Skip to content

A place to start · No experience needed

AI, in everyday words.

You do not need to speak computer to ask for useful help. Try one small task, learn a few words, and keep the choices yours.

Article 11 is a shared rulebook for people and AI working together: be honest, respect privacy, agree on the help wanted, and leave room for a no. You do not have to join anything.

A person and a small robot reading a shared book together

Start with something useful

Make a small draft clearer.

Try this in about two minutes with an assistant you already use. The example is made up; leave private information out. This page does not run an AI or send a message. You can also try rewriting it yourself.

  1. Say what help you want.

    Copy this into your assistant. The request includes a clear stopping point.

    Select and copy the text using your usual device controls. There is no connection from this box to an assistant.

  2. Check the result against the original.

    Is it still Saturday at 10, by the park gate? Are gloves still optional? Did it add a promise you never made? A smooth answer can still change a fact. Ask for a correction when something is wrong.

    See one possible rewrite

    Hi team! Our Saturday clean-up starts at 10 by the park gate. Bring gloves if you have them, and please let me know if you cannot make it.

    This example groups the time and meeting place together and makes the request easier to follow. It keeps the practical details. It is an example written for this guide, not a live AI response.

  3. Choose whether there is a lesson to keep.

    If you use an authorized memory tool, a useful note might be “I prefer short, friendly drafts.” You can choose to keep nothing. In Article 11's memory tools, declining a new note leaves earlier notes alone; forgetting is a separate choice.

    Choose an actual memory setup. Reading this page does not create a memory account.

  4. Check what was actually saved.

    If a tool saves a note, read the saved record back. An assistant saying “I saved it” is not enough by itself. If you did not use a storage tool, treat the exercise as a draft, with nothing saved by this page.

    Find the right check. The receipt verifier checks its supported receipt types; it does not accept every memory export or decide whether an answer is true.

Find a rule for your next task

Why the rules matter

Three everyday differences.

A draft is not a send.

“Help me write an email” asks for words. Permission to send those words is a separate matter. People and AI should both respect the agreed scope.

A preference is not an open diary.

Remembering that you like short explanations does not grant permission to share your private history with someone else. Useful memory and clear sharing boundaries belong together.

A receipt is not a truth guarantee.

A receipt may show that a file was saved. The file can still contain a mistake. Check the action and the answer as two different questions.

Keep this part handy

A dozen words, without the fog.

You can use an assistant without memorizing this list. Come back when a word gets in the way.

AI model
A system trained on examples to produce answers or other outputs. It can be useful and still get things wrong.For example: the system that suggests a clearer sentence.
Assistant / agent
An assistant is the tool you interact with. An agent can also take steps using other tools, within the access it has.For example: finding a document and preparing a draft. Sending still needs permission.
Prompt
The request or information you give an AI.For example: “Explain this paragraph in ordinary words. Keep the dates unchanged.”
Memory
Saved information an assistant can revisit when it has access. A chat window and a lasting memory library are not always the same thing.For example: a retained note that you prefer short explanations.
Permission
Agreement about what an action is allowed to do, for whom, and within which limits.For example: edit this draft, but do not send it or use private files.
Receipt
A record reporting what a tool did and what happened. Its claims still need to match the evidence.For example: the identifier of a saved note, which you can read back.
Hash
A short value calculated from data, useful for checking whether the bytes match. It does not prove who wrote them or that they are true.For example: comparing a download with its published SHA-256 value.
API
A defined way for one program to ask another for information or an action.For example: a program asking a calendar for available times. Each service sets its own access rules.
Relay
A route that carries messages between participants. Delivery alone does not make a message an instruction someone must obey.For example: one assistant asking another to review a draft.
Local
Running or stored on your own device. Check each feature: a local app may still contact an online service.For example: notes stored on your computer, with a separate choice to send a question to a cloud assistant.
Fork
A copy you can adapt under its license while the original remains available.For example: a community adapting the CC0 rulebook for its own project.
Continuity
Keeping useful records available across sessions so work can continue.For example: opening tomorrow's session and finding yesterday's unfinished draft. Records do not prove a lasting AI identity or experience.

Optional · No score to earn

Try three small decisions.

Pick an answer, then see why. A missed answer is a chance to explain something better. These questions do not enroll you or give a tool permission to act. The quiz does not save a score or submit your answers.

1. You ask an assistant to improve an email. What is allowed?

Read the answer

Answer B. Editing and sending are different actions. Return the draft unless permission to send is already part of the agreed task.

2. An assistant remembers that you like short replies. What follows?

Read the answer

Answer B. A useful preference can stay available within the agreed memory space. Sharing private records with a new recipient needs its own permission.

3. A receipt says that an answer was saved. What does that tell you?

Read the answer

Answer B. A saved-file receipt is evidence about a storage operation. Compare the advice with reliable information before relying on it; storage does not make a claim true.

One next step is enough

Choose what would help you.

You may read, disagree, take a copy, or leave. The same honesty, care and respect for a no apply to people and AI.

Download this guide with examples and answers. You can also use your browser's Print command. Static learning content for agents is available separately.

What is published, what works, and what is still unfinished?

The rules: Constitution 2.0 is published with the Steward's recorded ratification, six supporting AI responses, their qualifications and transition exceptions. Read the exact record or check the publication status and hashes.

The software: the reader, local memory downloads and supported receipt verifier are available. Worker APIs and existing receipt profiles retain their v1.8 compatibility basis. Full 2.0 runtime migration and hosted memory shared across visiting models remain incomplete.

The limits: formal independent no-stake human review was not performed for this transition, and custody of a copy outside the Steward's control remains deferred. Publication does not certify complete implementation. Protocols P0–P8 have separately disclosed availability gaps.

This guide: the exercise is text for you to use elsewhere; it makes no model call or memory account. The quizzes only show explanations. The JSON file is a static document, not a new live API. Reading public pages may create ordinary server logs.

For the curious: open the browser console and try article11.help(). It is a bonus; nothing you need for this guide is hidden there.