A new paper titled Science Is Open Software is making the rounds, and it is arguing that modern science itself is nothing more than open-source software. The author, who spends a lot of time working on software, wants researchers everywhere to treat code the way they treat experiments — as a testable, replicable, and improvable foundation for knowledge.
The argument rests on a simple claim: if computational science depends on software, and if that software is opaque, broken, or closed off, then the science built on top of it is broken too. The author goes further — they argue that open-source software is the scientific method itself, just expressed in code instead of equations.
What Wikipedia Says About Science
The paper opens with a definition of science from Wikipedia: science is a systematic discipline that builds and organizes knowledge in the form of testable hypotheses and predictions about the universe. The author then asks readers to grab a random arXiv paper and see whether it actually meets that standard.
The answer is rarely yes. Most papers contain knowledge, but testing that knowledge — repeating the experiment, checking the assumptions, adapting the model — is harder than it looks. The paper frames this as a failure of reproducibility: if you cannot take an idea, test it, and build on it, then it is not science.
The Inner Model Argument
Kenneth James Williams Craik, in his book The Nature of Explanation, argued that organisms use small simulations of reality to explain and predict the world around them. The author takes this idea and applies it to scientific research: the goal of science is to improve your inner model of the world until you can make better predictions than before.
Software, in this view, is how we encode and share those predictive models. A paper that does not let you reproduce its analysis is not improving anyone’s inner model — it is just presenting a claim without a way to verify it.
What Reproducibility Actually Means
The paper distinguishes between reproducibility and replication. Replication is the duplication of results — running the same experiment again and getting the same numbers. Reproducibility is something broader: it is the ability to take the scientific idea, embed it into your own model, adapt it, and build upon it.
The author offers a thought experiment: imagine replacing “software model” with “mathematical model.” Just as we would not accept a physics paper that says its equations predict X but refuses to show the math, we should not accept computational science that hides its methods.
Why Software Matters for Science
Greg Wilson has noted that many fields have been held back, and many careers disrupted, because of a buggy program. The paper agrees — software is ubiquitous in modern science, from CoVid models to search algorithms to lab protocols. Researchers are busy people. They do not bother to look through all software dependencies to verify correctness, understand implementation details, or check for potential errors.
The consequence is that scientific results depend on the software. If the software is wrong, the science is wrong. Software bugs already cause numerous retractions, including several cases cited in the paper.
The Case for Open Source
The paper argues that open-source software solves the reliability problem by putting imperfections on public record. Code that is executable, modifiable, and transparent can be amended and improved, just like scientific understanding. The author frames this as the scientific method in simulation.
Open-source software already generates trillions in value, and there is room for much, much more. But the paper acknowledges that open source is not a perfect cure — there are IP and security concerns, bugs can still occur, and stability can be a problem.
A Vision for Future Science
The paper closes with a picture of what truly open computational science might look like. Every result is instantly reproducible — when you read a paper claiming a new drug reduces symptoms by 30%, you click a link and watch the exact analysis run in your browser. The data processing, statistical tests, and visualizations execute in seconds using the same environment.
Schedule Table: Key Points in Order
| Section | Topic |
|---|---|
| Introduction | Definition of science from Wikipedia |
| Inner model argument | Craik’s small simulations of reality |
| Reproducibility | Distinction between replication and reproducibility |
| Software dependence | Greg Wilson’s note on buggy programs |
| Open source case | Transparency, modifiability, and public record |
| Vision for science | Instantly reproducible results in the browser |
What We Make of It
The paper’s argument is a direct challenge to how science is currently practiced. It pushes back against the idea that software is a time sink — it is the foundation of modern research, and treating it as secondary is a mistake.
The vision of instantly reproducible results is appealing, but it is also far from where we are. The paper’s own concessions about IP, security, and stability suggest that the transition will not be simple.
The paper is an important move towards better science, but it is also a challenge to the status quo. It asks researchers to trust other people’s work, but only insofar as that work is open and inspectable.
The paper’s honesty about its own position is refreshing. It characterizes the post as challenging many of the current trends in academia, but it is an important move towards better science that doesn’t turn us all insane. That humility is part of the appeal — the author is not claiming a silver bullet, just a direction.
The paper’s final image of a browser-based reproduction of a paper’s results is a useful target. Whether we reach it is another question.
Source material: “Science Is Open Software,” jepedersen.dk.
Get the Notebook.
The day's best stories and every fresh verdict, in plain English, in your inbox by seven. One email a day, no more.