The Art of the Stupid Idea: How AI could unlock your most creative work

We've been taught that group brainstorming is a creative free-for-all, but is it? The intent is good, but the human element is tricky. Foundational studies on group creativity show a clear productivity loss. People hold back their ideas due to a fear of judgment. We block or forget our own ideas while waiting for the most dominant voice in the room to finish.

So, we try brainstorming solo. But that's not a silver bullet, either. Often, the fear of judgment disappears, but it's replaced by a different set of cognitive traps:

1. The Blank Page Problem: The sheer terror of starting from scratch.
2. The "It's Just a Box" Problem: Our brains get tunnel vision, seeing things only for what they usually do. A box? It holds tacks. Not a candle-holder, not a stepping stool, just... a box. We miss the creative possibilities right in front of us.
3. Anchoring Bias: We fall in love with our first good enough idea and stop searching for a great one.

It's not that our brains are built to fail. It's that our brains are highly optimized for survival and pattern-matching, not for sustained, divergent, and fearless creativity.

I tried to find a cheat code. And I found it.

My new brainstorming partner isn't a person, it's AI and not in the way you think.
Using the foundations of design thinking, I use LLMs to create a psychologically safe sandbox where my stupid ideas can grow and systematically overcome both the group and solo brainstorming traps.

An important thing to note is that timing matters. Recent studies found that the timing of AI involvement during ideation profoundly shapes creativity outcomes. When LLMs are introduced too early, before humans have formed any context or emotional anchor, they can narrow originality and make people feel less ownership of their ideas.

AI is most powerful when it joins the conversation after you've grounded the problem in human understanding.

---

The Human Trap vs. The AI Trap#

My first Aha! moment was realizing that humans and AI are the perfect dysfunctional-but-brilliant team because they have opposite problems.

The Human Trap: We are biased, fearful of judgment, and get stuck in ruts (we can only see the obvious uses for things). We are great at taste and making connections, but often bad at fearless generation.

The AI Trap: AI has no fear and no ruts, but a 2025 Wharton/Nature study revealed its own flaw: The Homogeneity Trap. In one test, 94% of AI-assisted users suggested the exact same castle toy concept from a brick and a fan. The unassisted humans had wildly unique ideas.

AI is prolific, but it's also incredibly average. It defaults to the most statistically probable answer, which is often the most boring one.

This is the solution:

• Use the AI's speed and non-judgment to shatter your human fear and functional fixedness.
• Use your human taste and curation to overcome the AI's homogeneity trap.

My job is no longer Chief Idea Generator. I am the Chief Idea Curator.

In design thinking, the first step is always empathy, seeing the world through someone else's eyes. The 'Human Trap' is all about our biases and blind spots, and the 'AI Trap' is about missing the messy, beautiful uniqueness of real people. The magic happens when you use design thinking to define the real problem, then let AI help you break out of your rut and see new possibilities.

A Detailed 3-Phase Brainstorming Workflow#

This is how I move from a blank page to a pressure-tested idea, using J.P. Guilford's classic model of Divergent (go wide) and Convergent (narrow down) thinking, remixed with the best bits of design thinking: empathy, wild ideation, and rapid testing.

Phase 1: Divergence (Going Wide), aka Ideate & Empathize#

The goal here is pure, unadulterated quantity. This is the classic design thinking Ideate stage, but with a twist: you use empathy to see the problem from every angle, and then let the ideas fly.

⚠️ Start with your rough idea first: Remember, timing matters. Don't ask AI for ideas from scratch. That triggers the Homogeneity Trap and kills ownership. Instead, start with your own seed idea, problem observation, or domain context. Even if it's rough, incomplete, or "stupid", that human anchor is what makes the divergence phase powerful. You're not asking AI to brainstorm for you; you're asking it to explode your idea into 10 different directions.

• Prompting Technique 1: Prime the Persona.

  • Why: This is your #1 weapon against the Homogeneity Trap. By forcing the AI into a specific, opinionated role, you break its average default.
  • How:
  plaintextCopy

  # ❌ BAD PROMPT
  # Can you give me ideas for a new app?
  
  # ✅ GOOD PROMPT
  I'm thinking about productivity tools for construction workers, specifically around 
  tracking materials on chaotic job sites.
  
  You are a venture capitalist from Andreessen Horowitz who specializes in B2B SaaS. 
  You are skeptical of consumer apps and obsessed with network effects and high-margin 
  business models. 
  
  Take my rough idea and explode it into 15 wildly different directions. Think about 
  different business models, different end users (foremen vs workers vs suppliers), 
  different form factors (mobile, wearable, IoT), and adjacencies I haven't considered. 
  Go absolutely wild with quantity, I want range, not polish.
  

  • Design Thinking Note: This is empathy in action, seeing the world through a persona's eyes, not just your own. Notice how you lead with YOUR specific context, then ask the persona to explode it into many directions. 15+ ideas force true divergence.

• Prompting Technique 2: The Stupid Idea Seed.

  • Why: This is your antidote to fear of judgment. The AI cannot laugh at you. This is your safe space to throw your absolute worst, half-baked ideas, just to see what happens.
  • How:
  plaintextCopy

  # ❌ BAD PROMPT
  # My idea is kinda silly but maybe you can help?
  
  # ✅ GOOD PROMPT
  My half-baked idea is a social media app where you can only post *bad* news. 
  I know it sounds weird, but I'm serious. Take this idea and give me 10 feature names 
  that are witty and dark, each exploring a different angle of the concept.
  

  • Design Thinking Note: This is the Ideate stage, quantity over quality, and no judgment allowed. You own the seed; AI amplifies it.

• Prompting Technique 3: The 10x Expansion.

  • Why: This breaks functional fixedness. You use structured frameworks to force the AI to twist, bend, and combine ideas in ways your brain wouldn't.
  • How:
  plaintextCopy

  # ❌ BAD Prompt
  # Can you make this idea better?
  
  # ✅ GOOD Prompt
  I have an idea for a 'developer documentation summary' tool. 
  Take this concept and apply the SCAMPER framework: What can we Substitute? Combine? 
  Adapt? Modify? Put to another use? Eliminate? Reverse? Give me one concrete variation for each.
  

  • Design Thinking Note: You provide the raw material (the tool idea); AI provides the creative explosion using a proven framework.

✨ Bonus tip: I also typically use Deep Research in AI tools to get detailed insights and context to build upon and validate the idea in this phase. This helps ensure the brainstorming is grounded in reality and leverages up-to-date information. Learn more about divergent thinking here

Phase 2: Analysis (Pressure-Testing), aka Define & Test#

Now you have a hundred ideas. Are they any good? This is where design thinking's Define and Test stages come in: you clarify the real problem, poke holes in your ideas, and see what survives.

• Prompting Technique 1: The Pre-Mortem (or Grumpy CFO).

  • Why: This is the most powerful prompt I use. It instantly builds your case by finding all the holes before you invest months building the wrong thing.
  • How: You flip the persona from creator to critic with specific, targeted questions.
    plaintextCopy

    # ❌ BAD PROMPT
    # What are the problems with this idea?
    
    # ✅ GOOD PROMPT
    Act as a grumpy, risk-averse CFO who has seen 100 startups fail. I'm pitching 
    you an 'AI-powered documentation summarizer for developers.'
    
    Give me a brutally honest pre-mortem: it's now 18 months from launch, and the 
    product has failed. Tell me:
    1. The 3 most likely reasons it failed (ranked by probability)
    2. The hidden costs we didn't account for in month 1
    3. The one user adoption blocker we should have seen coming
    4. What successful competitor did that we didn't
    
    Be specific with numbers, timelines, and concrete scenarios.
    

  • Design Thinking Note: This is the Test stage, pressure-testing your ideas before you build. The more specific your failure scenarios, the better you can defend against them.

• Prompting Technique 2: The Red Team (or Rival CEO).

  • Why: This helps you find your competitive blind spots and forces you to think about defensibility, not just features.
  • How:
    plaintextCopy

    # ❌ BAD PROMPT
    # What would a competitor do?
    
    # ✅ GOOD PROMPT
    You are the CEO of [Real Competitor Name, e.g., Notion/Linear/Vercel]. You've 
    just seen my new feature pitch: 'Real-time collaborative debugging for remote teams.'
    
    You have 3 options:
    1. Build it yourself in 6 weeks
    2. Acquire a startup doing something similar
    3. Make it irrelevant by solving the underlying problem differently
    
    Which option do you choose and why? Walk me through your exact strategy to 
    neutralize my launch, including timeline, resources, and go-to-market.
    

  • Design Thinking Note: Still testing, but also defining what really matters. If your idea can be easily copied or made irrelevant, you know you need a deeper moat.

• Prompting Technique 3: The First Principles Validator.

  • Why: This helps you strip away the hype and validate if your idea actually solves a real problem that people will pay to fix.
  • How:
    plaintextCopy

    # ✅ GOOD PROMPT
    I have an idea for [your idea]. Before I build anything, help me validate the 
    first principles:
    
    1. What is the *actual* problem this solves? (Not the feature, the underlying pain)
    2. Who currently solves this problem, and how? (Existing alternatives, even manual ones)
    3. Why would someone switch from their current solution to mine? (What's 10x better?)
    4. What would need to be true about the world for this to be a $10M/year business?
    5. What's the smallest, ugliest version I could build to test if anyone actually cares?
    
    Be skeptical. Poke holes in my assumptions.
    

  • Design Thinking Note: This is the Define stage, making sure you're solving the right problem, not just falling in love with your solution.

Phase 3: Convergence (Finding the Gold), aka Prototype & Human Curation#

This is where you come back in. The AI generated, the AI critiqued, but you must curate. This is the Prototype stage, synthesizing, combining, and shaping ideas into something real.

• Prompting Technique 1: The AI Analyst.

  • Why: You're done with ideas and need structure. The AI can spot patterns and connections you might miss when you're drowning in options.
  • How:
    plaintextCopy

    # ❌ BAD PROMPT
    # Which of these ideas is best?
    
    # ✅ GOOD PROMPT
    I have 50 ideas from our brainstorming session. Analyze them and:
    1. Group them into 5 thematic categories based on the core problem they solve
    2. Identify which category has the highest potential impact based on market need
    3. For each category, highlight the 2-3 most unique or surprising ideas
    4. Flag any ideas that appeared multiple times with different angles (these might be worth exploring)
    

    plaintextCopy

    # ❌ BAD PROMPT
    # Combine these ideas into one.
    
    # ✅ GOOD PROMPT
    I've narrowed it down to ideas #4 (AI-powered inventory tracking), #19 (real-time 
    cost estimation), and #32 (supplier network integration). 
    
    Synthesize these into a single, cohesive product concept that:
    - Solves one core job-to-be-done
    - Has a clear primary user
    - Leverages the unique strength of each original idea
    
    Present it as: 1 product name, 1 elevator pitch (25 words max), and 3 core features.
    

  • Design Thinking Note: This is prototyping, combining and shaping ideas into something you can actually build. The more specific your convergence criteria, the better the synthesis.

• Prompting Technique 2: The Decision Matrix Builder.

  • Why: When you're stuck between multiple good options, you need a structured way to compare them against what actually matters to you.
  • How:
    plaintextCopy

    # ✅ GOOD PROMPT
    I'm torn between 3 final concepts. Create a decision matrix that scores each 
    concept (0-10) against these criteria:
    - Technical feasibility (can I build this in 3 months?)
    - Market differentiation (is this truly unique?)
    - Revenue potential (will people pay for this?)
    - Personal excitement (do I want to work on this for 2 years?)
    
    For each score, provide a 1-sentence justification. Then rank the concepts.
    

  • Design Thinking Note: This forces you to define your success criteria explicitly, making the final decision more objective and defensible.

• Prompting Technique 3: The Human Curator.

  • Why: This is the one step the AI can't do. This is taste. This is the shiver down the spine. This is where you override the data with your gut.
  • How: You read the synthesized options, the decision matrices, and the structured comparisons. You notice the idea that keeps pulling you back. You spot the unexpected connection between two seemingly unrelated concepts. You make the final, human decision based on what only you can feel.
  • Design Thinking Note: The final curation is pure human intuition, no framework can replace it. The AI gave you structure; now you bring the vision.

The Dark Side of the Sandbox (The Cons)#

This workflow isn't perfect. Using AI introduces its own set of very human problems.

1. The Information Overload Problem: The AI's ability to generate 1000 ideas in 10 seconds is a feature, but it's also a bug. You can drown in a sea of good enough options.

   • The Fix: Be brutally specific with your Divergence prompts. Don't ask for app ideas. Ask for 5 app ideas for the construction industry that solve inventory management. Constraints are your friend.

2. The Infinite Iteration Loop Problem: This is often the problem I face, every perfectionist's nightmare. Because you can always generate one more variation, or run one more pre-mortem, you can get stuck brainstorming forever. You never actually build anything.

   • The Fix: Time-box your sessions. Set a timer. Give yourself 20 minutes for Divergence, 10 for Analysis, and 10 for Convergence. At the end of 40 minutes, you must make a choice and move on. The goal isn't the perfect idea; it's a great, pressure-tested idea.

Your Future Isn't Generated; It's Curated#

We're at a new threshold. The research is clear: human brains have biases, and AI models have biases. Relying on either one alone is a recipe for failure, either through fear or through homogeneity.

Here's the paradox: AI ideas may increase collective diversity but not individual creativity.

When everyone uses AI tools the same way—asking for generic ideas from scratch—we all converge on the same statistically average outputs.

But when we each use AI differently—seeding it with our unique contexts, personas, and constraints—the collective range of ideas we explore as a group expands. The danger is that as individuals, we risk losing our unique creative voice, our idiosyncratic way of seeing problems that no one else does. This is why the curation step is non-negotiable—it's where you reclaim your individual creativity from the collective soup.

Don't use AI to think for you, and don't hand it a blank page before you've grounded yourself in the problem. Use it to think with you, once you've defined the context and your goals. Use it as the non-judgmental sparring partner who frees you from your fear of stupid ideas and lets you explore 100 paths you wouldn't have time for. But remember: the real work (the human curation, the taste, the judgment) remains irreplaceable. Creativity isn't about generating ideas; it's about choosing and building the right ones at the right time.

Now, go find your sparring partner.