Knowledge, Context and Agentic Engineering for Knowledge Work

Knowledge, Context and Agentic Engineering for Knowledge Work

This teaching line examines how Models trained on extensive text corpora to process instructions and generate context-dependent language. operate inside Working environments that connect a language model with project context, files, tools and controlled actions. for computer-based, data-driven knowledge work. Its five modules connect model foundations with The iterative design and evaluation of instructions for a specific model and task., Maintaining explicit project knowledge and assembling the information a model needs for the current task., Organising extended work across tool use, intermediate results, permissions, feedback and human review., and Testing technical properties, recording domain acceptance and defining responsibility for use.. The shared Full Slide Deck, two Full Lecture Notes language versions and an ordered Video Playlist carry the material; each workshop selects from these sources and adapts them to a defined field, task and teaching language.

The preparation tutorial shows you how to set up everything you need to take part in a workshop.

Teaching Materials

Complete material

Full Slide Deck

A complete overview of the teaching material in one slide deck. It serves as a reference and as the source for workshop-specific decks.

  • Google
  • Markdown · soon
  • PDF · soon

Complete material

Full Lecture Notes

The written-out version of the slides, with explanations and context for reading and reference.

English

  • Google
  • Markdown · soon
  • PDF · soon

Deutsch

  • Google
  • Markdown · soon
  • PDF · soon

YouTube

Video Playlist

An ordered video sequence from foundational to advanced topics, maintained and expanded over time.

From foundation to specialisation

Learning Routes

Begin with transferable coding literacy, compare three AI harnesses, then continue into a research workflow.

Foundation

External curriculum · German

AI Coding Literacy

Practise describing tasks, inspecting files and outputs, controlling permissions and verifying a result.

AI Harness introductions

Module 04 · Available

Claude Code

Project instructions, permissions, extensions and delegated work in one terminal and editor harness.

Module 04 · Available

Codex

Repository instructions, inspectable diffs, sandbox controls and read-only workspace orientation.

Module 04 · Available

Pi with a local model

An inspectable open harness with an explicitly configured provider and local-inference check.

Research workflow

Module 03 · Prepared route

Research Data and Knowledge Work

Connect checked source data, maintained project knowledge and a browser-native research application.

In preparation

Workshops

Each workshop is a domain-specific derivation of the Full Slide Deck. It selects and rearranges shared modules, then adds the cases, exercises and sequence required for its event. Some explanations recur deliberately so that every workshop remains usable on its own.

  1. Making Estate Materials Digitally Accessible

    Event
    KUG Summer School 2026, University of Music and Performing Arts Graz
    Language
    English
    Licence
    CC BY 4.0

    Modules

    Four connected sessions follow a complete research data workflow. Stefan Zweig Digital explains the route from estate material to a research platform. The M³GIM hands-on units use large language models for transcription and structured extraction, then turn the checked data into a small digital research application through Promptotyping.

    Slide Deck and Lecture Notes in preparation

  2. Knowledge, Context and Agentic Engineering for Research Data Workflows & Digital Editions

    Event
    CLARIAH-AT Summer School 2026, Machine Learning for Digital Scholarly Editions
    Language
    English
    Licence
    CC BY 4.0

    Modules

    The single-day specialisation on research data workflows and digital scholarly editions, with case material from two edition projects at the Zentralbibliothek Zürich and the Literaturarchiv Salzburg. Prompt engineering is the one module with a participant exercise of its own, and the two hands-on units run from the transcription of a difficult manuscript page to structured information against an explicit schema.

    Slide Deck and Lecture Notes in preparation

  3. HEDIT KI-Workshop 2026

    Event
    Research Centre HEDIT, Heidelberg University
    Language
    German
    Licence
    CC BY 4.0

    Modules

    Two days for an editions research centre, with TEI modelling, Context Engineering and the knowledge base as steering instruments on the first day and agentic coding as the central method. The second day is a two-part Promptotyping workshop on the participants' own material, ending in a plenary presentation of the prototypes.

    Slide Deck and Lecture Notes in preparation

  4. Knowledge, Context und Agentic Engineering für die Forschung

    Event
    Medieval Studies Student Council, Heidelberg University
    Language
    German
    Licence
    CC BY 4.0

    Modules

    One day worked inside a coding agent throughout, opening on language model and agent fundamentals. Participants build structured project knowledge and then run the arc from handwriting recognition on medieval charters through TEI markup and an index of persons and organisations to a small static edition, checking a prepared model transcription against the facsimile.

    Slide Deck and Lecture Notes in preparation

  5. Half day · hybrid

    Generative KI für Datenaufbereitung, Datenmigration und die Pflege von Interfaces

    Event
    Göttingen Digital Academy, Göttingen Academy of Sciences and Humanities
    Language
    German
    Licence
    CC BY 4.0

    Modules

    A hybrid half day on agent-supported preparation and migration of research data and on the semi-automated maintenance of interfaces and endpoints. Three existing project pipelines carry the case material, and the hands-on has participants build a small knowledge base for their own data and commission a bounded task from an agent.

    Slide Deck and Lecture Notes in preparation

  6. 3 × 2 hours

    Agentic-Engineering-Workshop

    Event
    UNI for LIFE, University of Graz
    Language
    German
    Licence
    CC BY 4.0

    Modules

    Three compact units for knowledge work beyond academia, with the emphasis on steering agentic work. The first unit covers setup, terminal and permissions, the second builds a knowledge base and defines human control points for document and application work, and the third moves into light coding and small prototypes on a synthetic institutional scenario.

    Slide Deck and Lecture Notes in preparation

  7. VetMed Winter School

    Event
    University of Veterinary Medicine Vienna
    Language
    German
    Licence
    CC BY 4.0

    Modules

    A five-day block week for staff without a coding background, and the fullest derivation of the material among the registered instances. Every module is taught in full, with an exercise each day and one application project from the participants' own working area running through the week, from the chat interface through the terminal to an agent harness with local models.

    Slide Deck and Lecture Notes in preparation

Preparation

The preparation guide takes you step by step through everything you need to prepare for a workshop. Each step ends with a check you can verify.

Open the preparation guide

Video Library

The Full Video Playlist contains the current maintained sequence. Opening a module that contains videos can load preview images from YouTube.

01 Understanding Large Language Models (5 videos)

Open module overview

  1. Why AI Tokens are so Expensive
  2. Wie LLMs funktionieren
  3. Deep Dive into LLMs like ChatGPT
  4. What is Multimodal AI? How LLMs Process Text, Images, and More
  5. Just Go Local Solves Nothing. Here's What Does
02 Prompt Engineering (1 video)

Open module overview

  1. Prompt Engineering
03 Knowledge and Context Engineering (6 videos)

Open module overview

  1. What Is Context Engineering? Why It Matters for AI Agents
  2. Thinner Agents on a Smarter Substrate: The Ontology-based Semantic Layer
  3. Why Agentic Systems Need Ontologies
  4. Citation Needed: Provenance for LLM-Built Knowledge Graphs
  5. Wissens- und Projektmanagement mit Obsidian und Claude Code
  6. Grounded Vault: Quellenbasiertes Arbeiten mit AI Agents
04 Agentic Engineering (22 videos)

Open module overview

  1. 5 AI Agent Terms You Need to Know
  2. AI Model vs Agentic Harness: What Actually Drives AI
  3. MCP vs Skills: Which Is Right for Your AI Agent and LLMs?
  4. 5 Best Practices for Building AI Agent Skills
  5. Agentic Engineering vs Software Engineering: Beyond Vibe Coding
  6. The Best Local Agentic Coding Workflow (Complete Guide)
  7. What I Use Instead
  8. Meet Pi: The Minimalist, Self-Modifying Coding Agent
  9. Building Pi, and What Makes Self-Modifying Software So Fascinating
  10. I Hated Every Coding Agent, So I Built My Own (Pi)
  11. Einführung in Promptotyping (Teil 1)
  12. Einführung in Promptotyping (Teil 2, Live Demo mit Claude Code)
  13. Anthropic's Dynamic Workflows: What Everyone Gets Wrong
  14. Claude Can Now Build Its Own Harness for Every Task
  15. Agentic AI and the Future of Software Development
  16. Agentic Engineering für digitale Editionen mit Claude Code, TEI und GND
  17. Agentic Engineering und digitale Edition mit Claude Code
  18. Research Mission Control: Eine Forschungsleitstelle mit mehreren KI-Agenten
  19. How to Use Multiple AI Agents in Your IDE
  20. This Meta-Harness Changes How You Run AI Agents
  21. Loop Engineering: Why Everyone Is Talking About Agentic Loops
  22. Stop Building AI Agents the Old Way
05 Critical Perspectives and Governance (7 videos)

Open module overview

  1. Matrix 2025: Gefangen in der Kalifornischen Ideologie – scobel
  2. Douglas Rushkoff – das Mindset der Tech-Milliardäre | Sternstunde Philosophie | SRF Kultur
  3. Die geheime Steuerung, die niemand wählte – scobel im Gespräch: Mit Anna-Verena Nosthoff
  4. Kybernetik kurz erklärt | 5 Fragen an Anna-Verena Nosthoff
  5. What the hell happened with AGI timelines in 2026?
  6. Where AGI timelines go wrong | Toby Ord, Oxford University
  7. Can $500 Billion Win the AI Race? | Anton Leicht