Your First Conversation

Choose the context

Open the intended project or conversation. Select an appropriate available agent and model when those controls are exposed.

State the task

Describe what you need, the constraints, relevant context, and the desired output format. For example: compare two approaches, state assumptions, and identify what remains uncertain.

Review the response

Wait for completion, then check source indicators, limitations, and important factual claims. A fluent answer is not proof of a completed operation.

Continue deliberately

Keep ongoing work in the appropriate project and verify saved history. Record durable guidance explicitly in the relevant memory scope.

Write a prompt that can be checked

Use four parts: goal, context, constraints, and output. For example: “Turn these three practice notes into a five-item checklist. Use only the notes. Mark missing owners or dates as unknown.” Review the answer, give one correction, and ask for a revised version.

Ask Friday to guide you

In an authorized FridayLocalAI conversation, try:

Help me turn my goal into a clear FridayLocalAI prompt. Ask for the goal, context, constraints, and desired output. Then guide me through reviewing the answer without assuming that fluent text is correct.

Tell Friday what you can see and which step you have reached. If the instructions do not match your screen, say so; ask for the relevant Help topic or contact your operator. Friday cannot be assumed to see your screen or know your current settings.

Steps to follow

1. Open the intended conversation in Organizer.
2. Enter a goal and the small amount of context needed to answer it. Use Shift+Enter for a new line; Enter sends the message.
3. Wait for the reply. Compare it with your inputs and look for unsupported details.
4. Ask for one specific revision, such as a shorter checklist or a correction.
5. Give the conversation a useful name using Rename conversation where available.

Choose one outcome

Before writing a prompt, decide what you want to have when the exchange is finished. An outline, a comparison table, a corrected paragraph, and an explanation are different outcomes. If you ask for all of them at once without enough context, it becomes difficult to decide whether the answer is useful. Start with one result and build from it. A good first task is small enough that you can check it yourself. Ask Friday to organize a short list of notes or explain a concept using an example you provide. Avoid making the first conversation depend on a large document, a remote service, or a system change. Learning the basic interaction is easier when the input and expected result are both visible and uncomplicated.

Give context that changes the answer

Context is useful when it helps Friday make a relevant choice. State the audience, purpose, constraints, and source material needed for the task. “Write an outline for a five-minute introduction aimed at new users” is more informative than “write about AI.” You can also state what to exclude, such as unsupported release dates or claims that a feature is available everywhere. Do not paste unrelated background merely because it exists. A focused prompt helps you notice whether the response used the information that matters. If important facts are missing, ask Friday to question you before drafting. This is especially useful when the answer depends on your preferences or circumstances, which the application cannot be assumed to know without context.

Make the output easy to review

Choose a format that supports the task. A short numbered list works for steps. A table can make a comparison easier to scan. A paragraph may be better for an explanation that needs connected reasoning. State any practical limit, such as a short introduction or a list of the three most important questions. For a practice prompt, try: “Here are three notes. Make a concise action list using only these notes. Keep unknown owners and dates marked as unknown.” You can check whether every item comes from the supplied material. The format gives the answer a clear structure without implying that the information is automatically correct. Review the content as well as whether it followed the requested shape.

Revise with a specific instruction

After reading the response, identify the change you need. “Remove the invented deadline and keep the original order” gives Friday more useful direction than “try again.” If the tone is wrong, describe the audience and desired level of formality. If a point is missing, supply the relevant information or ask which evidence would be needed. Keep the correction connected to the current objective. If a new requirement changes the task substantially, say so. You can ask Friday to summarize the revised goal before producing another draft. This reduces the chance that an earlier assumption continues unnoticed. Review the new result against both the original request and the correction; a second answer is not automatically an improvement merely because it is different.

Distinguish writing from doing

Friday may describe a plan, draft a message, or explain how a setting could be changed. Those are useful outputs, but they are not evidence that the message was sent or the setting was saved. If your task requires an actual operation, the application needs a permitted capability and a result you can verify. For a first conversation, make that boundary explicit. Ask for guidance or a draft and perform the intended interface action yourself when appropriate. If a reply claims a completed action, look for the corresponding artifact or saved state. You can ask, “What result can I inspect?” rather than treating confident wording as proof. This habit becomes more important as tasks involve files, tools, or external destinations.

Preserve a useful record

Give the conversation a name that describes its purpose where renaming is available. Return to it when continuing the same task, and review the last decision before proceeding. If the exchange becomes long, ask for a concise recap of established facts, open questions, and next steps, then check the recap against the conversation. Save reusable guidance separately when it belongs in Memory Notes and your role permits that action. Do not assume that saying “remember this” establishes a durable record without verification. Temporary details can remain in the current exchange. A useful conversation is one you can understand later: what was requested, what was supplied, what changed, and what still needs review. That is a stronger foundation than a collection of disconnected prompts with unclear outcomes.

Ask for a reviewable result

A useful follow-up is: “Review this answer against my stated goal and sources. Separate supported statements, assumptions, proposed actions, and unresolved questions.” Compare that review with the original documents. If a step would change files, publish content, or affect someone else, establish who may approve it before acting. End the session by requesting a concise handoff with accepted decisions and next steps, then save and reopen the accepted record using your installation’s supported controls. These habits appear throughout the Phronesis Intelligence publications and help make a conversation easier to resume.

Explore the publications and annotated reading guide.