A method project built from practice.
Manuale LLM is not a generic introduction to AI. It is a practical attempt to turn repeated work with language models into an explicit method.
The project grows from a simple observation: a strong result does not come from “using AI” in the abstract. It comes from context, structure, constraints, review, iteration and the willingness to reject outputs that do not deserve to survive.
Why this project exists
Language models are often discussed in extremes. They are either treated as magical substitutes for competence or dismissed as empty noise. Neither approach is useful.
What matters in practice is the relationship between tool and user. An LLM can help write, analyse, compare, reorganise and accelerate. It does not choose the aim of the work, the standards of the work or the responsibility for the work.
That distinction is the centre of the manual.
What it covers
The repository presents a structured body of notes and chapters about how to work with language models more deliberately. Themes include:
- prompting as architecture rather than impulse;
- method, iteration and diagnostic use;
- responsibility, verification and human direction;
- using models for writing, structure and project support;
- extending capability without surrendering judgement.
The current public repository also includes appendices dedicated to text work, worldbuilding and image-related use.
Tool, responsibility, direction
The core principle of the manual is simple: a language model is a powerful instrument, but still an instrument.
It can amplify a process. It does not inherit accountability. Intent, precision, judgement, acceptance criteria and ethical responsibility remain human.
That is why the manual focuses not only on what a model can do, but on how a person should work when using one.
Project status
Manuale LLM is a public draft in active development. It should be read as an evolving method project rather than a finished institutional publication.
The repository is the current public reference point.