> Okay, so let’s consider a different chart. We start by gathering the 15,000 most downloaded Python packages on PyPI in December 2025.2 Then we split the packages into cohorts based on their birth-year, and for each cohort we plot their median release frequency over time.3 This seems like a reasonable proxy measure of the production of real, actively-used software.

No. Many projects explicitly release on a fixed schedule. Even if you don't, you're going to get into a rhythm.

There's a deeper problem with using PyPi to measure the success of vibecoding: Libraries are more difficult to program then apps. Maybe vibecoding is a good way to create apps that solve some specific problem, but not to create generally useful libraries.