Much of what makes science work, such as the assumptions behind a hypothesis, the rationale behind a study design, the conceptual choices behind a statistical model, lives implicitly in researchers' minds and in the practices of their communities. In this talk, I will share a vision our lab is working towards: a richer comptuational substrate that supports novel reasoning and discourse about scientific validity, feasibility, and priority. A key design challenge is how to represent and interact with this hidden reasoning and, in doing so, make scientific research more robust. This is particularly important now, as AI is increasingly incorporated into scientific discovery.