Science Is Open Software

Modern computational science is essentially open source software: software encodes the executable, testable models scientists use to predict and explain the world, so reproducibility requires runnable, inspectable, and modifiable code. Science builds internal models that must be shareable and improvable; if implementations are hidden or buggy, published results lose their testability and cannot be reliably incorporated into others’ mental models. Because research outcomes depend on software and its dependencies, opaque or fragile code can invalidate findings and stall cumulative progress.

Open source practices provide the transparency, modifiability, and communal scrutiny needed to make computational work scientific in practice. The envisioned future includes instantly reproducible papers running in preserved environments, community-maintained models that evolve like Wikipedia, faster discovery, and stronger public trust. Concrete recommendations are to always share and document code, prefer stable reproducible environments (the piece advocates NixOS for long-term bit-for-bit repeatability), build on existing community tools rather than reinventing them, and reward software development in academic evaluation. Making code open and reproducible is presented as a necessary condition for robust, testable, and cumulative computational science.