Question format
A question has 4 answers: one correct, 3 distractors. This is what the QUESTIONS tab’s ADD QUESTION form and the bulk-upload template both collect, and what the server checks on every save.
Fields
Section titled “Fields”| Field | Type | Required | Notes |
|---|---|---|---|
| Domain | one of your dimension’s domains | yes | A platform domain, or one of your own (Add domains and subtopics). |
| Tier | 1, 2 or 3 |
yes | See Questions and tiers for what each tier tests. |
| Subtopic | text, 1–100 characters | yes | Must belong to the question’s domain. |
| Question (prompt) | text, 10–2000 characters | yes | The stem players see. |
| Correct answer | text, 1–500 characters | yes | |
| Distractors | 3 entries, each 1–500 characters | yes | Exactly 3 — no more, no fewer. |
| Question type | one of 19 values | yes | The cognitive shape of the question — see below. |
| AI relevance | one of A–D |
yes | How exposed the question is to AI — see below. |
| Bloom’s level | 1–7 | yes | Cognitive demand. Roughly: Tier 1 → 1–3, Tier 2 → 3–4, Tier 3 → 4–7. |
| Explanation (correct) | text, 1–2000 characters | yes | Why the correct answer is right. Must share at least two substantive words with the correct answer — it has to explain this answer, not the topic in general. |
| Explanation (distractors) | one entry per distractor, each 1–2000 characters | yes | Why each distractor is wrong. |
| Tags | up to 10 entries, each max 50 characters | no |
Question type
Section titled “Question type”One of: definitional, scenario, debugging, tradeoff, architectural, analytical, comparison, production_incident, ai_limitation, human_judgment, metacognitive, ai_code_review, constraint_architecture, prompt_engineering, cross_domain, last_mile, ai_output_validation, agent_orchestration, context_engineering.
Pick the one that best matches what the question asks the player to do — definitional recall, scenario reasoning, debugging, tradeoff analysis, architectural design, a production-incident response, or judging AI output, among others.
AI relevance
Section titled “AI relevance”| Value | Meaning |
|---|---|
A |
AI-Obsolete — pure recall, an LLM answers in seconds. |
B |
AI-Commoditized — AI handles it well; humans need context to apply it. |
C |
AI-Augmented — AI helps, but human judgment matters. |
D |
AI-Resistant — AI demonstrably fails; human expertise required. |
Length budgets by tier
Section titled “Length budgets by tier”A question must fit its tier’s length budget: the prompt (excluding fenced code) and every answer, measured in characters.
| Tier | Prompt | Longest answer |
|---|---|---|
| 1 | 200 | 70 |
| 2 | 266 | 90 |
| 3 | 333 | 100 |
AI-generated questions are held to a tighter budget than hand-authored ones.
The correct answer’s length also can’t drift more than 15/20/25 characters (Tier 1/2/3) from the distractors’ average length — a correct answer that’s conspicuously longer or shorter than the others is itself a tell. See Write questions that pass validation for the full list of content rules this feeds into.
Who can author
Section titled “Who can author”The dimension’s owner and admins. For a dimension owned by an organisation, that organisation’s owner and admins too.