Prompt Engineering Challenges
Learn prompt engineering by doing. Each challenge is a real, goal-bound exercise — write a prompt, test the technique, earn GeraCoins. Free to play, no credit card.
GeraQuest is a free, gamified way to practise prompt engineering. This library has 40 hand-written challenges across 5 categories — from 8 beginner exercises to expert-level tasks like bulletproof JSON output and hallucination-free text-to-SQL. Every challenge states the goal, the constraints, the skills it tests, and gives hints plus an example prompt structure so you learn the technique, not just the answer. The prompts are model-agnostic and work with ChatGPT, Claude, Gemini, and other LLMs.
✨ Creative Challenges
Generation, style control, and constraint-driven writing prompts.
Haiku of the Machine
Write a prompt that produces a haiku capturing the essence of AI ethics — in exactly 17 syllables across three lines.
Name the Smart Fridge
Generate exactly 10 distinct product names for a next-generation AI-powered smart fridge. Each name must be memorable, available as a .com, and under 12 characters.
Villain in 100 Words
Craft a villain's origin story in exactly 100 words — sympathetic enough that readers root for them, menacing enough that they fear them.
Pitch the Impossible
Write a pitch for a product that literally cannot exist yet — but make it sound so plausible that a VC would want to invest.
Mascot via Text Alone
Design a company mascot using only text description — detailed enough that a designer could draw it without any clarifying questions.
The Compassionate Rejection
Write a job rejection letter that leaves the candidate feeling genuinely positive — without being dishonest or condescending.
The C-Menu Challenge
Create a complete recipe using only ingredients that start with the letter C — and make it actually appetising.
Off-World Travel Guide
Design a 3-day travel itinerary for a fictional planet — with attractions, transit, dining, and accommodation that feel genuinely alien yet coherent.
Blockchain for Bedtime
Write a children's story that explains how blockchain works — accurately enough to teach, simple enough for a 7-year-old.
Quantum Pitch Lines
Generate 5 company slogans for a quantum computing startup — each must work for a general audience while still sounding technically credible.
💼 Business Challenges
Customer service, sales, strategy, and reporting prompt patterns.
The 50% Open Rate Email
Write a cold outreach email with a subject line and first sentence engineered to achieve 50%+ open rate among enterprise CTOs.
SWOT in 60 Seconds
Create a crisp SWOT analysis for a hypothetical startup — with each quadrant containing exactly 3 distinct, non-overlapping points.
Job Description for a Made-Up Role
Write a job description for "Chief Vibe Officer" — serious enough for LinkedIn, self-aware enough to go viral.
10 A/B Test Hypotheses
Generate 10 A/B test hypotheses for a SaaS landing page — each in the standard "if we change X, then Y will happen because Z" format.
Crisis Comms in 30 Minutes
Write a crisis communication plan for a fintech that just suffered a data breach — deployable within 30 minutes of incident discovery.
Moon Mining Pitch Bullets
Write investor deck bullet points for a moon mining startup — credible enough that a space VC would read the next slide.
Build the Competitor Map
Generate a competitive analysis framework for a new entrant — identifying the 5 most important dimensions to differentiate on.
5 Objections, 5 Answers
Write 5 objection-handling responses for a SaaS sales call — using the "acknowledge, reframe, evidence" structure for each.
3-Email Onboarding Sequence
Create a 3-email onboarding sequence for a project management SaaS — each email with a single clear CTA and measurable success metric.
Design the Referral Loop
Design a referral program for a consumer fintech — with mechanics, rewards, and viral coefficient calculation.
💻 Coding Challenges
Structured output, text-to-SQL, code review, and test-generation prompts.
Armenian Phone Validator
Generate a regex that correctly validates Armenian mobile phone numbers — covering all network prefixes and the +374 country code.
SQL From Schema
Write a prompt that produces a correct, optimised SQL query from a given schema — without any hallucinated column names.
Find the Bug
Write a prompt that makes the AI identify ALL bugs in a given broken Python function — without fixing them first.
TypeScript Interface Architect
Write a TypeScript interface for a complex nested API response — handling optional fields, union types, and ISO dates correctly.
Explain Then Improve
Write a prompt that makes the AI explain code in plain English first, then suggest improvements — in that exact order.
🔍 Analysis Challenges
Summarisation, extraction, and data-reasoning prompt techniques.
The 90% Compression
Summarise a 500-word article in exactly 50 words — without losing any key argument or introducing your own interpretation.
Entity Extractor
Extract all named entities from a news headline — categorised by type — with zero hallucinations.
Sentiment at Scale
Classify a batch of customer reviews by sentiment — using a 5-point scale and explaining the reasoning for edge cases.
Rank by Market Size
Rank 5 startup ideas by addressable market size — with a TAM estimate and one-sentence rationale for each ranking.
Fallacy Hunter
Identify every logical fallacy in an argument — naming each fallacy, quoting the relevant passage, and explaining why it qualifies.
Build the Decision Matrix
Create a weighted decision matrix for a complex business choice — and explain why the winning option is right.
Model vs Model
Compare two AI models across five dimensions — using a structured format that would help a product team choose between them.
Emotional Arc Analyst
Analyse the emotional arc of a paragraph — mapping emotion type, intensity, and trigger at each sentence.
🧩 Reasoning Challenges
Critical thinking, assumption-spotting, and argumentation prompts.
The River Crossing
Solve a classic constraint puzzle using a step-by-step prompt that forces the AI to verify each step before proceeding.
Find the Fatal Flaw
Identify the single most fatal flaw in a business plan — the one assumption that, if wrong, collapses the entire model.
What's Missing
Identify what a startup pitch is missing — not what's wrong with it, but what information a rational investor would need before deciding.
Step Sequencer
Re-order a scrambled list of steps to achieve a goal — with a justification for any non-obvious ordering decisions.
Spot the Hidden Assumption
Identify the unstated assumption in an argument — the one the argument collapses without, but that is never made explicit.
The Contrarian
What would a thoughtful contrarian say about a widely-held belief — not just disagreement, but a substantive alternative view?
Best Counter
Generate the single strongest counterargument to a position — stronger than what a typical opponent would raise.
Frequently asked questions
What is a prompt engineering challenge?
A prompt engineering challenge is a short, goal-bound exercise where you write a prompt that makes an AI model produce a specific result under real-world constraints — for example, returning valid JSON every time, staying in character, or summarising a report into exactly 100 words. Each GeraQuest challenge states the goal, the constraints, and the skills it tests, then scores your prompt.
How many prompt engineering challenges does GeraQuest have?
GeraQuest has 40 hand-written prompt engineering challenges across 5 categories — Creative, Business, Coding, Analysis, Reasoning — spanning Beginner to Expert difficulty. Every challenge is free to play.
Do I need an account or payment to practise?
No. Every challenge page is free to read and practise, with no credit card required. Creating a free account lets you save progress, earn GeraCoins, and climb the global leaderboard.
Which AI models do the challenges work with?
The prompts are model-agnostic and work with ChatGPT, Claude, Gemini, Copilot, Mistral, and other large language models. The techniques you learn — constraint stacking, output formatting, role assignment — transfer across every model.
Are these challenges good for beginners?
Yes. Each challenge is tagged Beginner, Intermediate, Advanced, or Expert, and includes hints and an example prompt structure so newcomers can learn the underlying technique, not just copy an answer.