This repo hosts materials related to a presentation given on 25 June 2026 at the Beilstein Dialogues Symposium 2026 "Representation and Reality: Rethinking Scientific Models" at the Jagdschloss Niederwald near Rüdesheim, Germany. For some granular notes on the event, see https://w3id.org/spaces/BeilsteinDialogues2026. Title: How granular formalization transforms modeling into a social machine Abstract Traditionally, scientific models have been seen as coherent representations of natural systems – idealized in some ways but still mirrors of reality. Yet contemporary research practices are eroding this picture from multiple directions. Distributed computation, large-scale collaboration, and machine-readable infrastructures now allow knowledge to be continuously revised, recombined, and verified across disciplinary boundaries. This shift is changing not only how models are constructed, but the very nature of what a scientific model is.This transformation is driven by the growing formalization of knowledge into discrete, addressable and increasingly interoperable units. This granularity is becoming visible across diverse fields. For instance, in chemistry, reaction databases are organized as machine-readable graphs, where each unit of chemical interest is individually addressable. In biology, modular and version-controlled models of signaling pathways can be independently validated and recombined. In mathematics, proofs are formally verified, and each lemma used on the way is a stand-alone, executable object. In computer science, computational notebooks can blend narrative, code and data in complex and even interactive ways, turning scientific reasoning into a distributed, executable artifact.In all these cases, scientific claims – whether about parameters, observations, or logical assertions – no longer serve merely as supporting material for a final representation. Instead, they become composable, machine-actionable components of shared epistemic infrastructures. Such components can be linked, validated, reused, and extended within evolving research workflows, often across disciplinary boundaries.This development forces a fundamental philosophical reorientation. A model is no longer evaluated solely by its explanatory adequacy or empirical fit. It is also judged by additional characteristics like its interoperability, verifiability, its environmental impact and its participatory potential within a distributed network of human and machine agents. In this emerging landscape, modeling the world and modeling the knowledge process are converging, and modeling becomes a form of social machine building that I am looking forward to discuss with the other participants of the symposium.