The knowledge behind the world
AI helps draft field overviews, framework articles, timelines, curricula, and concept maps in Noosaga’s Atlas. That structure supplies Nousfarer’s academic planets and the lessons inside their institutes.
AI Transparency
Nousfarer is a world for exploration and learning. AI helps create the knowledge you encounter there, and that content can be wrong. This page explains where AI is involved, what review means, and how your requests are used.
Last updated: 2026-09-081. AI in Nousfarer
Nousfarer: The Idea Expanse is Noosaga’s main experience. Noosaga’s Atlas supplies its fields, frameworks, curricula, articles, quizzes, and learning progress. AI helps generate and organize parts of that knowledge foundation and helps Copilot turn a learning goal into an expedition.
Learning content is saved and reused. Awakening an institute or preparing a missing scroll can start generation, while returning to an existing article uses the stored material. A familiar article is not necessarily newly generated, recently checked, or written with the current model and prompts.
2. What AI Creates
Concept scrolls and quiz trials
AI generates concept explanations and multiple-choice questions from the available learning context. In the ordinary workflow, the article is written first and the quiz is generated using that article. A quiz can therefore repeat an error from its source article.
Copilot expeditions
Copilot uses your learning goal, preparation, selected options, and relevant Atlas context to help plan a route. Custom plans are generated automatically and are not read by a person before you see them. Prepared starter expeditions reuse existing routes.
Content maintenance
Selected review and repair workflows use a separate AI reviewer to flag possible defects or propose changes. Atlas Review supports discussion and human approval of maintenance proposals. These processes can miss errors and do not certify the content.
Content generation uses a centrally configured generation model. A separately configured reviewer may critique content in workflows that call for it; independent AI review is not a step every article or quiz receives. Requests are processed through OpenRouter and the configured model providers.
3. Checks and Human Review
Automated checks can reject mechanically invalid content, such as a broken reference or an invalid quiz answer index. They cannot establish that a statement is true, an explanation is coherent, or a question teaches the right distinction.
AI reviewers can flag suspected errors, omissions, and weak explanations. An accepted result, a high score, or a review with no reported issues is an automated status—not evidence of educational quality. A framework’s verification status is a limited identity or labeling check, not peer review.
Educational review requires a person to read the exact article and every quiz question, answer, and explanation. Do not assume all content you encounter has received that review. Reviewing one sample or approving one correction does not approve an entire institute. When content changes, the changed version needs to be reviewed again.
Passing a Sphinx trial or earning an institute reward records progress through the game. It does not certify expertise or validate the assessment itself.
4. Content Limits
Generated material can contain factual mistakes, misleading examples, unclear questions, invented references, or inaccurate dates. It can overrepresent familiar traditions, omit important perspectives, or present a disputed interpretation as settled. Coverage also varies between fields.
Using Atlas context constrains a request, but does not make that context true. An error in a framework article can pass into its curriculum, a concept scroll, and then a quiz. Consistency across those materials is not independent confirmation.
Copilot routes depend on available coverage and what you tell it about your preparation. Read the plan before setting off and question a stop that does not fit your goal. References and reading suggestions are pointers: check that a source exists and supports the claim before relying on it.
5. Data and Providers
AI-assisted requests may send your question or learning goal, answers about your preparation, selected options, and relevant Atlas context to OpenRouter and configured model providers. Copilot planning can also use relevant learning progress, your current Atlas location, and a selected saved expedition as context.
Generated learning material is stored in the Atlas. Saving an expedition and recording quiz completion also create application records. Do not assume an AI-assisted interaction is ephemeral or processed only on your device. Provider processing, network transit, and operational logging can occur.
Avoid including sensitive personal information, confidential third-party data, or material you are not authorized to share in a learning request. The Privacy Policy describes processing, retention, and providers. You can export or request deletion of account-linked data through Account & Privacy on your profile.
6. Using and Sharing Content
Use Nousfarer to explore a subject, practice explaining concepts, and identify what to study next. Check important claims against primary sources and expert references. For teaching, read the actual lesson and every quiz question before using them with learners.
When sharing generated material, describe its AI-assisted origin and the review you actually performed. Do not present it as independently verified, human-authored, or professionally certified without a basis for that description. A manual correction does not change how the original draft was produced.
Noosaga’s educational outputs should not be used to make high-impact decisions about people, including decisions about health, employment, credit, admissions, housing, or access to essential services. See Trust & Provenance for further guidance on assessing learning content.
7. Corrections and Contact
Report a content issue by emailing axel@noosaga.com. Include the field, framework or concept, the exact passage or question, and a supporting source if you have one.
How corrections are made explains how reported problems are reviewed. You can also contact us about AI use, review practices, or data processing.