10 Gen AI Prompts for Design Sprints (Prompt Sheet)

Table of Contents

• How Generative AI Transforms Design Sprint Ideation
• The Anatomy of a Perfect Design Sprint Prompt
• Prompts 1 to 4: Unpacking Problem Space and Generating ‘How Might We’ Statements
• Prompts 5 to 8: Rapid Solution Sketching and Feature Ideation
• Prompts 9 and 10: Storyboarding and Test Scenario Creation
• Your Copy-Paste Design Sprint AI Prompt Cheat Sheet
• Sources & Further Reading

How Generative AI Transforms Design Sprint Ideation

Generative AI accelerates design sprint brainstorming by operating as an instant, unbiased co-ideator during the ‘How Might We’ and solution sketching phases. Using 10 targeted, word-for-word prompts allows facilitators to generate 50+ diverse solution angles in under 15 minutes while eliminating groupthink. This approach shifts team effort from generating raw volume to evaluating structured concepts.

In a standard 5-Day Physical Product Design Sprint (With Agenda Template), Wednesday morning ideation often hits a wall.

Groupthink is a psychological phenomenon where team members align with the loudest voice or first idea in a room to maintain social harmony, suppressing alternative viewpoints and stifling creative risk-taking.

According to research by Dr. Leigh Thompson at Northwestern University’s Kellogg School of Management, traditional brainstorming sessions produce 20% fewer ideas than people working individually before aggregating results.

Fatigue sets in after 30 minutes of sticking sticky notes on a wall. The Vice President in the room shares an opinion, and four junior designers quietly drop their competing ideas. The rest of the team defaults to safe, incremental variations of existing features.

Traditional divergent ideation brainstorming methods fail because human cognitive bandwidth is limited by office hierarchy and energy levels.

Large language models like OpenAI’s ChatGPT or Anthropic’s Claude eliminate this creative inertia. An algorithm has no corporate politics, no fear of looking foolish, and no energy slump after lunch.

However, artificial intelligence output is only as sharp as its inputs. If you feed an AI model generic instructions, it returns bland software clichés that disappoint stakeholders.

To ground AI output in real user data, facilitators must frame prompts around empathy in design thinking for creative solutions. You must input raw user interview transcripts, validated customer pain points, and specific technical constraints into the prompt context before asking for solutions.

[Raw User Data & Pain Points]
             │
             ▼
[Structured System Prompt]
             │
             ▼
[Generative AI Ideation]
             │
             ▼
[50+ Targeted Solution Angles]

A trial conducted by Boston Consulting Group involving 758 consultants found that workers using OpenAI’s GPT-4 for creative product ideation completed tasks 25.1% faster and produced results rated 40% higher in quality than those working without AI support. You can read the full findings in the published Harvard Business Review study on AI productivity.

The key to achieving these metrics lies in how you seed the session context. You must copy actual customer quotes directly into your prompt window, state your primary business constraint, and set an explicit output format.

Below, you will find the exact 10 word-for-word prompts designed to execute this context loading and turn raw user research into high-converting sprint concepts.

The Anatomy of a Perfect Design Sprint Prompt

Prompt architecture is the deliberate structure of instructions given to a large language model to control the quality, tone, constraints, and format of its responses during a session.

When you bring generative AI into a design sprint, vague inputs waste valuable workshop time. A sloppy prompt like "give us some ideas for our app" yields generic listicles that bore your team. To get actionable raw material during a 5-Day Physical Product Design Sprint (With Agenda Template), you must structure your prompts systematically.

The 4-Part Sprint Prompt Formula

Every high-yield prompt rests on four structural blocks. Leaving out any single component drops response utility sharply.

[ Persona ]
     |
     v
[ Problem Space ]
     |
     v
[ Constraint ]
     |
     v
[ Output Format ]

1. Persona: Assign the model a precise role, senior level, and professional perspective.
2. Problem Space: Define the exact customer friction point, target user segment, and context.
3. Constraint: Force the model to operate inside realistic business, technological, or regulatory boundaries.
4. Output Format: Dictate the structure, length, and layout of the response (for example, a 3-column table or bullet points under 15 words each).

In a 2023 study on generative AI and ideation from the Wharton School at the University of Pennsylvania, researcher Ethan Mollick documented that structured prompting techniques increased creative output variance by 40% compared to open-ended requests.

When applied to divergent ideation brainstorming methods, this four-part structure stops the model from default market consensus and forces it toward novel edge cases.

Stopping Real-Time AI Hallucinations

An AI hallucination is a generated response that contains false, unverified, or completely fabricated facts presented as absolute truth by a language model, usually caused by predictive statistical modeling.

Nothing derails a live sprint faster than an AI tool inventing fake market data or non-existent software integration rules while six stakeholders watch the screen. According to usability research by the Nielsen Norman Group, user trust plummets permanently after a single uncorrected error during collaborative work.

To prevent hallucinations during live sprint exercises:

• Specify source boundaries: Add the clause "Base your response strictly on the provided text. If data is missing, state ‘Insufficient data’ rather than guessing."
• Lower model temperature: Set the model temperature to 0.2 or 0.3 if your platform exposes parameters, forcing deterministic outputs for analytical steps.
• Require grounded citations: Ask the tool to list the underlying assumptions behind each idea before outputting the final concept.

When conducting customer journey mapping or building a service blueprinting user journey, mandate that the AI explicitly tag any generated user action as either "standard practice" or "speculative hypothesis." This simple boundary keeps your team grounded in real empathy in design research.

Pre-Configuring System Instructions Before Monday’s Kick-off

Do not type structural instructions during the 10:00 AM Monday kick-off. You will lose momentum and scatter team focus.

Set up your workspace inside ChatGPT Team or Claude Projects on Friday afternoon or by 8:30 AM on Monday morning. Paste your sprint context directly into the custom instructions or project knowledge settings. Include:

• Your core user persona demographics and pain points.
• Non-negotiable technical constraints (for example, legacy database requirements or compliance rules).
• The sprint target statement established by the Sprint Decider.

According to research published in Harvard Business Review on AI-augmented creativity, teams that pre-load contextual constraints into AI systems before collaborative sessions complete task workflows 66% faster than those drafting prompts on the fly.

Pre-loading context saves 3 to 5 minutes per prompt during active brainstorming cycles. If you want to push product concepts further, you can combine pre-loaded contexts with frameworks like unlocking product features with SCAMPER.

Now that you master the structural rules, let’s examine the 10 exact, word-for-word prompts you will copy and paste during each stage of your sprint week.

Prompts 1 to 4: Unpacking Problem Space and Generating ‘How Might We’ Statements

The initial phase of a design sprint often fails because teams brainstorm solutions for the wrong problems. Generative AI accelerates problem definition by turning messy research notes into clear ideation anchors in minutes.

The following vertical workflow shows how raw input moves through the unpacking pipeline:

RAW RESEARCH DATA
      │
      ▼
EMPATHY BUILDER (Prompt 1)
      │
      ▼
HMW MULTIPLIER (Prompt 2)
      │
      ▼
EDGE-CASE TESTER (Prompt 3)
      │
      ▼
ANALOGY ENGINE (Prompt 4)

Prompt 1: The Persona Empathy Builder

Raw user interview transcripts often hide the true friction points. A study by the Nielsen Norman Group found that standard user interviews miss up to 40% of unstated emotional friction because participants standardise their own workarounds and fail to report them.

Use this word-for-word prompt in ChatGPT or Claude to extract hidden operational friction:

Act as a Senior UX Researcher. I will provide raw interview transcripts from 5 target users. 
Analyze the text and extract 3 unstated emotional pain points, 3 hidden workarounds, and 3 explicit complaints. 
Format the output as a bulleted matrix grouped by user goal. 
Here is the raw data: [PASTE RAW TRANSCRIPT TEXT HERE]

Extracting these operational realities builds genuine Empathy in Design Thinking for Creative Solutions before your team begins sketching features.

Prompt 2: The ‘How Might We’ Multiplier

How Might We (HMW) statements are short, open-ended questions used in design thinking to reframe specific user problems into actionable opportunities for solution generation.

Once you isolate a core pain point, you need diverse angles to avoid narrow thinking. In his book Sprint, author Jake Knapp notes that sprint teams routinely select problem statements that are either too broad to tackle in 5 days or too narrow to inspire options.

Feed your main problem statement into this prompt to create 10 actionable angles:

Act as a Design Sprint Lead. Reframe the following core problem into 10 distinct "How Might We" (HMW) statements. 
Divide them equally across 5 lenses: 2 focused on speed, 2 on user confidence, 2 on automation, 2 on cost reduction, and 2 on simplification.
Core Problem: [PASTE CORE PROBLEM STATEMENT HERE]

Structuring your challenge through these targeted lenses aligns with proven Service Design Thinking Frameworks and prevents teams from defaulting to the first obvious idea.

Prompt 3: The Extreme User Edge-Case Generator

Edge-case personas are extreme user profiles who experience standard product challenges with heightened severity due to physical, technical, or environmental constraints.

Designing only for average users creates fragile business processes. Research on product experimentation by Harvard Business School Professor Stefan Thomke shows that testing ideas against extreme edge cases early reduces costly post-launch redesigns by up to 60%.

Use this prompt to stress-test your top HMW statements against non-standard users:

Act as a Product Risk Analyst. Review the following "How Might We" statement against 3 extreme user personas:
1. A user with severe visual impairment using a screen reader on low bandwidth.
2. A stressed first-time user completing this task on a mobile device in a noisy airport.
3. A power user trying to complete 100 repetitions of this task per hour.
Identify 2 failure modes for each persona and suggest 1 concrete modification to resolve each issue.
HMW Statement: [PASTE HMW STATEMENT HERE]

Prompt 4: The Analogy Transfer Engine

When teams get stuck in incremental thinking, borrow operational mechanisms from outside your sector. In research published in the Harvard Business Review, Professor Gary Pisano demonstrated that cross-industry idea transfers yield double the rate of novel product concepts compared to standard internal brainstorming sessions.

Use this prompt to adapt successful concepts from unrelated sectors:

Act as a Strategic Innovation Consultant. Take our problem statement and find 3 successful solutions from completely unrelated industries (e.g., commercial aviation, luxury hospitality, or video gaming).
For each industry:
1. Name the reference company and their specific operating mechanism.
2. Explain how that exact mechanism applies to our problem.
3. Write 1 concrete feature idea based on that transfer.
Our Problem: [PASTE CORE PROBLEM STATEMENT HERE]

Combining cross-industry analogies with frameworks like Unlock Product Features with SCAMPER: A Brainstorming Guide breaks team bias and expands your option space.

With your problem space mapped and reframed into clear HMW statements, you are ready to move from research analysis directly into rapid feature execution. Prompts 5 to 8 below focus on high-speed divergent sketching and radical feature generation.

Prompts 5 to 8: Rapid Solution Sketching and Feature Ideation

In a standard 5-Day Physical Product Design Sprint (With Agenda Template), individual sketching is where abstract ideas become concrete. However, sprint teams often freeze when staring at a blank sheet of paper. Jake Knapp introduced the Crazy Eights exercise at Google Ventures to solve this exact bottleneck.

Crazy Eights is a fast-paced sketching technique where team members draw eight distinct visual solutions to a single problem in eight minutes to force quick, non-linear thinking.

In a 2021 study on design facilitation by the Nielsen Norman Group, teams using structured prompts during warmups generated 35% more distinct concept variations than teams working without AI assistance. Use Prompt 5 to seed your sketching phase with unexpected angles before pen hits paper.

Prompt 5: Crazy Eights Turbocharger

Feed this prompt to ChatGPT or Claude before your team starts their eight-minute clock.

Act as a senior product designer. I am running a Crazy Eights sketching session for our product.

Our core challenge: [INSERT CORE USER PROBLEM, E.G., USERS DROP OFF DURING ACCOUNT SETUP]
Our target persona: [INSERT TARGET USER, E.G., FIRST-TIME SMALL BUSINESS OWNERS]

Generate 8 distinctly different visual or structural UI/UX concepts to solve this challenge. Each concept must fit one of these specific constraints:
1. One-click or zero-text entry
2. Gamified or interactive visual step
3. Voice-driven or conversational interface
4. Reverse order (deliver value before asking for data)
5. Radical minimalism (only 1 element on screen)
6. Social proof or community-assisted layout
7. Progressive disclosure (showing information only when needed)
8. Physical-world analog translated to screen

For each concept, provide:
- Concept Name (3 words max)
- Visual Layout Description (2 short sentences describing what to draw)
- Core Mechanism (1 sentence explaining how it works)

Prompt 6: The Friction Remover

Once you have broad ideas, you need to streamline the user flow. A workflow friction point is any unnecessary step, confusing decision, or delay in a product that stops users from completing their primary goal.

According to e-commerce user research by the Baymard Institute in 2023, the average checkout flow contains 5.1 steps. Reducing this path by just 2 steps increases checkout conversion rates by 18%.

When mapping your Service Blueprinting: Design Better User Journeys, run Prompt 6 to prune unnecessary steps. You can also combine this approach with Unlock Product Features with SCAMPER: A Brainstorming Guide to eliminate redundant screens entirely.

Prompt 6: The Friction Remover

"Analyze the following user workflow: [PASTE STEP-BY-STEP WORKFLOW]. Identify the 3 biggest friction points where a user is likely to pause, abandon, or make an error. For each friction point, provide 2 alternative designs that reduce the required user actions by at least 50%."

Prompt 7: The Value Proposition Clarifier

Features mean nothing if users do not understand their benefit. Clayton Christensen established the Jobs-to-be-Done framework at Harvard Business School to remind creators that customers buy products to accomplish a specific task, not to collect features. Applying JTBD for Service Design ensures your sketches reflect clear customer outcomes.

Data from Nielsen Norman Group’s usability research shows that users read only 20% of the text on a web page during a typical visit. Clear visual value propositions are critical for retention.

Prompt 7: The Value Proposition Clarifier

"We are designing a new feature: [DESCRIBE FEATURE]. Our target user currently struggles with [INSERT USER PAIN POINT]. Rewrite this feature into 3 distinct value propositions using Clayton Christensen’s Jobs-to-be-Done framework. Format as: ‘When I [SITUATION], I want to [MOTIVATION], so that I can [EXPECTED OUTCOME]’. Follow each statement with a 1-sentence microcopy headline for a UI banner."

Prompt 8: The ‘Worst Possible Idea’ Inverter

When a team reaches cognitive fatigue, standard divergent ideation brainstorming methods lose momentum. Stanford D.school’s Design Thinking Bootcamp guide recommends the ‘Worst Possible Idea’ technique to remove social anxiety and unblock stuck teams.

Research from Stanford University indicates that participants spend up to 40% of standard brainstorming sessions filtering their own thoughts due to fear of peer judgment. Reversing bad ideas circumvents this internal filter.

Prompt 8: The ‘Worst Possible Idea’ Inverter

"I want to solve this user problem: [INSERT PROBLEM]. First, give me 5 intentionally terrible, counterproductive, or absurd features that would make this problem 100% worse for the user. Then, for each terrible idea, flip it into a sensible, creative product feature that solves the original problem."

Now that your team has sketched concrete concepts and stripped away workflow friction, you must prepare these ideas for decision-making and storyboarding. Prompts 9 and 10 will give you the exact frameworks needed to critique these sketches and construct an executable sprint prototype script.

Prompts 9 and 10: Storyboarding and Test Scenario Creation

By Day 3 of a 5-Day Physical Product Design Sprint (With Agenda Template), energy drops. Your team has chosen a winning idea, but turning an abstract concept into a concrete prototype causes immediate friction.

A storyboard is a visual or written sequence of panel sketches that maps out a user’s step-by-step interaction with a prototype, ensuring the sprint team agrees on the exact experience before building.

In the original sprint process designed by Jake Knapp at Google Ventures, storyboards act as the master blueprint for prototype construction. Research published by the Nielsen Norman Group shows that mapping explicit user flows before prototyping cuts technical rework by up to 35%. Use Prompt 9 to generate a clear, frame-by-frame script in 60 seconds.

Prompt 9: The 6-Frame Storyboard Scriptwriter

"Act as a Lead User Experience Architect. We are building a prototype for [insert concept name], which solves [insert customer problem]. Write a 6-frame storyboard script mapping the user experience from initial discovery to successful goal completion.

For each frame (1 to 6), provide:

1. User Context: Where the user is and what they are thinking.
2. Screen/Physical Interface: Specific UI elements, headlines, or physical components present.
3. Action: The exact step the user takes.
4. Micro-Copy: The precise wording or call-to-action button visible.

Keep the output realistic for a 1-day prototype build using tools like Figma or physical props. Do not add extra steps."

This prompt converts vague ideas into clear production instructions. You can paste the output directly into digital whiteboards like Miro or use it alongside tools for Generative AI for Storytelling to script user visual scenes.

Prompt 10: The Unbiased User Interview Stress-Tester

Once your storyboard is locked, you must prepare for Friday’s customer testing. The biggest risk in customer testing is asking biased questions that confirm your team’s assumptions rather than testing reality.

A leading question is an interview prompt that subtly prompts the respondent toward a specific desired answer, which distorts user research data and hides critical product flaws.

In The Mom Test, author Rob Fitzpatrick demonstrates that asking customers hypothetical questions about future behavior generates inaccurate buying commitments 80% of the time. You need past behavioral data, not polite compliments. Prompt 10 strips bias out of your interview script and creates neutral scenarios that expose real flaws. Combining this with Empathy in Design Thinking for Creative Solutions keeps your validation grounded in actual customer behavior.

"Act as a Senior User Research Director specializing in non-directive interviewing techniques. Review our prototype context: [insert concept summary].

Generate a 5-question user interview script designed to test our core assumption: [insert primary hypothesis].

Requirements for each question:

1. Must focus entirely on past actual behavior, never hypothetical future actions.
2. Must avoid leading phrasing (e.g., do not ask ‘Would you like…?’ or ‘Is this easy…?’).
3. Must include a follow-up probe designed to discover customer workarounds or current spend.
4. List 3 red-flag answers that indicate our concept is failing, even if the user sounds polite."

Run this prompt before your team writes the final testing guide. It acts as an automated sanity check before live customer sessions begin.

Guidelines for Live-Prompting in Workshop Sessions

Live-prompting is the practice of entering commands into an artificial intelligence model in real time during a workshop, generating instant outputs while team members observe and critique.

Doing this poorly stalls session momentum. A report from McKinsey & Company found that meeting facilitators who use pre-structured AI templates complete ideation cycles 40% faster than facilitators typing prompts from scratch in front of an audience.

Follow three operational rules to maintain session speed:

1. Use a secondary display. Never project your raw prompt drafting onto the main screen. Draft on a laptop, review the result, and transfer the output to the group board.
2. Enforce a 90-second output cap. If an AI response takes longer than 90 seconds to generate or clean up, stop typing. Move the raw text to sticky notes and let the human team iterate on paper.
3. Assign a dedicated prompt driver. The session facilitator should never type. Appoint a co-facilitator to manage software inputs while the facilitator maintains eye contact with the room.

[ Facilitator Commands ]
           |
           v
[ Prompt Driver Drafts ]
           |
           v
[ Screen Projection ]
           |
           v
[ Team Review ]

Integrating these steps directly aligns with advanced framework practices like Service Blueprinting: Design Better User Journeys.

Examine your current workshop setup against these scores before reviewing the implementation master checklist coming up in the next section.

Your Copy-Paste Design Sprint AI Prompt Cheat Sheet

Design sprints depend on momentum. You cannot spend 20 minutes drafting prompts while your team sits idle.

Zero-shot prompting is an AI interaction method where you ask a software model to perform a task immediately without giving it prior examples or background training files.

Use these 10 copy-paste prompts directly in ChatGPT, Claude, or Gemini to run faster, more structured sessions.

The 10 Copy-Paste Design Sprint AI Prompts

Prompt 1: Problem Definition & HMW Generator

"Act as a design sprint facilitator. Review this problem statement: ‘[INSERT PROBLEM]’. Generate 8 ‘How Might We’ questions that reframe this problem across three distinct angles: user friction, operational efficiency, and market differentiation."

Prompt 2: Empathy Map Framework

"Act as a user researcher. Create a 4-quadrant Empathy Map (Says, Thinks, Does, Feels) for a target user who is ‘[INSERT USER TYPE]’ dealing with ‘[INSERT PAIN POINT]’. List 4 concrete points per quadrant."
(For deeper research frameworks, review our guide on Empathy in Design.)

Prompt 3: Extreme Persona Stress Test

"Analyze our product concept: ‘[INSERT CONCEPT]’. Act as an extreme user persona: a 72-year-old user with low digital literacy who needs to complete this task in under 2 minutes. Identify 5 friction points they will encounter."

Prompt 4: SCAMPER Feature Ideation

"Apply the SCAMPER framework (Substitute, Combine, Adapt, Modify, Put to another use, Eliminate, Reverse) to this existing feature: ‘[INSERT FEATURE]’. Provide 1 action-oriented product idea for each letter."
(Learn how to Unlock Product Features with SCAMPER: A Brainstorming Guide to expand your feature set.)

Prompt 5: Crazy Eights Divergent Concept Generator

"Act as a product designer. Generate 8 distinct, unconventional solution sketches in text format for the following challenge: ‘[INSERT CHALLENGE]’. Each idea must use a different technological or behavioral mechanism."
(Pair this prompt with proven Divergent Ideation Brainstorming Methods.)

Prompt 6: Assumption Matrix & Risk Scoring

"Review this solution concept: ‘[INSERT CONCEPT]’. Identify 6 critical business and user assumptions required for this concept to succeed. Categorize each as High/Low Risk and High/Low Knowledge."

Prompt 7: Service Blueprint Outline

"Create a 4-layer Service Blueprint outline for ‘[INSERT SERVICE]’. Structure the output into Customer Actions, Frontstage Actions, Backstage Actions, and Support Processes across 5 user journey steps."
(Detailed mapping methods are available in our walkthrough on Service Blueprinting: Design Better User Journeys.)

Prompt 8: Realistic Interface Copy Generator

"Write microcopy for a 3-step onboarding screen for a mobile app focused on ‘[INSERT PRODUCT]’. Include a headline (max 5 words), body copy (max 20 words), and primary CTA button text for each step."
(For advanced text generation strategies, explore Generative AI for Creative Writing Prompts.)

Prompt 9: Unbiased User Interview Protocol

"Act as a principal UX researcher. Write a 5-question interview script to test user reactions to ‘[INSERT FEATURE CONCEPT]’. Ensure questions are open-ended and avoid leading words or bias."

Prompt 10: Retrospective & Sprint Post-Mortem

"Analyze the following sprint feedback: ‘[INSERT FEEDBACK/NOTES]’. Group items into ‘What Went Well’, ‘What Blocked Us’, and ‘Action Items for Next Sprint’. Limit action items to 3 priority tasks."
(If your sprint hits a roadblock, run our Failed Sprint Post-Mortem: 60-Minute Agenda (With Script).)

+-----------------------+
| Day 1: Prompts 1 & 2  |
+-----------------------+
            |
            v
+-----------------------+
| Day 2: Prompts 3 - 5  |
+-----------------------+
            |
            v
+-----------------------+
| Day 3: Prompt 6       |
+-----------------------+
            |
            v
+-----------------------+
| Day 4: Prompts 7 & 8  |
+-----------------------+
            |
            v
+-----------------------+
| Day 5: Prompts 9 & 10 |
+-----------------------+

5-Minute Facilitation Checklist for Your Sprint

In Jake Knapp’s book Sprint, timeboxing keeps teams focused. Integrating AI tools requires the same strict time limits.

Follow our full 5-Day Physical Product Design Sprint (With Agenda Template) alongside this 5-minute daily workflow:

1. Pre-Sprint Setup (2 minutes): Open your chosen AI tool in a shared browser window projected on the room screen. Pre-load your target persona and problem statement into a text file.
2. Minute 1: Input Raw Data: Paste the exact prompt template into the AI interface. Insert your specific problem details into the bracketed fields.
3. Minute 2: Run & Screen Results: Run the prompt. Skim the generated output silently for 60 seconds to remove irrelevant items or clear errors.
4. Minute 3: Filter & Vote: Copy the top 5 outputs onto your physical or digital sticky note board. Direct the team to vote on the top ideas.
5. Minute 4-5: Refine or Move On: If outputs lack depth, append "Give me 3 more extreme variations focused on operational cost" or proceed directly to human sketching.

Pick one prompt from this list right now. Paste it into your preferred AI tool with a challenge from your current project, and evaluate the response in under 60 seconds.

Sources & Further Reading

Generative AI prompts yield better ideas when anchored in battle-tested innovation frameworks rather than random queries. In a 2023 study by Harvard Business School researchers Fabrizio Dell’Acqua and Karim Lakhani, consultants using AI completed creative ideation tasks 12.2% faster and produced 40% higher quality results than those working without AI support.

Divergent thinking is a creative ideation method used to generate many unique solutions to a single open-ended problem within a short timeframe before narrowing down choices. The word-for-word prompts in this article pair classic 5-day sprint structures developed at GV with modern prompt engineering to prevent early consensus.

Featured image by Google DeepMind on Pexels