Between 2026-04-19 and 2026-04-21, the sole human author of this paper, working as the operator-and-co-author in an interactive Claude-Code-agent-assisted loop, caused seven independent research papers to be drafted, figures-generated, metadata-written, PDF-rendered, and published with DOIs on Zenodo. The seven papers span three distinct subject domains (systems infrastructure; legacy-CCTV face-recognition retrofit; AYUSH/FSSAI regulatory data), four authorship configurations, two licence regimes (CC BY 4.0 and CC BY-NC 4.0), and two distinct correction lifecycles. Across the window, the pipeline produced 10 Zenodo records (7 papers, 2 companion datasets, 1 correction version), approximately 36,400 body words, 15 figures, 103 references, and 86 pages of typeset material. We report this not as promotional copy but as a case study, under the simple observation that the research-publication community has been extensively discussing agent-assisted writing without having many openly-documented examples of agent-assisted publication end-to-end. We describe the pipeline architecture, the per-paper templates that emerged, the execution timeline, the specific errors that occurred and how they were corrected, and the uncomfortable questions the process surfaces about authorship norms, peer-review readiness, citation velocity, and the appropriate frame for the human author's contribution. No claim is made that this pipeline replaces peer-reviewed venues; all seven papers remain preprints. The aim is a frank technical and ethical record.
Vibhav Aggarwal (Tue,) studied this question.