AI‑assisted coding has matured to everyday usability among professional programmers, lowering the barrier for digital humanists: less time on syntax and framework details, more on framing research problems. Yet the opportunity exposes a gap: robust, maintainable software still depends on a grasp of software‑engineering fundamentals. At the same time, engineering practice is shifting under new AI paradigms. This workshop advances documentation‑oriented programming (DOP) as a practical foundation.

The workshop starts with a 15‑minute framing followed by three hands‑on stages in a single iterative track. Stage 1 turns a concise specification into a simple TypeScript/JavaScript script. Stage 2 refactors it into a small reusable library and integrates it as the core of a Zotero plugin. Stage 3 publishes the same core as a lightweight browser‑side library with a minimal web interface. A prepared repository includes starter code, minimal tests, and instructions; widely available models ensure accessibility.

The workshop situates AI coding within a broader text‑first shift: complex GUIs, RAG pipelines, and graph‑based builders give way to document‑driven workflows requiring less effort but more productivity. The Model Context Protocol (MCP) exemplifies a documentation‑defined, middle‑ground interface where agents act on descriptions in natural language while remaining predictable. Workshop participants will leave with up‑to‑date paradigms, practical vocabulary, and confidence to build small, durable tools and evolve them into proper software projects.