Modern science is synonymous with open source software, according to this compelling argument. While academia often dismisses software as a time sink overshadowed by the pressure to publish, the author contends that open source software is computational science itself. Science, by definition, builds testable hypotheses and predictions. A paper that doesn't help readers predict or model the world fails this fundamental test. Software is how we encode and share these predictive models.

Drawing on Kenneth James Williams Craik's concept of "inner models," the author explains that scientists improve internal simulations of reality to make better predictions. If an arXiv paper doesn't aid this process, it isn't truly science. Computational reproducibility is therefore essential—not merely duplicating results, but enabling others to embed, adapt, and build upon scientific ideas. Unreproducible findings are useless because they cannot be expanded.

As Greg Wilson noted, software is ubiquitous in modern science, from Covid models to lab protocols. Researchers rarely verify every dependency, making open, reproducible code critical. Insisting on scientific method compliance in software development is a vital move toward better, more sane science that doesn't hold back entire fields or disrupt careers.