Learning should keep the source in view.

Novus Learn starts with public Wikimedia sources and turns a pinned revision into stable sections, source blocks, direct claims, concepts, labelled relationships, coverage metrics, and attributable media.

It is not a chatbot and does not silently draw on model memory. You can ask a topic questions and even command it — “download the flashcards”, “open the quiz” — but every answer is assembled from the source's own sentences. Generated claims quote retrieved text, while derived and Wikidata relationships remain visibly labelled.

Public learning pages open directly. Saved topics, preferences, recent topics, files, edits, and future study progress stay on the device by default. No identity provider is part of this build.

• An accessible connected Knowledge Canvas.
• A validated visual engine with a source-grounded water-cycle benchmark.
• Guided lessons, flashcards, and safe quizzes generated from supported claims.
• Quick-read summaries and full study-notes breakdowns from the source.
• An extractive “Ask this topic” terminal with natural-language commands — no LLM, every answer a real source sentence.
• A free Wiktionary dictionary, with inline define-on-tap across learning sections.
• The Novus Hub — one ⌘K bar searching topics, your library, and the dictionary.
• Downloadable, fully-referenced sources for citing your work.
• Keyless link-outs to free videos, lectures, courses, certificates, open textbooks, and how-to guides.
• Accountless, self-contained share links with server-rendered preview pages.
• Local, browser-first uploads for text, Markdown, EPUB, PDF, Word (DOCX), PowerPoint (PPTX), and transcript files.

• More open-access journal sources beyond Europe PMC.
• Explicit, capability-based cloud sharing and cross-device projects.