Make a course with ChatGPT and study it in Lernoa

Make a course

Four steps, about five minutes. The prompt below is Lernoa's own: it teaches the assistant the exact file format the import reads.

1. Copy the prompt

   Type what you want in the box (a topic, a level, how many chapters) and press Copy prompt. Or copy it as it is and replace {DESCRIBE YOUR COURSE HERE} at the end in your chat.

   You write course files for the Lernoa app. Output ONE file and nothing else, following this format exactly (it is an all-tag grammar, NOT markdown — no "#" headings, no "- Key: Value" bullets, no "|" tables, no "Q1./A./Answer:" lines): The whole file is one <course> root element: <course format="lernoa-course/v2" subject="Folder name" level="beginner|intermediate|advanced|exam" language="English" tags="tag1, tag2" icon="graduation-cap"> <title>Course Title</title> (exactly one, required, plain text) <summary>Optional course summary paragraph.</summary> All <course> attributes are optional. "format" must be exactly "lernoa-course/v2" if present. Then one or more <chapter title="Chapter Title"> elements (1-50 total): <chapter title="Chapter Title"> <summary>Optional one-line chapter summary.</summary> <content> Chapter body as a sequence of these block tags, one blank line between blocks: <nutshell>Key takeaway.</nutshell> <heading>Section heading</heading> <paragraph>Body text.</paragraph> <points> <item>Bullet</item> <item depth="1">Sub-bullet under the bullet above (depth 0-3)</item> <item>Another bullet</item> </points> <table> <row header="true"><cell>Column A</cell><cell>Column B</cell></row> <row><cell>Value</cell><cell>Value</cell></row> </table> <callout style="key">Fact worth remembering (styles: key, tip, warning).</callout> <analogy>A comparison that makes the idea concrete.</analogy> <example>A worked example.</example> <code>verbatim code; escape &lt; &gt; &amp; as &amp;lt; &amp;gt; &amp;amp;</code> <code lang="d2">direction: right request: Request cache: "Cached?" {shape: diamond} serve: Serve from cache fetch: Fetch and store request -> cache cache -> serve: Yes cache -> fetch: No</code> </content> <quiz> Optional. 0-25 <question> elements — see the question format below. </quiz> </chapter> Each text block is one paragraph: line breaks inside it become spaces, so start a new block (or use <points>) for a new line. Only <code> keeps line breaks. Inline formatting inside any text (paragraph/heading/nutshell/analogy/example/callout/item/cell text, titles, summaries, prompts, choices, explanations): <b>bold</b>, <hl>highlight</hl>, `inline code` (backticks), <color c="red|green|blue|amber|purple">colored text</color>, <icon name="rocket" /> (self-closing, lowercase letters/digits/hyphens only). Use only these Lernoa tags — no arbitrary HTML, no markdown, no links, no images inside titles/prompts/etc. Escape literal <, >, & as &lt; &gt; &amp; anywhere outside backticks. <table> is for comparisons and reference grids: <row> elements holding <cell> elements, header="true" on the heading row, at most 12 columns and 60 rows, inline formatting only inside a cell (no lists, code blocks or nested tables). Keep every row the same number of cells. <code lang="d2"> holds D2 diagram SOURCE (d2lang.com) that Lernoa draws as a diagram when the chapter is read (the editor shows the source). Use it where a decision, flow, sequence of interactions, lifecycle or structure is clearer drawn than described. Declare each shape on its own line as id: Label and connect ids with -> (or <-, <->, --) plus an optional ": label"; pick a shape with {shape: diamond} for a decision, cylinder for a data store, or person, queue, cloud, oval, circle, hexagon, page, document, package, step; nest shapes in a container (vpc: VPC {web: Web server; db: Database}) and reach them as vpc.web; start a flow with direction: right (or down); for who-talks-to-whom over time put shape: sequence_diagram on the first line; for a data model give each table {shape: sql_table}. Keep it small (about 12 shapes or fewer, short labels). Put a label in double quotes when it contains # ; { } | or a quote (# starts a comment), never use a D2 keyword (shape, label, style, direction, icon, link, near, tooltip, width, height, class, vars) as an id, and add no style, classes, vars, icons or links — the app themes it. Same escaping as any <code> block; no markdown fences. <code lang="mermaid"> (Mermaid source) still works, but write new diagrams in D2. <image src="/cards/images/..." alt="..." /> is allowed inside <content> but ONLY with a Lernoa-managed /cards/images/ URL — never write an <image> tag pointing anywhere else; omit images entirely if you have none to use. Question format (used inside both <quiz> and the optional trailing <exam>): <question kind="mcq"> <prompt>Single-answer question?</prompt> <choice>Wrong choice</choice> <choice correct="true">Right choice</choice> <choice>Wrong choice</choice> <choice>Wrong choice</choice> <explanation>One sentence explaining the answer.</explanation> </question> <question kind="multi"> <prompt>Multi-answer question? Select all that apply.</prompt> <choice correct="true">Right choice</choice> <choice correct="true">Right choice</choice> <choice>Wrong choice</choice> <explanation>One sentence explaining the answer.</explanation> </question> <question kind="ordering"> <prompt>Put these steps in order, first to last.</prompt> <item>First step</item> <item>Second step</item> <item>Third step</item> <explanation>One sentence explaining the order.</explanation> </question> <question kind="matching"> <prompt>Match each item to its answer.</prompt> <pair><item>First item</item><match>Its answer</match></pair> <pair><item>Second item</item><match>Its answer</match></pair> <pair><item>Third item</item><match>Its answer</match></pair> <choice>An extra option that matches no item</choice> <explanation>One sentence explaining the matches.</explanation> </question> Rules: "kind" is "mcq" (exactly one correct choice), "multi" (two or more correct choices), "ordering" or "matching" — you may omit "kind" and it will be inferred (from <pair>s, from <item>s, else from how many choices are marked correct="true"). An mcq/multi question needs 2-10 <choice> elements, in the order they should be shown, with correct="true" on every correct one (at least one). An ordering question lists 2-10 different <item> elements in the CORRECT order and no <choice> (learners see them shuffled). A matching question has 2-10 <pair> elements, each with one <item> and its one <match>; the same <match> may answer several items, optional extra <choice> elements (no correct attribute) are options that match nothing, and there are at most 10 different options. <prompt> and <explanation> appear at most once each; <explanation> is optional. Every question is graded all-or-nothing. After all <chapter> elements, optionally add ONE trailing <exam> (sibling of the chapters, not nested in one) holding up to 500 <question> elements in the same format, then close </course>. Limits: 1-50 chapters, up to 25 quiz questions per chapter, up to 500 exam questions, 2-10 choices (or ordering items, or matching pairs) per question. Now write the complete course file for: {DESCRIBE YOUR COURSE HERE}

   Select the text and copy it, or press Copy prompt.

2. Paste it into ChatGPT

   Open ChatGPT, Claude or Gemini and paste the prompt. The reply is one course file: chapters, a quiz for each, and a final exam.

3. Save the reply as a file

   Paste the reply into a plain text editor and save it as course.md (or course.txt). The file should start with <course. If the assistant puts it in a code block, copy what is inside the block.

4. Import it into Lernoa

   In Lernoa open Learn, then Build a course, then Import from files, and drop the file in. The course appears with its chapters, quizzes and final exam, ready to study.