## Notes from 22 May 2025 [[2025-05-21|← Previous note]] ┃ [[2025-05-23|Next note →]] On 20–21 May, I attended a [[COREX]] meeting in Bratislava, my first time participating in one of the group's activities. In light of [Alexey Guzey](https://guzey.com/what-im-thinking-about/)’s recent newsletter, which linked to the statement that '[the "it" in AI models is the dataset](https://nonint.com/2023/06/10/the-it-in-ai-models-is-the-dataset/?utm_source=substack&utm_medium=email)', what struck me most was how the COREX initiative is quietly building something rare: comparative data on the institutional configuration of top-level administrative roles and the career paths leading to them. What kinds of positions exist at the intersection of politics, advisory roles, and senior bureaucracy? Who selects for them, based on what criteria, and how do political and administrative logics interact in these choices? This kind of data doesn’t yet exist in a structured, usable form. Over the two days, I witnessed firsthand the difficulty and necessity of creating a database that is comparable across contexts and usable by others. Today, academic incentives still reward novel arguments and publications more than foundational work such as database construction or replication studies. However, as with AI, where model performance often depends more on the dataset than the architecture, it is evident that the future of science, particularly social science, will increasingly rely on these undervalued forms of knowledge production.