Prompts for Brainstorming and Creative Writing

How Creative Tasks Differ Most tasks in this tutorial have a correct answer: a right label, an exact amount, a faithful summary. Brainstorming and creative writing are different:

Analytical Tasks Creative Tasks One correct answer Many good answers Consistency is a goal Variety is a goal Examples help by being copied closely Examples can limit originality if copied closely Success is measured by correctness Success is measured by usefulness, originality, and fit Language models tend to produce the most likely response (Lesson 2). For a fact, that is helpful. For ideas and stories, the most likely response is often the most predictable one. Creative prompting is largely about steering the model away from the obvious and toward what is useful, without losing control of the constraints that matter.

This lesson is a practical collection of brainstorming and creative writing prompts.

Recipe Produces

1. Divergent brainstorming Many varied ideas
2. Constraint-based ideas Ideas that fit real limits
3. Perspective shifting Ideas from different viewpoints
4. Idea evaluation Ideas scored against criteria
5. Idea development One idea expanded into a plan
6. Naming Names and taglines
7. Short creative writing Stories, scenarios, and dialogues
8. Educational content Exercises, quiz scenarios, examples
9. Variations Several distinct versions of the same piece Recipe 1: Divergent Brainstorming Use it when: you need many ideas quickly and want to see the range of possibilities.

Divergent thinking means generating many different options; convergent thinking means narrowing them down. Brainstorming prompts should be divergent first.

Prompt Generate 15 ideas for {goal}.

Requirements:

• Each idea must be different in approach, not a variation of another idea.
• Avoid these common ideas: {obvious_ideas}
• Include at least 3 unusual ideas that most people would not think of first.
• One line per idea: a short name, then a one-sentence description. Example Settings goal: increasing the number of students who finish a 6-week online coding course obvious_ideas: email reminders, certificates, discounts Why It Works A large number pushes past the first few predictable ideas. "Different in approach" prevents ten variations of one idea. Excluding obvious ideas forces the model away from its most likely answers (Lesson 17). Asking explicitly for unusual ideas makes room for options that are less likely but potentially more valuable. Removing Near-Duplicates in Code Even with instructions, similar ideas appear. A simple similarity check removes them before review.

import re from difflib import SequenceMatcher

def normalize(text: str) -> str: return " ".join(re.findall(r"[a-z0-9]+", text.lower()))

def dedupe_ideas(ideas: list[str], threshold: float = 0.8) -> list[str]: """Keep each idea only if it is not too similar to an idea already kept.""" kept = [] for idea in ideas: if all(SequenceMatcher(None, normalize(idea), normalize(k)).ratio() < threshold for k in kept): kept.append(idea) return kept

ideas = [ "Weekly coding challenge with leaderboard", "Weekly coding challenges with a leaderboard!", "Peer code review circles", "Alumni mock interview nights", "peer code-review circles", "Monthly hackathon with local startups", ] print(dedupe_ideas(ideas)) Output:

['Weekly coding challenge with leaderboard', 'Peer code review circles', 'Alumni mock interview nights', 'Monthly hackathon with local startups'] Code Explanation normalize() lowercases text and keeps only letters and digits, so punctuation and capitalization do not affect the comparison. SequenceMatcher(...).ratio() measures similarity from 0 to 1 (Lesson 27). An idea is kept only if it is less similar than the threshold to every idea already kept. The near-duplicates "Weekly coding challenges with a leaderboard!" and "peer code-review circles" are removed. Text similarity only catches ideas with similar wording. Two ideas with the same concept in different words still need a human eye or a model-based grouping step.

Recipe 2: Constraint-Based Ideas Use it when: ideas must be realistic for your situation: budget, team size, time, or technology.

Prompt Suggest 8 ideas for {goal}.

Every idea must fit these constraints:

• Budget: {budget}
• Team: {team}
• Time to launch: {time}
• Must not require: {exclusions}

For each idea, give: name, one-sentence description, and which constraint it is closest to breaking. Why It Works Creative ideas that cannot be implemented are not useful. Stating constraints before generation produces ideas that fit. Asking which constraint each idea is closest to breaking highlights risks early.

Recipe 3: Perspective Shifting Use it when: your ideas feel similar to each other, or you need to see a problem from viewpoints you have not considered.

Prompt We want to {goal}.

Generate 3 ideas from each of these perspectives:

1. A first-year student who is struggling with motivation
2. A working professional studying at night
3. A mentor who reviews student projects
4. A small company that hires our graduates

For each idea, explain in one sentence why that perspective leads to it. Why It Works Each perspective changes which problems and solutions seem important (Lesson 34, roles). Combining several perspectives produces more varied ideas than a single request, and the explanations show why each idea matters to someone.

Recipe 4: Idea Evaluation Use it when: you have many ideas and need to choose which to pursue. This is the convergent step after divergent brainstorming.

Prompt Evaluate each idea in against these criteria, scoring 1 to 5:

• impact: how much it would help {goal}
• effort: how easy it is to implement (5 = very easy)
• risk: how safe it is (5 = very low risk)

Return only a JSON array: [{"idea": "...", "impact": n, "effort": n, "risk": n, "main_concern": "one sentence"}]

{ideas} Why It Works Evaluation in a separate step from generation produces more critical judgments (Lesson 10, Tip 6). Defined criteria and scales make scores comparable. Weighted totals and ranking belong in code (Lesson 48), and close scores should be treated as ties. Recipe 5: Idea Development Use it when: an idea has been chosen and needs to become concrete.

Prompt Develop the idea in into a plan for {context}.

Include:

1. What participants actually experience, step by step
2. What we need to prepare
3. How we will know it is working (2 or 3 measurable signals)
4. The biggest risk and how to reduce it
5. The smallest version we could test in 2 weeks

{idea} Why It Works The structure turns a one-line idea into something testable. Asking for the smallest version to test encourages experimentation rather than large, risky launches.

Recipe 6: Naming Use it when: naming a course, feature, event, or product.

Prompt Suggest 12 names for {thing}.

The name should:

• Suit {audience}
• Suggest {qualities}
• Be easy to pronounce for speakers of English and {other_languages}
• Be at most {max_words} words

Give names in 4 different styles: descriptive, playful, single invented word, and metaphor-based. For each name, add a one-line rationale. Why It Works Different styles produce a range instead of twelve similar names. Pronunciation for your actual audience matters in multilingual markets. Rationales help a team discuss and choose. Important Checks Generated names may already be used by other businesses or be registered trademarks. Check availability, domain names, and trademark registries yourself before adopting a name. The model cannot reliably know whether a name is taken.

Recipe 7: Short Creative Writing Use it when: writing short stories, scenarios, or dialogues, for example to make a lesson more engaging or to illustrate a concept.

Prompt Write a short story of about {length} words for {audience}.

Story elements:

• Main character: {character}
• Setting: {setting}
• Situation or conflict: {conflict}
• The story should naturally show this idea: {concept}

Style: {style} Avoid: clichés such as {cliches}, and a moral stated directly at the end. The story must be original; do not imitate or retell an existing story. Example Settings character: Divya, a second-year student learning Git setting: the night before a group project deadline conflict: two teammates changed the same file concept: why branches and merging exist style: light, warm, with realistic dialogue cliches: "suddenly everything made sense", a dream ending Why It Works Specific story elements prevent a generic plot. A concept to show makes the story serve a teaching purpose, which is useful for tutorials and courses. Naming clichés to avoid pushes the model away from predictable phrasing. "Must be original" keeps the output from reproducing or closely imitating existing copyrighted works. Recipe 8: Educational Content Use it when: creating practice material: exercises, quiz scenarios, discussion questions, or project ideas.

Prompt Create 5 practice exercises about {topic} for {level} learners.

Each exercise should:

• Be set in a realistic situation from {domain}
• Require applying the concept, not just recalling a definition
• Increase in difficulty from exercise 1 to exercise 5
• Include the expected answer and one common mistake learners make

Return the exercises as a numbered list. Why It Works Realistic situations make practice meaningful. "Applying, not recalling" produces exercises that test understanding. Increasing difficulty supports progressive learning. Common mistakes prepare explanations for feedback. Verify Educational Content Exercises and answers must be correct. Check every answer yourself, and run any code answers. For multiple-choice questions, use a separate review step to check that exactly one option is correct (Lesson 10, Tip 6).

Recipe 9: Variations Use it when: you need several versions of the same content, such as headlines for A/B testing or different openings for an article.

Prompt Write 5 variations of {content_type} for the content in .

Each variation must use a different approach:

1. A question
2. A surprising fact from the content
3. A direct benefit statement
4. A short story or scenario
5. A bold, simple claim supported by the content

Use only facts from the content. Maximum {max_chars} characters each.

{text} Why It Works Named approaches guarantee genuinely different variations instead of rephrasings. "Only facts from the content" keeps creative headlines honest. A length limit suits real placements; check it in code (Lesson 8). Variations created this way are well suited to testing with real users (Lesson 69).

Controlling Creativity Lever More Variety More Control Instructions "Different approaches", "avoid the obvious", "include unusual ideas" Precise constraints and required elements Examples None, or several very different ones One close example of the desired style Number requested Ask for many, then filter Ask for a few Perspectives Several roles or viewpoints A single role Sampling settings Higher temperature, if supported (Lesson 52) Lower temperature, if supported Process Generate first, evaluate later Generate and evaluate in one step A common, effective pattern is diverge, then converge: generate broadly with few restrictions, remove duplicates, evaluate against criteria, and develop the best ideas (Lesson 48).

Common Mistakes Mistake 1: Asking for a Few Ideas What it is: "Give me 3 ideas."

Why it happens: Three seems manageable.

What it causes: The three most predictable ideas.

How to fix it: Ask for many, require different approaches, then filter and evaluate.

Mistake 2: Generating and Judging Together What it is: "Give me the best idea."

Why it happens: It seems efficient.

What it causes: The model chooses before exploring, usually picking a safe, common idea.

How to fix it: Separate generation (Recipe 1) from evaluation (Recipe 4).

Mistake 3: No Constraints for Practical Ideas What it is: Brainstorming without budget, team, or time limits.

Why it happens: Constraints feel like they limit creativity.

What it causes: Exciting ideas that cannot be implemented.

How to fix it: Include real constraints when ideas must be actionable.

Mistake 4: Trusting Facts Inside Creative Output What it is: Publishing creative content that contains statistics or claims.

Why it happens: Creative output feels harmless.

What it causes: Invented facts in headlines, stories, or educational examples.

How to fix it: Restrict facts to provided content, and verify any factual claim.

Mistake 5: Skipping Originality and Name Checks What it is: Using generated names, slogans, or stories without checking them.

Why it happens: They appear new.

What it causes: Accidental similarity to existing names, trademarks, or works.

How to fix it: Ask for original work and check names and close similarities yourself.

Best Practices Diverge first, then converge. Many varied ideas, then structured evaluation.

Exclude the obvious explicitly. Name the ideas or clichés you do not want.

Use perspectives and named approaches. They produce genuine variety.

Add real constraints when ideas must be practical. Budget, team, time, and exclusions.

Remove duplicates and evaluate in separate steps. Use code for deduplication and scoring.

Keep facts and originality under control. Facts from provided content; original work; verified names.

Try It Yourself Exercise 1: Diverge and Deduplicate Use Recipe 1 to generate 20 ideas for improving attendance in live online classes. Run dedupe_ideas() on the result, then manually group the remaining ideas by concept. Count how many genuinely different concepts you received.

Exercise 2: Evaluate and Develop Take the deduplicated ideas from Exercise 1 and evaluate them with Recipe 4. Compute a weighted score in code (Lesson 48), choose the top idea, and develop it with Recipe 5.

Exercise 3: A Teaching Story Use Recipe 7 to write a 300-word story that illustrates a concept you teach, such as variables, APIs, or database indexes. Ask a learner, or imagine one, to explain the concept after reading the story. Revise the story elements until the concept comes through clearly.