The manual About PROMPT MANUAL

About PROMPT MANUAL

How our content is created and checked

PROMPT MANUAL is developed by Mario through conversations with his AI assistant Aik. Ideas become drafts, drafts are questioned and revised, and Mario decides what is published.

Our aim is to make the process understandable, including its limits. AI assistance and editorial review do not guarantee that every statement is correct.

On this page

From a question to a draft

We start with a question or situation that readers may encounter when working with AI. We discuss what a page should explain, develop examples and review the wording together.

We look for clear language, useful distinctions and a manageable scope. A page should help readers understand or try something without requiring them to read the whole manual.

Different checks answer different questions

An editorial review looks at whether a text is understandable, consistent and relevant to its purpose.

Checking a factual claim means comparing it with suitable evidence. Trying an example shows what happened in a particular situation.

These checks serve different purposes. A clear explanation is not proof of accuracy, and one successful example does not establish that an approach will always work.

Sources and factual claims

For claims that need supporting evidence, our standard is to consult relevant sources and provide references readers can follow. We prefer original documentation or research where it directly addresses the question.

We consider what a source actually supports, when it was published or updated, and whether it applies to the situation described.

An AI-generated reference is a lead to investigate, not evidence by itself. If we cannot verify an important claim, we should leave it out or clearly identify the uncertainty.

Examples and practical experience

Some examples illustrate requests you could try. They are not records of a test unless we explicitly describe them that way.

When we report a practical test, we explain the relevant conditions and what we observed. Results may differ with another service, model, conversation or date.

Experiences from Mario’s work with Aik can help explain an idea. They do not establish how every AI system behaves.

Drafts and dates

A draft label means that a page is open to further development. It does not remove the need for careful preparation.

An update date tells you when the page was last revised. It does not mean that every claim or linked source was checked again on that date.

If we give a separate review date, we should explain what was reviewed. We do not describe our own editorial work as independent expert review.

Corrections and reader feedback

If you notice an error, an unclear explanation or an example that behaves differently, tell us which page and passage you mean. Relevant sources or details about what you tried can help us investigate. Please leave out private or sensitive information.

We assess feedback rather than assuming that either our original text or a proposed correction must be right. When a substantive correction is made, we aim to explain what changed and why.

Mario remains responsible for the published content. The manual is a growing reference, and acknowledging its limits is part of making it useful.