Greg WilsonOct 28
Bridgeford et al 2025: "Ten Simple Rules for AI-Assisted Coding in Science" https://arxiv.org/abs/2510.22254 #nwit
Ten Simple Rules for AI-Assisted Coding in Science

While AI coding tools have demonstrated potential to accelerate software development, their use in scientific computing raises critical questions about code quality and scientific validity. In this paper, we provide ten practical rules for AI-assisted coding that balance leveraging capabilities of AI with maintaining scientific and methodological rigor. We address how AI can be leveraged strategically throughout the development cycle with four key themes: problem preparation and understanding, managing context and interaction, testing and validation, and code quality assurance and iterative improvement. These principles serve to emphasize maintaining human agency in coding decisions, establishing robust validation procedures, and preserving the domain expertise essential for methodologically sound research. These rules are intended to help researchers harness AI's transformative potential for faster software development while ensuring that their code meets the standards of reliability, reproducibility, and scientific validity that research integrity demands.

arXiv.org
Show threadzenkat
@gvwilson Excellent and well-balanced perspective. The authors understand the strengths and weaknesses of DevAI tools, and provide strategies to leverage the strengths while mitigating the weaknesses.
Oct 28 at 1:54pmWeb
Show threadzenkatOct 28
@gvwilson The key insight -- one that every user of these tools should keep in mind -- is that LLMs do *not* have a mental model of the problem domain or the solution space. They are good at summarizing, predicting, and reasoning over local context windows. But they do not "think", and pretending that they do will get you into all sorts of trouble.