ai-publication
~/ai-mac-studio/publication/ is the publication registry for all scholarly output by Gerd Graßhoff (Humboldt-Universität zu Berlin, ORCID 0000-0002-4450-357X). All preprints target the Zenodo community grasshoff-scholarly-ai (not created yet) Zenodo community.
13 papers registered in catalog.yml across 5 themes · 1 arxiv-submitted · 12 draft
Platform & Infrastructure — 3 papers
A Hierarchical Modular Skill System
Full title: A Hierarchical Modular Skill System — Declarative Agent Knowledge for a Scholarly AI Platform Across 42 Coordinated Projects
Presents a hierarchical modular skill system for a scholarly AI platform spanning 42 coordinated projects. Skills are declarative Markdown files that give Cursor AI agents up-to-date, actionable knowledge about tools, workflows, and conventions. Three tiers: general operation space, project-level skills, and machine-local skills.
Source: ai-agents/infrastructure/ai-skills/arxiv/main.tex · arXiv: — · DOI: —
A Visionary AI-Optimized Working Platform
Full title: A Visionary AI-Optimized Working Platform — Transparent Data, Declarative Skills, and Natural-Language Operation across 35+ Coordinated AI Tools
Describes an AI-optimized working platform integrating over 35 coordinated tools for scholarly research. The platform uses transparent data models, declarative skill definitions, and natural-language operation. Key components: voice memo processing, AI-assisted correspondence, document OCR, and a web dashboard.
Source: ai-agents/management/ai-voicememo/arxiv/main.tex · arXiv: — · DOI: —
Managing Scholarly Output Across a Coordinated AI Research Platform
Full title: Managing Scholarly Output Across a Coordinated AI Research Platform — The Grasshoff Arxiv System
Describes the Grasshoff Arxiv System — an integrated platform for managing scholarly preprints across a coordinated multi-project environment. Covers the central catalog, automated Zenodo community publishing, VG-Wort METIS pixel counter management, and the three-tier publication workflow: local draft, Zenodo deposition, and arxiv.org submission.
Source: ai-publication/arxiv/ai-arxiv-project/arxiv/main.tex · arXiv: — · DOI: —
PDAI Document Processing Pipeline — 3 papers
PDAI: A Page-Centric Document Format
Full title: PDAI: A Page-Centric Document Format and Processing Pipeline for AI-Driven Scholarly Archive Processing
Presents PDAI (Page-Document AI), a page-centric document format with a strict two-layer architecture: an immutable page layer (OCR XML) and a derived segment layer (logical units with page provenance). The format is supported by a DTD schema hierarchy, Jinja2 instruction templates, and a two-tool pipeline: ai-scholar for OCR/structure/translation and ai-classifypdf for PDF harvesting/classification. Demonstrated on 8 publications totalling 2,152 pages across 5 centuries and 4 languages.
Source: ai-agents/document-processing/ai-scholar/arxiv/pdai-document-structure.qmd · arXiv: — · DOI: —
AI Scholar: From PDF to Portable Document AI Infrastructure
Full title: AI Scholar: From PDF to Portable Document AI Infrastructure — A Three-Package Architecture for Scholarly Document Processing in the Age of Large Language Models
Describes the three-package architecture (ai-scholar, ai-viewer, ai-classifypdf) for end-to-end scholarly document processing. From raw PDFs to structured PDAI XML with OCR, bibliographic extraction, entity recognition, and translation. Supports large-scale processing via Tailscale-connected cluster dispatch across four machines.
Source: ai-scholarly/research/ai-scholar/arxiv/ai-scholar-grasshoff.qmd · arXiv: — · DOI: —
A Multi-Layer Index and Retrieval Architecture for Historical Document Collections
Full title: A Multi-Layer Index and Retrieval Architecture for Historical Document Collections — DuckDB-Backed Entity Extraction, Token-Inverted Indices, and an OED-Style Concept Model Grounded in Wittgensteinian Use-Theory
A six-layer indexing architecture combining DuckDB-backed entity extraction, token-inverted indices for sub-millisecond retrieval, and an OED-style concept model grounded in Wittgensteinian use-theory. Enables deterministic, reproducible search across 10,000+ page-level XML documents — demonstrated on the Eneström corpus of Euler’s Opera Omnia. Living document, evolves with ai-index package.
Source: ai-agents/data-ml/ai-index/arxiv/main.qmd · arXiv: — · DOI: —
Scientific Communication & Epistemology — 3 papers
AI-Reporter: A Path to a New Genre of Scientific Communication ★
Full title: AI-Reporter: A Path to a New Genre of Scientific Communication — From Presentation to Publication through Agentic LLMs
Traces the historical development of scientific communication from manuscript to print through Regiomontanus, Copernicus, Kepler, and the Royal Society. The AI-Reporter continues this trajectory, transforming multimodal presentations into globally accessible, assessment-ready records within minutes. Published on arXiv July 8, 2025. Revision in preparation (author name fix, figure improvements, arXiv compatibility).
Source: ai-scholarly/research/ai-reporter/arxiv/ai-reporter-grasshoff.qmd · arXiv: 2507.05903 · DOI: — · Revision: ai-reporter-revision-final.tar.gz ready for upload
Epistemic Systems and Communicative Action
Full title: Epistemic Systems and Communicative Action: AI-Mediated Modelling of Agenda Evolution through Email Correspondence
Subtitle: A Computational Framework for Reconstructing Scientific Discovery Processes from Documentary Evidence
Presents the concept of epistemic systems — a novel theoretical framework for modelling scientific discovery — applied to email correspondence analysis. The computational system Mail Kanzlei operationalises Actor, Agenda, and Communicative Act to model how scholarly agendas evolve through correspondence. Uses LLMs with DTD-controlled structured XML output to extract intentional structure from correspondence. First validated in the computational reconstruction of the discovery of the urea cycle.
Source: ai-agents/communication/ai-mail/arxiv/epistemic-mail-grasshoff.qmd · arXiv: — · DOI: —
Computational Epistemology: Foundations for the Assessment of Propositional Validity in AI Systems
Subtitle: Why Retrieval Cannot Establish Truth and What Must Replace It
Foundations for a computational epistemology addressing how AI systems can assess propositional validity. Argues that retrieval-augmented generation cannot establish truth and proposes epistemic validation frameworks grounded in formal epistemology and the assessment practices of historical science. Developed in connection with the Mail Kanzlei epistemic modelling system.
Source: ai-scholarly/research/computationalEpistemology/arxiv/computational-epistemology-outline.qmd · Status: outline only
Ancient Sundials — 4 papers
Construction Geometry of the Conic Sundial with Horizontal Gnomon
Subtitle: Complete Analytical Framework for Shadow Projection on an Inclined Cone
A complete analytical framework for the construction geometry of conic sundials with horizontal gnomon. Derives shadow projection equations for an inclined cone, establishes conditions for hour-line placement, and validates the geometry against 3D scan data from surviving ancient sundial fragments. The mathematical foundation for the ai-sundial3d reconstruction pipeline.
Source: ai-scholarly/research/ai-sundial3d/arxiv/conic-sundial-geometry.qmd · arXiv: — · DOI: —
Automated Feature Detection on Ancient Conic Sundial Fragments
Subtitle: From Groove Geometry to Line Parameters via Cross-Sectional Profile Analysis
An automated pipeline for detecting and characterising groove geometry on ancient conic sundial fragments from 3D scan data. Extracts cross-sectional profiles, fits parametric line models to groove edges, and produces structured output for the ai-sundial3d reconstruction workflow. Validated against manually annotated reference fragments.
Source: ai-scholarly/ai-sundial3d/arxiv/automated-feature-detection/automated-feature-detection.qmd · arXiv: —
Automated Model Optimization for Sundial Feature Detectors
Subtitle: Nonlinear Parameter Estimation for Conic Sundial Reconstruction from 3D Scan Data
Nonlinear parameter estimation for optimising feature detector models in conic sundial reconstruction. A gradient-free optimisation framework tunes detector parameters against annotated 3D scan ground truth, achieving robust performance across diverse sundial fragment geometries.
Source: ai-scholarly/ai-sundial3d/arxiv/model-optimization/model-optimization.qmd · arXiv: —
Literate Coding with AI Agents
Full title: Literate Coding with AI Agents: A Methodology for Computational Analysis of Scientific Objects — Complexity Reduction through Structured Context, Executable Specification, and Iterative Refinement
A methodology for computational analysis of complex scientific objects using AI agents in a literate coding paradigm. Integrates research writing and AI-assisted development in a closed feedback loop, reducing complexity through structured context, executable specification, and iterative refinement. Demonstrated on the ai-sundial3d reconstruction pipeline as a case study.
Source: ai-scholarly/research/ai-sundial3d/arxiv/literate-coding-methodology.qmd · arXiv: —
Summary Table
| # | Short Title | Theme | Date | Status | |
|---|---|---|---|---|---|
| 1 | A Hierarchical Modular Skill System | platform | 2026-03 | draft | — |
| 2 | A Visionary AI-Optimized Working Platform | platform | 2026-03 | draft | |
| 3 | Managing Scholarly Output (Grasshoff Arxiv) | platform | 2026-03 | draft | — |
| 4 | PDAI: A Page-Centric Document Format | pdai | 2026-03-05 | draft | — |
| 5 | AI Scholar: From PDF to Document AI | pdai | 2025-11 | draft | |
| 6 | Multi-Layer Index and Retrieval Architecture | pdai | 2026-03 | draft | |
| 7 | AI-Reporter | communication | 2025-07-08 | arxiv-submitted | |
| 8 | Epistemic Systems and Email Correspondence | epistemology | 2025-01 | draft | |
| 9 | Computational Epistemology | epistemology | — | outline | — |
| 10 | Conic Sundial Construction Geometry | sundial | 2026-02-15 | draft | |
| 11 | Automated Feature Detection on Sundials | sundial | 2026-03-30 | draft | — |
| 12 | Automated Model Optimization for Sundials | sundial | 2026-03-30 | draft | — |
| 13 | Literate Coding with AI Agents | methodology | 2026-04-02 | draft | — |
Registered Paper Directories (arxiv/papers/)
Symlinks in ~/ai-mac-studio/publication/arxiv/papers/ linking to paper dirs in ai-scholarly/:
| Directory | Target |
|---|---|
ai-sundial3d |
~/ai-mac-studio/scholarly/ai-sundial3d/arxiv/ |
ai-sundial3d-research |
~/ai-mac-studio/scholarly/research/ai-sundial3d/arxiv/ |
ai-scholar |
~/ai-mac-studio/scholarly/research/ai-scholar/arxiv/ |
ai-reporter |
~/ai-mac-studio/scholarly/research/ai-reporter/arxiv/ |
computationalEpistemology |
~/ai-mac-studio/scholarly/research/computationalEpistemology/arxiv/ |
ai-nepi2 |
~/ai-mac-studio/scholarly/projects/ai-nepi2/communication/arxiv/ |
Zenodo Community & Publication Workflow
Community: grasshoff-scholarly-ai (not created yet) · Author: Gerd Graßhoff · ORCID: 0000-0002-4450-357X · Affiliation: Humboldt-Universität zu Berlin
| Step | Tool | Action |
|---|---|---|
| 1. Draft | LaTeX / Typst / Quarto | Write in project/arxiv/main.* |
| 2. Register | Edit catalog.yml |
Add entry with status: draft |
| 3. Symlink | ln -s |
Add to papers/ in ai-publication |
| 4. Zenodo | tools/zenodo_publish.py |
Deposit → get DOI → update catalog.yml |
| 5. arXiv | Manual upload | Set arxiv_id in catalog.yml |
| 6. VG-Wort | METIS pixel counter | Register for royalties |
| Badge | Meaning |
|---|---|
| draft | Local draft, not yet deposited |
| zenodo-published | Deposited on Zenodo, DOI assigned |
| arxiv-submitted | Submitted to arXiv.org |
| Compiled PDF exists locally |
Research data lives in ai-scholarly. For the PDAI pipeline producing source data, see PDAI Collections.