Part 1 runs from 04:00 to 06:00 UTC. It needs no programming: everything is done in a chat window.

Learning outcomes

After Part 1 you can:

1. Explain next-token prediction, tokens, context window, parameters and quantization in plain words, and connect them to file size, memory and speed.
2. Place a model on the four-tier openness ladder and say why libraries care (provenance, reproducibility, vendor independence).
3. Install Ollama or LM Studio, adjust the privacy settings, download a model that fits your hardware, and check that it works.
4. Write prompts for cataloging-style tasks and improve them step by step.
5. Recognize a confident wrong answer and know two checks for it.

Timeline

Minutes are counted from 04:00 UTC.

Minutes	Segment	Mode
0-10	Welcome: what changed since 2024, why libraries care, how the four hours work, how to get help (shared document, Zoom reactions)	demo
10-35	How an LLM works: next-token prediction; tokens (live tokenizer, multilingual costs); context window and its memory cost; parameters and quantization (file size equals memory); live ollama ps	demo with prediction checks
35-55	Open source, open weights, fully open: the ladder; Apertus as the case; the EU public summary; “was my library in the training data?”	demo
55-80	Setup recap and privacy settings; first chat with Apertus in the Ollama app or LM Studio; choosing a model size; measure your own speed	type-along (GUI)
80-110	Prompting for library tasks: anatomy, iterative refinement, few-shot classification, structured answers, multilingual prompts, the hallucination probe	exercise (three tiers)
110-120	Recap, self-check, what Part 2 assumes, how to leave or join at the break	demo

Three ways to take part

Follow along with the graphical tools only, run the notebooks in Part 2, or read the executed notebooks afterwards. An executed notebook has already been run and shows its outputs. The workshop page says what each needs.

The chapters