In general, the courts distinguish between functions and algorithms, and they require patent applicants to disclose algorithms to cure perceived deficiencies in functions. The problem with this line of reasoning is that both algorithms and functions under 35 U.S.C. § 112(f) are composed of the same things: steps. So the result of the algorithm requirement is to simply make patent applicants “fix” one step by specifying more steps. Accordingly, if the algorithm requirement is taken to its logical conclusion, then each step would be fixed with more steps, and each of those steps would be fixed with even more steps, like Russian dolls. Instead, the courts do not take the algorithm to its logical conclusion and, instead, only require a single layer: the original step and the further steps (i.e., algorithm) for it. This is arbitrary, confusing for patent applicants and examiners, and a poorly calibrated solution to concerns about software patents.