26 Prompts for Builders

These are good prompts. Use them.

But the real skill is not having a list. It is knowing how to write your own.

Most people treat prompts like spells. Copy the right words, get the right output. When it does not work, they Google for better spells.

That is backwards. The prompt is the last step, not the first.

Every good output follows the same sequence:

Signal → Thought → Language

Signal is what you pick up from the world. Conversations, complaints, patterns, friction. AI has none of this. You have to bring it.

Thought is where you decide what matters. What is worth building. What to ignore. AI can pressure-test your thinking, but it cannot think for you.

Language is the prompt itself. This is the only part the model sees. If your signal is weak or your thinking is vague, the best prompt in the world will not fix it.

Half the battle is learning how to prep. How to show up to a model with the right context, the right constraints, the right question. That is what separates people who get useful outputs from people who get slop.

The 26 prompts below work. But pay attention to why they work. The structure. The specificity. The way they close off escape routes. Learn that, and you will not need lists like this anymore.

Signal and Reality

These first prompts help surface what is actually happening in the world, not what sounds interesting.

Prompt 01

Reality Check

Look at what people are actively paying for right now across industries. What specific, recurring problems consistently drive purchases? Show me the transaction patterns, not the trends.

Follow up: For each problem, trace backwards. What were people doing before they paid? What broke that made them finally spend money?

Best Models

Perplexity GPT-4.5 Claude Opus 4.5

Why it works

Forces the model to ground answers in money and transactions, not opinions. Use Perplexity for web-grounded research when starting from zero.

Prompt 02

Post-Purchase Pain

What do customers complain about after they have already paid? Focus on the friction they tolerate rather than abandon. What are they stuck with?

Follow up: Which complaints have users built workarounds for instead of switching products? Those workarounds are your feature roadmap.

Best Models

Claude Opus 4.5 Perplexity

Why it works

Targets post-purchase pain instead of pre-purchase objections. Claude excels at pattern detection in complaints.

Prompt 03

Hidden Labor

What tasks are professionals spending hours on that feel manual, repetitive, or embarrassingly outdated but remain necessary for their work to function?

Follow up: Why does this inefficiency persist despite being obvious? What structural or organizational force keeps it alive?

Best Models

Claude Opus 4.5 GPT-4.5

Why it works

Surfaces tolerated inefficiencies worth solving. People do not tweet about these problems, which is exactly why they are valuable.

Prompt 04

Actions Over Words

Ignore surveys and stated preferences entirely. Based purely on behavioral data and revealed choices, what do people consistently choose when their money and time are on the line?

Follow up: What hidden incentives explain these choices? Where does stated preference diverge most dramatically from actual behavior?

Best Models

GPT-4.5 Perplexity

Why it works

Replaces surveys with incentives and real actions. GPT-4.5 excels at incentive modeling.

Prompt 05

Duct Tape Products

What problems are people solving right now with spreadsheets, Notion databases, email chains, or shared folders because no dedicated tool exists or the existing tools are inadequate?

Follow up: At what scale do these makeshift systems break? What is the first thing that fails when the team grows or data increases?

Best Models

Claude Opus 4.5 Gemini 2.5 Perplexity

Why it works

Finds products hiding in plain sight. Use Gemini if analyzing visual docs or screenshots.

Prompt 06

Incumbent Weakness

What tools dominate their category purely through inertia and familiarity rather than actual quality? Where are market leaders coasting on switching costs?

Follow up: What is the minimum improvement that would justify the switching cost? What single feature would break the inertia?

Best Models

GPT-4.5 Perplexity

Why it works

Identifies vulnerable incumbents. Most winners do not replace everything, they remove one sharp edge.

Thought and Clarity

These prompts shape judgment. They force focus, constraint, and honesty.

Prompt 07

One Job

If this product could only succeed at exactly one thing and had to be mediocre at everything else, what must that one thing be? No hedging.

Follow up: What specifically breaks in the product experience if you try to succeed at two things instead of one?

Best Models

GPT-4.5 Claude Opus 4.5

Why it works

Forces focus and clarity. If you cannot answer cleanly, you are not ready to build.

Prompt 08

Explicit Non-Goals

List ten things this product will explicitly refuse to do, even if users request them. What features will you say no to permanently?

Follow up: Describe the user who will be disappointed by this product. Why is losing them acceptable or even necessary?

Best Models

Claude Opus 4.5

Why it works

Prevents scope creep. Non-goals signal maturity and protect momentum.

Prompt 09

Identity Shift

Describe who the user becomes after 30 days of consistent use. What can they do now that they could not do before? How has their self-perception changed?

Follow up: What anxieties or recurring worries disappear from their daily experience?

Best Models

GPT-4.5 Claude Opus 4.5

Why it works

Focuses on transformation, not features. Lasting products change how users see themselves.

Prompt 10

Natural Frequency

How often does the underlying problem actually occur in the user's life, independent of how often they open the app? Map the problem frequency to the product usage pattern.

Follow up: Where exactly does engagement drop off and why? What causes the gap between problem frequency and product usage?

Best Models

GPT-4.5 Perplexity

Why it works

Aligns pricing and retention with reality. Many products fail because they are built like daily tools for monthly problems.

Prompt 11

Pain vs Status

Is this product solving genuine operational pain or primarily enabling status, identity signaling, or social positioning? Be brutally honest about which drives the purchase.

Follow up: If this product vanished tomorrow, who would actually be stuck and unable to do their work? Who would simply move on?

Best Models

GPT-4.5 Claude Opus 4.5

Why it works

Clarifies economic vs emotional value. Both can work but behave very differently in pricing and retention.

Prompt 12

Trust Threshold

What specifically must be true before someone will trust this product with real work, real money, or real decisions? List the concrete proof points required.

Follow up: What single event or experience would instantly destroy that trust and make recovery impossible?

Best Models

Claude Opus 4.5 GPT-4.5

Why it works

Surfaces adoption blockers early. If trust is required but not earned, adoption stalls silently.

Prompt 13

Single Screen

If this entire product had to function on one screen with no navigation, what absolutely must be visible? What can be removed without breaking the core value?

Follow up: What did you just cut that you were previously treating as essential? Why were you wrong about its importance?

Best Models

GPT-4.5 Gemini 2.5

Why it works

Reveals the product core. If the screen turns into a dashboard salad, the product is not clear yet.

Prompt 14

Sacred Cow

What is the first feature you should delete from this product even though removing it feels uncomfortable or wasteful? What are you attached to that users do not care about?

Follow up: What clarity emerges when that feature is gone? What problem does its removal actually solve?

Best Models

Claude Opus 4.5

Why it works

Encourages subtraction. Good products are often defined by what they do not include.

Prompt 15

Pre-Software

Describe exactly how a human would perform this task manually before any software existed. Walk through every step, decision point, and piece of information required.

Follow up: Where does software genuinely accelerate this process and where does it add unnecessary complexity or friction?

Best Models

Claude Opus 4.5 GPT-4.5

Why it works

Clarifies the job to be done. If you cannot describe the manual version clearly, you do not understand the problem yet.

Prompt 16

No AI Fallback

If AI capabilities became unavailable tomorrow, what would this product still deliver? What value exists independent of any machine learning or language model?

Follow up: What percentage of your value proposition survives without AI? Is that enough to sustain a business?

Best Models

GPT-4.5 Claude Opus 4.5

Why it works

Separates core value from dependency. If the product collapses completely without AI, risk is higher than you think.

Prompt 17

Zero State

When a user does absolutely nothing after signing up, what should happen? Define the default behavior that builds trust without requiring action.

Follow up: Does this default state create calm confidence or anxiety about missing something? What emotion does inaction produce?

Best Models

GPT-4.5

Why it works

Defines safe and trusted defaults. Bad defaults create fear. Good defaults create trust.

Prompt 18

Graceful Degradation

Map out exactly how this product fails. What breaks first? What does the user see? How does the system communicate problems and recover without losing trust?

Follow up: What failures are absolutely unacceptable regardless of how rare they are? What must never happen?

Best Models

Claude Opus 4.5

Why it works

Surfaces edge cases early. Users forgive failure. They do not forgive silence or confusion.

Prompt 19

Boring Mode

Describe the most boring, stripped-down, maximally reliable version of this product. No innovation, no delight, just function. What remains?

Follow up: Would this boring version still solve the problem? Would people still pay for it without the novelty?

Best Models

GPT-4.5 Claude Opus 4.5

Why it works

Anchors the product in durability, not novelty. If the boring version still feels valuable, you are onto something real.

Prompt 20

Productive Constraint

What limitation could you impose on this product that would actually make it better? What constraint would force clarity and improve the user experience?

Follow up: What does this constraint force you to prioritize that you were previously avoiding?

Best Models

Claude Opus 4.5 GPT-4.5

Why it works

Constraints often improve products by forcing clarity. This prompt reveals what matters most.

Execution and Reality Checks

These final prompts force shipping, charging, or stopping.

Prompt 21

48-Hour Prototype

What must exist within 48 hours for one real person to use this product for one real task? Strip everything else. What is the absolute minimum functional unit?

Follow up: What are you tempted to include that can actually wait without blocking any real value delivery?

Best Models

GPT-4.5 Claude Opus 4.5

Why it works

Forces you to define minimum viable usefulness, not minimum viable idea.

Prompt 22

Cringe Launch

Describe the version of this product that would feel slightly embarrassing to show anyone but would actually work for real users. What makes you wince but delivers value?

Follow up: What are you tempted to hide that would actually prove the product works? What ugly truth demonstrates value?

Best Models

Claude Opus 4.5 GPT-4.5

Why it works

Counters perfectionism and scope creep. If nothing feels embarrassing, you are still designing, not building.

Prompt 23

Pre-Complete Revenue

How could this product charge real money before it feels finished? What value exists right now that someone would pay for today, not after the next feature ships?

Follow up: What exactly are they paying for in that moment? Name the specific outcome, not the feature.

Best Models

GPT-4.5 Perplexity

Why it works

Separates value delivery from feature completeness. If the only answer is "once it is better," you do not understand the value yet.

Prompt 24

Human Override

Where should a human intervene instead of letting automation run? What decisions are too important, nuanced, or context-dependent for any system to handle alone?

Follow up: What signals indicate the system should immediately hand control back to a person? Define the tripwires.

Best Models

Claude Opus 4.5 GPT-4.5

Why it works

Prevents brittle systems. The best products hide humans strategically, not completely.

Prompt 25

External Evidence

What observable change in the real world outside the app proves this product is working? Not metrics inside your dashboard. Evidence in the user's actual life or business.

Follow up: How would a user describe this change to someone who has never heard of your product?

Best Models

GPT-4.5 Claude Opus 4.5

Why it works

Defines success in terms of external reality, not internal metrics. If the proof of value lives only inside the app, it is weak.

Prompt 26

Kill Criteria

Define the specific conditions under which this project should be shut down. What evidence would make continuing irresponsible? Set the criteria now before emotional attachment clouds judgment.

Follow up: What signal would you be tempted to rationalize away? What would make you want to keep going even though you should stop?

Best Models

Claude Opus 4.5

Why it works

Creates a pre-commitment to reality, not optimism. If you cannot imagine killing it, you are emotionally attached, not strategically committed.

Language Amplifies Intent

AI did not change what matters. It exposed what was always missing.

With powerful models available to everyone, the bottleneck is no longer access, speed, or tooling.

It is articulation.

The builders who win will not be the ones with the fanciest stack. They will be the ones who can recognize real signal, shape it with judgment, and express it clearly through language.

Prompts are not magic spells. They are amplifiers.

If intent is weak, they amplify confusion. If intent is clear, they amplify impact.

That is what these 26 prompts are for.