How it works
Answer. See why. Get better.
TokIQ is built around one loop: make a prompting decision, find out if it was right, and understand why. Here is how it works, step by step.
- 🔥 6-day streakBeginner
Pick a topic
Short quizzes on how to write prompts that work.
FFundamentalsClear instructions, context and constraints72%RRole & contextSystem prompts, personas and framing58%EExamplesTeaching by example, and when it backfires40%RReasoningGetting a model to think before it answersSStructured outputJSON, schemas and answers you can parseGGroundingRetrieval, sources and staying factualQuizProgressProfile01Open the app and pick a topic
No sign-up needed for your first two questions. On the Quiz tab, choose one of eight topics, from Fundamentals to Safety. After you create an account you can tell TokIQ your level, which AI tools you use and what for.
- Structured outputIntermediate
Your app parses the model’s JSON, but sometimes the reply starts with “Sure! Here is the JSON:”. What is the most robust fix?
Check answerAAdd “please” to the promptBUse structured-output mode with a schema, then validateCLower the temperature to 0DStrip the first line of every reply02Answer one question at a time
Each question describes a real prompting situation: a prompt that fails, a choice between two instructions, an agent that loops. Pick the answer you think is best and tap Check answer.
- Structured outputIntermediate
Your app parses the model’s JSON, but sometimes the reply starts with “Sure! Here is the JSON:”. What is the most robust fix?
AAdd “please” to the prompt✓Use structured-output mode with a schema, then validateCLower the temperature to 0DStrip the first line of every replyThat’s right! +10 XP
Structured-output modes constrain the reply to valid JSON. Validation catches the rest.
Why?Next03See if you were right, and earn XP
You find out immediately, with a one-line reason. Correct answers earn XP, and answering on consecutive days builds your streak.
- Why?
Why string hacks break JSON parsing
A model that adds a friendly sentence before its JSON is doing what chat models are trained to do. Telling it “no preamble” helps, but only a schema-constrained mode makes the shape a guarantee.
{ "title": string,
"priority": "low" | "high" }Validate every reply. When it fails, retry once and include the validation error.
3 follow-up questions
What should a field be when the source is silent? ›When is JSON mode not enough? ›Where should the schema live? ›04Tap Why? to go one level deeper
Why? opens a short article on the idea behind the question, followed by three follow-up questions on the same point. This is where the learning sticks.
Progress
68%
Accuracy
6
Day streak
1,240
XP
By topic
Fundamentals82%Role & context64%Examples71%Reasoning45%Structured output58%Safety37%QuizProgressProfile05Track your progress and move up
The Progress tab shows accuracy per topic, your streak and XP, and an optional global ranking. When you keep getting questions right, the app suggests moving up a level.
Write better prompts, a few questions a day.
Short quizzes on real prompting decisions, with an explanation for every answer. Free to start on iPhone and Android.