The practical constraints that long sustained the qualitative-quantitative divide in social science—scarce data, inaccessible tools, and narrow technical expertise—are dissolving. Digitization, advances in natural language processing and machine learning, open-source software, and generative AI now place rich data from natural social settings, and the methods to analyze it, within reach of nearly any researcher. The binding limitation is no longer access to data or tools but the substantive understanding needed to use them—and the cost of collaborating at the scale that complex, linked data demands. These coordination costs—the learning costs of onboarding contributors and the attention costs of tracking their work—grow superlinearly with a project's size, a problem generative AI amplifies rather than solves.

We propose a framework pairing Domain-Driven Design (DDD) with a layered "Ports and Adapters" architecture: DDD organizes a project around the concepts domain experts already hold, reducing learning costs from a function of project size to one of domain complexity, while Ports and Adapters decouples components to localize attention and contain cascading failures. Both rest on qualitative methods—interviews, ubiquitous language, and domain modeling—making qualitative rigour load-bearing rather than ancillary. We recast these costs in tokens, the observable successor to Brooks's person-months, and introduce AI agents as reproducible instruments for testing the framework on PATRON, a live political-data infrastructure.

The advance of data and computing does not displace qualitative scholarship: there can be no quantitative turn in political science without a qualitative turn in quantitative methods.