Choose your journey

You do not have to become a programmer or knowledge engineer to benefit from bundled knowledge. Start with a task, and stop when you have a useful result. All routes begin with an existing example.

Find an answer

For consumers and anyone exploring an unfamiliar subject:

1. Choose an example about something you recognise.
2. Ask one question using the supplied evidence and prompt.
3. Open the cited record and identify its source and date.
4. Ask what the evidence cannot establish.

You are done when you have an understandable answer and know its main limit. A bundle could later organise your own reading, hobby research or local-history notes. You do not need to publish a repository or build an app.

Use knowledge at work

For knowledge workers preparing a briefing, comparison or handover:

1. Compare two or three records and retain their sources and observation dates.
2. Check a near-neighbour: a similar title, different jurisdiction or old version.
3. Try the small fictional starter before using work material.
4. Choose ten suitable source notes and five recurring questions for a private experiment, subject to your organisation's access and data-handling rules.

You are done when the collection supports a repeatable, sourced task. A configured SharePoint/Copilot derivative is an optional later route, not a prerequisite or a guarantee about every tenant.

Check your subject

For domain experts reviewing whether important distinctions survive:

1. Choose two easily confused concepts from the example catalogue.
2. Ask an AI to compare them using only identified records.
3. Inspect aliases, explicit relationship direction, authority, omissions and dates.
4. Correct a source note in a collection you may edit, rebuild and repeat the check.

You are done when you can explain the correction and show its effect on the answer. A candidate link is not identity; an inferred connection is not an official statement; missing data do not prove that something never happened. Do not modify the submitted government WebMCP candidate for this exercise.

Learn by making

For students and people who learn through experiments:

1. Predict the answer to the fictional starter's question before asking an AI.
2. Inspect the relevant Markdown record and its Explorer representation.
3. Change one recorded time, rebuild and check the new answer and citation.
4. Test a question whose answer is absent and record what the AI does.

You are done with a working small bundle and a checked experiment. Continue through the project studio if you want to research a larger subject. Building a personal interface is an optional final stage.

Keep a small receipt

Record the collection and version, material supplied, AI host if used, question, answer, cited record IDs, unsupported claims and what you changed. Do not record private prompts in public evidence. Use the studio's evaluation scorecard when comparing runs.

Next: make your first bundle.