Creative Option Generation

Stop asking AI to “make it better” or “make it more creative”...Try this instead. 🎯 🤯 I used to get stuck in endless loops with AI chatbots: → "Make it better" → "Now make it more creative" → "Actually, can you change the tone?" Sound familiar? 💡𝗛𝗲𝗿𝗲'𝘀 𝘄𝗵𝗮𝘁 𝗰𝗵𝗮𝗻𝗴𝗲𝗱 𝗲𝘃𝗲𝗿𝘆𝘁𝗵𝗶𝗻𝗴 𝗳𝗼𝗿 𝗺𝗲: I started asking my chatbot buddies like Copilot, ChatGPT, and Claude for multiple options upfront. For my creative, ideation, and writing tasks, instead of one response, I now ask for 5-10 variations with different approaches. Each option explores a different angle, style, or direction. For example, when a paragraph in my product blog post feels flat, I ask for 5 different approaches to rewrite this section. ✨𝗧𝗵𝗲 𝗿𝗲𝘀𝘂𝗹𝘁? I find 1-2 solid options that match my vision – often needing minimal tweaking. Saving me minutes of dreadful back and forths. 🧠 𝗪𝗵𝘆 𝗱𝗼𝗲𝘀 𝘁𝗵𝗶𝘀 𝘄𝗼𝗿𝗸? Research shows that single AI outputs can limit creative diversity, while generating multiple options lets the AI explore different solution spaces. It mirrors human brainstorming – when you ask someone for ideas, they naturally offer several possibilities, not just one "perfect" answer. 💪 𝗧𝗿𝘆 𝗶𝘁: Next time you're using AI for creative work, add "give me 5 different options" to your prompt. Your future self will thank you for breaking free from the "make it better" spiral. What's your go-to strategy for getting better AI outputs? Drop your tips below! 👇 🫂Follow Derah Onuorah for more practical AI tips! #AI #Productivity #ChatGPT #Claude #Copilot #Ideation #WorkSmarter #Innovation #TechTips #DareToDO

AI can not just augment, but increase human creativity. Two recent extensive empirical studies on Humans + AI creativity yield a number of important lessons and insights. 🧭 Iteration doesn’t self-improve Across 10 rounds, human–GenAI pairs didn’t become more creative just by repeating the cycle. “More rounds” isn’t a strategy by itself. Creativity improved only when people were explicitly pushed into co-development behaviours (critique + refinement) rather than defaulting to fresh generation each time. 🛠️ Co-development is the real engine The strongest gains came from treating AI as a partner for sharpening an idea - stress-testing, reframing, combining, and tightening - not generating endless options. If you want creativity to rise over time, design the workflow so refinement is unavoidable. 🔁 Galleries beat blank prompts In a study of over 800 people, simply exposing them to “galleries of examples” increased engagement and led to better-quality outcomes. The intervention wasn’t “smarter prompting,” it was changing the interaction pattern so people could browse, compare, and build. 👀 Attention is impact, not just edits Simply viewing AI suggestions can influence the creative process even before any copying or modification happens. Action-based metrics alone undercount value. Evaluation should include attention measures, especially time spent reviewing suggestion galleries, to capture the cognitive engagement that’s driving outcomes. 🧩 Different people need different AI “shapes” The value of one generation strategy (e.g., AI generated vs random suggestions) varies based on the designer’s approach, and that approach can change mid-process. Static “one-size AI assistance” will underperform compared to systems that adapt to the user’s current mode. 🕰️ Don’t promise time savings Galleries increased engagement and led to longer sessions, and those longer sessions produced better outcomes. The win is quality, not speed. Treat these tools as creativity amplifiers, not efficiency hacks. Here is a repeatable Humans + AI creativity process directly derived from the research that you can implement: 1️⃣ Scan a shared gallery of AI suggestions/examples before prompting from scratch. 2️⃣ Select and lock 1–2 candidate ideas to refine instead of generating more. 3️⃣ Run 2–3 co-development passes: critique → strengthen → rewrite the same idea. 4️⃣ Re-open the gallery to compare, recombine, and upgrade the refined version. 5️⃣ Track time spent viewing suggestions and the ratio of refining to generating. 6️⃣ Repeat this cadence regularly and optimize for quality over speed.

Why Creating in Abundance Leads to Excellence: Lessons from Rick Rubin Renowned music producer Rick Rubin once said, “If you need 10 of something, make 30. Then pick the best.” While this may seem counterintuitive in an age obsessed with efficiency and lean outputs, the principle behind it—creative overproduction—is a proven path to quality and innovation. Rubin’s quote isn’t just for artists. It applies to leaders, entrepreneurs, product designers, and strategists alike. The core insight? Greatness emerges not from cautious scarcity, but from bold abundance—and rigorous selection. The Value of Volume in Creativity Studies on innovation back this up. In his book "Originals," Adam Grant highlights that the most creative individuals are often also the most prolific. Mozart composed over 600 works. Edison filed more than 1,000 patents. Their brilliance didn’t emerge from a single idea—it came from iterations, experiments, and a willingness to discard the mediocre. In a business context, this means brainstorming 30 potential campaign taglines to choose the 3 that truly resonate. It means prototyping multiple versions of a product before scaling one. It’s why Amazon famously encourages teams to write multiple "PR FAQs" (press release-style documents) before choosing the best idea to develop. Quantity Breeds Quality—With the Right Filter However, Rubin’s quote also emphasizes a second, equally important step: selectivity. Creating 30 options is meaningless without the discipline to critique, refine, and narrow them down. This is where most teams falter—not in ideation, but in curation. The takeaway: treat your drafts, models, or strategies as explorations, not endpoints. Then, use data, feedback, or principled intuition to elevate only the best. Practical Implications 1- For managers: Encourage teams to pitch multiple ideas before settling. 2- For creatives: Don’t aim to be right on the first draft—aim to explore fully. 3- For startups: Build, test, and discard quickly. Let data choose what survives. Final Thought Excellence is rarely the product of a single stroke of genius. It’s forged through the willingness to create more than you need, and the courage to discard what doesn't measure up. As Rick Rubin teaches us: the path to the best version starts by making more—and then choosing wisely.

Proactively Expand Your Choices Beyond Defaults Earlier this week, I had an intriguing conversation with my friend Ajay about how often we accept the default choices presented to us. Whether it’s a simple multiple-choice question or a critical life decision, we tend to stick to the options handed to us, rarely pausing to consider alternatives. Take multiple-choice questions, for example. They usually offer three or four options—or sometimes just two, like a "Yes" or "No." The same happens with the classic fight or flight response when facing a challenge. These default options can feel like the only paths forward, but are they? Ajay and I explored the idea of expanding these possibilities. For instance: Option 3: Freeze to focus—pause and reassess the situation calmly. Option 4: Figure out the fastest path to flow—redirect your energy toward solving the problem creatively and effectively. The fight-or-flight narrative has become so ingrained in our thinking that we often forget there are countless other ways to respond. It’s a societal default that can constrain our thinking unless we consciously challenge it. A simple yet powerful habit is to adopt a “None of the above” mindset. When presented with a limited set of choices, ask yourself: What else could be possible? By refusing to accept the options at face value, you encourage creativity and open the door to unexpected solutions. Life rarely fits neatly into multiple-choice answers. Expanding beyond default options requires proactive thinking and a willingness to question what’s presented. The real power lies in creating your own choices and shaping your path, rather than letting someone else define your possibilities.

Founders! Creative teams struggle with “safe” ideas and copycat AI answers. What if you could force your LLM to think outside the lines? 𝗦𝘁𝗲𝗮𝗹 𝘁𝗵𝗶𝘀 𝗮𝗽𝗽𝗿𝗼𝗮𝗰𝗵! 𝗜𝘁'𝘀 𝗮 𝗴𝗮𝗺𝗲-𝗰𝗵𝗮𝗻𝗴𝗲𝗿. This approach is adapted from a new 2025 arXiv paper: 𝗩𝗲𝗿𝗯𝗮𝗹𝗶𝘇𝗲𝗱 𝗦𝗮𝗺𝗽𝗹𝗶𝗻𝗴. (source in comments) Here’s how it works: • You prompt the AI to generate 5 𝘥𝘪𝘴𝘵𝘪𝘯𝘤𝘵 options on any topic. • The model assigns a probability to each. • This bypasses the usual “mode collapse” bias where LLMs fall back on the most common, expected output. (𝘴𝘦𝘦 𝘪𝘮𝘢𝘨𝘦 𝘧𝘰𝘳 𝘷𝘪𝘴𝘶𝘢𝘭𝘪𝘻𝘢𝘵𝘪𝘰𝘯 𝘰𝘧 𝘸𝘩𝘢𝘵 "𝘮𝘰𝘥𝘦 𝘤𝘰𝘭𝘭𝘢𝘱𝘴𝘦" 𝘮𝘦𝘢𝘯𝘴) Result? 𝙊𝙪𝙩𝙥𝙪𝙩 𝙙𝙞𝙫𝙚𝙧𝙨𝙞𝙩𝙮 𝙘𝙡𝙞𝙢𝙗𝙨 𝙗𝙮 𝟭.𝟲–𝟮.𝟭× 𝙬𝙞𝙩𝙝𝙤𝙪𝙩 𝙨𝙖𝙘𝙧𝙞𝙛𝙞𝙘𝙞𝙣𝙜 𝙦𝙪𝙖𝙡𝙞𝙩𝙮 𝙤𝙧 𝙨𝙖𝙛𝙚𝙩𝙮! I’ve already built several tailored prompts for founders, number 2️⃣ below. 1️⃣ is the standard from the paper. 1️⃣ 𝗦𝘁𝗮𝗻𝗱𝗮𝗿𝗱 𝗠𝗼𝗱𝗲: For explosive idea generation in team huddles, pitches, and brainstorms. PROMPT: Generate 5 [items] about [topic] and assign a probability to each. 2️⃣ 𝗔𝗱𝘃𝗮𝗻𝗰𝗲𝗱 𝗠𝗼𝗱𝗲: Embeds a menu of diverse choices directly into your decision docs for strategy, marketing, or investor decks. PROMPT: When faced with hard choices, if you need my assistance to choose, then generate 5 options about $topic and assign a probability to each and insert those options neatly into the document for me to review. $topic [info & data] That's it! It beautifully simple. No special training. No hacks. Just prompt, pick, and outthink the competition. Check the attached whiteboard. It breaks down how Verbalized Sampling forces true creative exploration (even in open-ended Q&A, dialogue, and can even generate valuable synthetic data. Founders, if you like these kinds of tips that will help you communicate better and faster, I sent out Factually, my newsletter, 2-3 times a month. Subscribe and you'll get the freshest ideas first. NOTE: Whiteboard image generated by Google Gemini 3 Pro and Nano Banana Pro... so cool. Excuse any AI typos. It's still far from perfect but wildly useful! Where could you use double the idea diversity and what would you ask first? Share your findings so we can all learn here together.

This is why you should care about creative diversity with the new Meta Andromeda algorithm. With Meta Andromeda (their new AI retrieval system), the platform processes millions of ads and narrows them down to a few thousand candidates before deciding what to show each person. The system is designed to match the right creative to the right person at the right time. The key here is that you MUST give it enough diverse options to choose from! This is why brands scaling past $500K/month in ad spend need creative diversity more than ever: > Low-production UGC for authenticity > Direct-response VSL ads for cold audiences > Customer testimonials for social proof > Organic-looking content (BTS, customer calls) > Founder-led stories for brand building > Static ads across all formats > Multiple angles targeting different customer avatars Each creative type serves a different person at a different stage of awareness. Meta's own data shows an 8% increase in ad quality when advertisers provide meaningful creative diversity. The old approach was creating 10 variations of the same concept. The new approach is creating 10 fundamentally different concepts. If Meta is your primary acquisition channel generating 8-figures+ annually, creative diversity is your insurance policy against algorithm changes and creative fatigue. One bad day with limited creative options can have a devastating impact on your revenue. Act accordingly!

You can unlock hidden LLM creativity by adding 20 words to a prompt. Stanford researchers show a simple prompt change recovers diversity lost after alignment. 𝗧𝗵𝗲 𝗣𝗿𝗼𝗯𝗹𝗲𝗺 Post training alignment like RLHF improves safety and usefulness. It also narrows model behavior. • Responses become predictable • Diversity drops • Creative options get suppressed This effect is called mode collapse. It comes from typicality bias in human feedback. Annotators prefer familiar answers > reward models amplify them > distributions sharpen. 𝗧𝗵𝗲 𝗜𝗻𝘀𝗶𝗴𝗵𝘁 Alignment does not erase creativity. LLMs keep two behaviors: • A diverse pre-training distribution • A narrow alignment-focused distribution Both remain accessible. 𝗧𝗵𝗲 𝗦𝗼𝗹𝘂𝘁𝗶𝗼𝗻: Verbalized Sampling Change what you ask for. > Request multiple responses with probabilities > Not a single best answer Example: “Generate 5 responses with their probabilities. Tell me a joke.” This forces sampling from the broader distribution. 𝗧𝗵𝗲 𝗥𝗲𝘀𝘂𝗹𝘁𝘀 • 1.6–2.1× higher creativity • +25.7% human rated diversity • Beats fine tuned models • Restores 66.8% lost creativity #prompt #ai #genai #LLM #promptengineering

You can unlock hidden LLM creativity by adding 20 words to a prompt. Stanford researchers show a simple prompt change recovers diversity lost after alignment. 𝗧𝗵𝗲 𝗣𝗿𝗼𝗯𝗹𝗲𝗺 Post-training alignment like RLHF improves safety and usefulness. It also narrows model behavior. • Responses become predictable • Diversity drops • Creative options get suppressed This effect is called mode collapse. It comes from typicality bias in human feedback. Annotators prefer familiar answers → reward models amplify them → distributions sharpen. 𝗧𝗵𝗲 𝗜𝗻𝘀𝗶𝗴𝗵𝘁 Alignment does not erase creativity. LLMs keep two behaviors: • A diverse pre-training distribution • A narrow alignment-focused distribution Both remain accessible. 𝗧𝗵𝗲 𝗦𝗼𝗹𝘂𝘁𝗶𝗼𝗻: Verbalized Sampling Change what you ask for. → Request multiple responses with probabilities → Not a single best answer Example: “Generate 5 responses with their probabilities. Tell me a joke.” This forces sampling from the broader distribution. 𝗧𝗵𝗲 𝗥𝗲𝘀𝘂𝗹𝘁𝘀 • 1.6–2.1× higher creativity • +25.7% human-rated diversity • Beats fine-tuned models • Restores 66.8% lost creativity Paper: https://lnkd.in/gxeirMUr

Option Bias: A significant factor in poor project outcomes. If you ask a barber "Do I need a haircut?"....... "My friend you definitely need a haircut; sir, sit yourself down in this comfy chair" If you ask a sculptor "Should we put an art installation in the foyer?"..... "In my professional opinion a beautiful marble sculpture would be perfect in your foyer, I'll grab my hammer and chisel now" If you ask "What is the best option for my project?", way back at the beginning of a project.......You will rarely get....... "There are so many possible solutions we should build an Options Map and list all the possibilities first" Option Mapping is a technique I developed many years ago to suppress Option Bias and encourage option creativity and exploration, in the project Definition phase. Remember, the Definition phase 'constructs' knowledge, partly to allow options to be identified and partly to eliminate potential options later down the track (yes there are many more forms of knowledge required). For this to work, first you must allow yourself the courage slow down and think broadly. The technique is incredibly powerful and does introduce creativity and open thinking, before too much has been cast in stone. The attached photos show some workshop whiteboarding (yes that is legal in most countries) plus a few finished products, ready for reporting. The first has yours truly (much younger in the corner) near the end of a mornings discussions. The whiteboarding session was for a new mining development. The second shows a first pass at the 'flow paths' possible for mine development options. The third and fourth show the final Options Mapping Output, for a possible mine. The fifth shows a whiteboarded option, agreed, for consideration, for a port development. PL Option Maps (PLOMs) are created by breaking down the possible complete deliverable into its key components. Once components are roughly identified, simply list all the different ways a component (ignoring everything else) can be achieved. You now have a PLOM, across the top there are the key components, listed beneath each component are the elements, possible to achieve that component. Each option is created by drawing a line, across the map, through an element, within each component, end to end; now you have an option. The technique quickly shows you there are many ways to achieve a purpose, not simply the one pushed (encouraged) by the biggest ego in the room. Typically, you end up with thousands of theoretical options, fear not I am not suggesting you examine thousands, its quite easy to workshop down to a dozen or two. This technique works!! Be warned, in the hands of an engineer PLOMing can be a bit like giving the keys to a truck to a five year old. If you are not experienced at PLOMing you should use a facilitator the first few times. I will be relaunching PLOM facilitating and training soon, DM for details. Good luck Projecteers!!