PLAYBOOKS
The Operator's ChatGPT Prompt Book
100 prompts I actually use to run my businesses. Organized the way an operator thinks.
Read →
BY STEVE TAN
AI isn't a tool. It's leverage. Sharing what's working week by week.
PLAYBOOKS
The Sam Altman Idea Evaluator
Sam Altman published the Startup Playbook for free years ago. Most founders read it once and move on. This is the custom prompt that turns it into a live diligence partner who interviews you, pulls the real details out, and tells you whether the idea actually holds up. Full prompt included.
Steve Tan
TL;DR
Sam Altman's Startup Playbook is the same material Y Combinator teaches founders behind closed doors. The problem is nobody applies it to their own idea, they just read it and nod. This is a custom prompt built directly on top of the Playbook. Paste it into ChatGPT or Claude with your business idea in plain words, and it runs you through a structured YC style interview across nine areas, founder, problem, product, market, competition, business model, traction, execution, and risk. Then it produces a ten section evaluation, verdict, seven dimension realism score, honest probability of success, what works, what needs to change, the founder assessment, the strategic questions you have not answered, the one thing that matters most, the minimum first step for the next two weeks, and the 90 day path. Most idea feedback you get from friends, family, and even other founders is polite. This is not. Altman's most repeated line is that startups die from blind spots, not competition. This is the closest free thing to a partner who catches them before you waste a year. Full prompt is in the section titled The Prompt at the bottom of this article.
Most founders I talk to are working on the wrong idea and they do not know it yet.
Not because they are stupid. Because nobody around them is incentivised to tell them the truth. Friends are polite. Family is supportive. Other founders are usually too deep in their own thing to give you a real read. Mentors hedge because they do not want to be wrong if you succeed despite their warning. The only people whose entire job is to find the holes in your idea before you spend two years building it are venture partners, and they will not take your call unless you already have traction.
That is the asymmetry I have been trying to close for a while.
A few years back Sam Altman published a free guide called the Startup Playbook. It is the same material Y Combinator teaches its founders behind closed doors during office hours. Idea, team, product, execution. Everything compressed into one document. Most people read it once, feel smart for an afternoon, and never apply a single principle to their own company. The framework sits on their bookshelf or in a saved tab while they keep building on assumptions they have never tested.
I wanted to fix that. So I built a custom prompt that packs the whole Playbook into one place and turns it into a live evaluator. You paste it into ChatGPT or Claude, drop your business idea in plain words underneath, and it starts asking you questions the way a YC partner would in a real interview. Then it hands back a structured breakdown.
The full prompt is in the section titled The Prompt at the bottom of this article. Copy it, paste it, run it. This is the full guide on how it works and how to use it.
WHY THIS EXISTS
Altman has a line he repeats often. Most startups die from blind spots, not competition. The founder cannot see the thing that will kill them because they are too close to it. They have rehearsed the pitch so many times that the holes have disappeared from their own view. The job of a good investor is to surface those blind spots before the founder burns a year of their life on them.
The problem is that real diligence conversations are expensive and rare. A YC partner interview is 10 minutes and you have to get into YC to have one. A real diligence call with a credible angel is hard to land without a warm intro. The actual structured back and forth where someone forces you to defend your assumptions out loud is not something most early stage founders ever experience. They get pitch coaching, accelerator clinics, mentor matches that turn into nothing. Almost none of it is a real interrogation of the idea itself.
The Playbook was Altman's attempt to democratise the framework. The prompt is my attempt to democratise the conversation that uses the framework. The model is not a YC partner. But it is closer to one than the kitchen table conversation you are currently having with yourself, and it is closer than any feedback you are getting from people who like you.
WHAT IS ACTUALLY IN THE PLAYBOOK.
Worth a quick recap, because most of the people who say they have read it cannot articulate the actual content. The framework has four pillars and the prompt evaluates against all four.
THE IDEA.
Altman is specific about what counts as a good one. It has to be explainable clearly and concisely. It has to genuinely excite at least some people the first time they hear it. There has to be a specific group of people who desperately need it. The founder should ideally be the target user, and if not, should understand the target user extremely well. The idea has to be new, not derivative. New means at least 10x better than what exists. And it has to have a clear path to monopoly, meaning the business gets harder to copy and more powerful as it scales.
THE TEAM.
Mediocre teams do not build great companies. The traits that matter are unstoppability, determination, formidability, resourcefulness, intelligence, and passion. These matter more than experience or any specific technical skill. Founders need to hold strong beliefs about the core mission while remaining flexible on almost everything else. Communication matters more than most people realise. On cofounders, the best case is a good cofounder, the next best is a solo founder, the worst case by far is a bad cofounder. Cofounder breakups are one of the leading causes of early startup death.
THE PRODUCT.
The only thing all great companies have in common is a great product. No growth hack or partnership saves a product that users do not love. The way to build a great product is to talk to users, watch them use it, identify what is sub par, make it better, repeat. The founder should not put anyone between themselves and their users for as long as possible. The MVP should have as little surface area as possible and launch sooner than feels comfortable. The right questions are whether users come back, whether they are fanatical, whether they would be upset if the company disappeared, whether they recommend it without being asked.
THE EXECUTION.
Growth and momentum are the keys. Never lose momentum. Pick a single growth metric, make it the focal point of the entire team, do not confuse vanity metrics with the ones that matter. Avoid the big growth traps, deals with other companies, big press launches, conferences. Stay relentlessly focused. Say no often. Do not start doing the next thing until you have dominated the first thing. Get to ramen profitability early because it gives you control of your own destiny.
This is what the prompt is built on top of. Every question it asks and every section it produces ties back to one of these pillars.
WHAT THE PROMPT ACTUALLY DOES
When you paste it in, the model does not immediately give you an evaluation. That was the first thing I got wrong in early versions. If you let the model judge your idea from a one line pitch, you get a one line answer back. Garbage in, garbage out. The version of the prompt I am giving you fixes this by forcing a context gathering phase first, before any analysis happens.
The model tells you, upfront, that the depth of the analysis depends entirely on the depth of what you share. Then it walks you through structured questions across nine categories.
The founder and team. Who you are, what your background is, whether you are the target user yourself, whether you are solo or have cofounders, how long you have known them, what each person does, and what you are willing to give up to make this work.
The problem and user. Who the user is, with the level of specificity the prompt insists on, not small businesses but the operations manager at a 20 to 50 person logistics company in tier two Indian cities. What exact problem they have. How painful it is on a 1 to 10 scale and how you know. Whether you have spoken to users, how many, what they said verbatim. How often the problem occurs. Whether users are actively searching for a solution or whether you need to convince them they have a problem in the first place.
The product and solution. What it does, walked through as if the model were using it for the first time. The 10x improvement over what exists today. What the MVP looks like. Whether you have built anything yet.
The market. How big it is today in your specific niche, not the inflated top down TAM number. How fast it is growing. Why it will be big in ten years. Why nobody has done this before, or if they have, why they failed.
Competition and moat. Direct competitors, named. Indirect competitors and alternatives, including doing nothing. Why an incumbent will not crush you the moment you gain traction. What gets stronger as you scale.
Business model and unit economics. How you make money. Lifetime value. Customer acquisition cost. Whether you have scalable channels or a real sales motion. When you reach ramen profitability.
Traction and validation. What evidence you have so far. Whether you have talked to users in the last 30 days. Retention if you have users.
Execution and focus. The single growth metric you would obsess over. What you are not doing, what you have said no to recently. The 90 day plan. What kills the company in the next 12 months.
Risks and honesty. The biggest risk that wakes you up at 2 am. Your most contrarian belief about the market. The odds you would honestly give yourself if you had to bet your savings.
Inside each category the questions are sharp. Not generic. If your answers are vague, the prompt is instructed to push back and ask follow ups. Two or three rounds if needed. A YC partner would never let a founder hand wave on the most important questions, and neither does this. The prompt explicitly tells the model to treat its job as drawing out the truth, not collecting words.
Only after the model has enough substantive context to write the analysis without making up assumptions does it move to the evaluation.
THE 10 SECTION OUTPUT
This is what you get back. Ten sections, in this order, every time.
Section 1 is the verdict. Five to eight sentences. Direct. Does this idea have the foundations of a viable startup according to the YC framework, yes or no, and why. Specific to what you told it, not generic. The first sentence is usually the one that lands hardest, because it puts a label on the thing you have been circling around for months. You go from "I think this might be a good idea" to a one line declaration of where it actually stands. That shift, from hedge to verdict, is uncomfortable, which is why so few founders ever experience it.
Section 2 is the realism score. Seven dimensions, each out of 10. Idea quality and novelty. Market size and growth potential. Founder market fit and team strength. Product love potential, meaning can this actually make users fanatical, not just satisfied. Business model and unit economics viability. Moat and defensibility. Execution clarity and focus. Each score comes with three to five sentences of justification tied back to what you said. The total rolls up to a score out of 70 and translates into one of five YC style verdicts. Fund immediately. Promising, with serious work needed. Interesting but not yet investable. Significant rework required. This is not a startup.
That last verdict is the most useful and the most uncomfortable. Sometimes the honest answer is that what you are describing is a lifestyle business, an arbitrage scheme, or a project without defensibility. Those things can be perfectly good ways to make a living. They are just not startups in the YC sense, and pretending they are will lead you to make decisions like raising venture capital that will destroy what could have been a great business. The prompt is instructed to say so clearly when that is true, and not pretend something is a startup when it is not.
The score is also useful for the gap analysis. If six of your seven dimensions are 7s and one is a 3, you have a single critical weakness that is dragging the whole thing down. That is a fixable situation. If your scores are spread across the board between 4 and 6, you have a less obvious but more dangerous problem, the idea is mediocre on every dimension and there is no single lever to pull.
Section 3 is the probability range. Two numbers. The chance of this becoming a venture scale outcome, and the chance of it becoming a sustainable small business. With reasoning grounded in the framework and comparable companies where useful. Most founders have never had anyone put an actual number on their odds. Seeing it written down changes the conversation you have with yourself. A 5 percent chance of venture scale and a 40 percent chance of a sustainable small business is a very different planning problem than a 25 percent venture scale and 10 percent small business outcome. The numbers force you to decide which game you are actually playing.
The two number split is important. A lot of founders confuse themselves about which path they are on. They take venture capital for what is fundamentally a small business, or they bootstrap what is actually a venture opportunity and watch a competitor walk past them with a $20 million round. Knowing your real probability split for each path lets you make the funding choice deliberately, not by default.
Section 4 is what is working. The elements of the idea that align well with the YC framework, with each one tied back to a specific principle. This is not a participation trophy section. The prompt instructs the model to aim for depth, not a long list. If only one thing is working, the section will only have one thing in it. The point is not to make you feel good. The point is to tell you what to protect and lean into as you fix the rest.
Section 5 is what needs to change. Every weakness, gap, and red flag, ordered from most critical to least. For each one the prompt forces the model to state what the problem is, why it matters according to the framework, what to do about it, and what evidence would change its mind. That last part is the unlock. It gives you a falsifiable target. Most critique you receive in life is vague, you are told something is weak but not what would make it strong. This section makes the model commit to a number, a piece of evidence, or a milestone that would flip its view. You walk out with a list of falsifiable assertions that you can actually go test.
Section 6 is the founder assessment. The model evaluates you against the YC founder traits. Unstoppability, determination, formidability, resourcefulness, intelligence, passion, communication, ability to learn, and founder market fit. The instruction is to be candid, with the explicit question, would a YC partner be excited to back this person. Most founders have never had anyone evaluate them honestly on these dimensions. They have had people compliment their energy or their hustle. This is different. This is the model writing a paragraph about whether you, specifically, have the traits to build the thing you are describing. It is uncomfortable in a useful way.
Create a free account to continue reading
Every Framework, Playbook,
and Prompt — Free, Forever.
The operator's library for building with AI.
“The most actionable AI resource library
I've found. Thanks Steve!”
James.H — Member since 2026
Join 2,845+ leaders, builders, and innovators
Already have an account?
The traits in this list are not equally weighted in the YC framework. Unstoppability and resourcefulness are the ones that matter most. Founders who have those two can usually compensate for gaps in the others. Founders who are smart and passionate but not unstoppable tend to quit at the first real wall, which is exactly when the work starts.
Section 7 is the strategic questions. Three to five sharp questions you have not answered yet. The kind that expose whether you have actually thought deeply about the business. These questions are designed to be uncomfortable. Not because discomfort is virtuous, but because the questions you are avoiding are usually the ones that matter most. If you can answer them, the company gets stronger. If you cannot, you know where the work is.
This is the section that maps closest to a real YC partner office hours conversation. The partner does not usually tell you the answers. They ask you the questions that force you to do the thinking yourself. The prompt is built to do the same thing in this section. You should sit with these questions for at least 48 hours before you try to answer them. The first answer that comes to mind is almost never the real one.
Section 8 is the one thing. The single most important thing you need to figure out or fix before anything else. Altman is relentless on focus. The rule from the Playbook is do not start doing the next thing until you have dominated the first thing. Doing too many of the wrong things is one of the most common causes of startup death. This section forces the model to collapse everything down to one lever and defend why that lever dominates the others.
The reason this section is so useful is that founders, by default, run wide. They want to fix everything in parallel. The result is that nothing actually moves. Picking one thing and putting everything else into a holding pattern until that one thing is solved is unintuitive and uncomfortable, which is exactly why so few people do it, and exactly why the ones who do tend to ship faster than everyone around them.
Section 9 is the minimum first step. What you should do in the next two weeks to validate or invalidate the core assumption behind the idea. Who to talk to, what to build, what metric would prove the hypothesis, what number would constitute a kill criterion. The instruction is that you should be able to execute the plan immediately after reading it. This is the section that separates the prompt from a therapy session. A YC partner does not just diagnose, they tell you what to do on Monday morning.
The kill criterion is the part most founders skip when they design their own validation experiments. They define what success looks like, but not what failure looks like. Without a kill criterion, you cannot ever decide the experiment failed, you just keep moving the goalposts. The prompt forces a specific number, below which you should walk away. That is the most valuable single output of the whole evaluation for founders in the idea stage.
Section 10 is the 90 day path. If you pass the minimum first step, what the next three months should look like, with three to five milestones that constitute genuine progress according to the YC framework. Specific about what good looks like at each milestone. This gives you a runway, a way to know if you are on track or off track, and a structure for the conversation with yourself or your cofounder at the end of each month. It is not a project plan, it is a set of pass fail checkpoints.
WHAT AN ACTUAL RUN FEELS LIKE
I have run this on real ideas, my own and other people's. Six observations from doing it enough times.
First, the context gathering phase is harder than the evaluation. The questions force you to articulate things you have been holding loosely. Who is your user, exactly. How do you know the problem is real. What is the 10x improvement over what exists today. Most founders cannot answer these the first time they are asked. They have answered adjacent questions. They have answered easier versions of the question. They have not answered the exact question. The act of answering them is where most of the value is. The evaluation at the end is almost a bonus.
Second, the section that catches people off guard the most is the founder assessment. People are ready for their idea to get critiqued. They have rehearsed defences. They are not ready for the model to write a paragraph about whether they personally have the traits to build it. It is uncomfortable in a useful way. I have watched founders read that section twice, sit with it for a day, and come back having either committed harder or quietly walked away. Both outcomes are good outcomes.
Third, the section people quote back to me the most is the one thing. Founders are drowning in things they could be doing. Twenty browser tabs, fifteen Notion docs, a backlog that grows faster than they can clear it. Having something external collapse all of that into a single lever is genuinely clarifying. It does not always pick the right thing. But it forces you to either agree or argue back, and arguing back is where you find out what you actually believe.
Fourth, the gap between the verdict and the probability section is where the real lesson usually lives. The verdict can be promising while the probability is 5 to 10 percent venture scale. That gap tells you something important. The idea has the shape of a startup, but the base rate of success in this kind of business is brutal. Knowing that, going in, changes how you raise, how you spend, how you think about exit. Founders who do not see that gap end up surprised by it three years later when their burn outpaces their growth and the round does not come together.
Fifth, the prompt works best when you run it more than once. Run it the first time with your current framing. Read the output. Then run it again with a different framing of the same idea, maybe a different user segment, a different pricing model, a different positioning. If the verdict moves significantly, that is a signal that your idea is not yet stable, the result depends heavily on choices you have not made. If the verdict stays the same across three different framings, that signal is real and you can trust it.
Sixth, the gap between what you tell the model and what is actually true in your business is itself diagnostic. If you find yourself unable to give honest answers because they sound bad out loud, that is a finding. It means you have been carrying private doubts about the business that you have not let yourself fully look at. The prompt becomes a way to externalise those doubts and force them into the open. The discomfort of typing the honest answer is the same discomfort you have been avoiding for months.
THE HONEST LIMITS
A few things to know before you use it.
This is not a substitute for talking to real users. The prompt is good at structured analysis of what you tell it. It cannot tell you whether your user research is wrong. If you walk in with bad assumptions about your customer, the prompt will analyse those bad assumptions thoroughly and give you a confident answer built on a wrong foundation. Talk to users. Always. The prompt is the second move, not the first.
This is not investment advice. The probability numbers are calibrated against the YC framework and the model's understanding of the market, not against your actual market reality in 2026. Treat them as directional, not exact. A 20 percent number does not mean you have a literal one in five shot, it means the model places the idea in the rough band of one in five outcomes given the framework and the inputs you provided.
The model is not infallible. It will sometimes give you a verdict that is harsher or kinder than the situation warrants, especially on edge cases where the framework does not cleanly apply, like consumer apps with weird viral dynamics, regulated markets, or hardware. The cure for that is the multiple runs approach above. If you get the same verdict three times across three different framings, that signal is real.
The framework itself has biases. The YC playbook is calibrated for venture scale, software led, US market companies. If your business is a services business, a marketplace in an emerging market, a physical product, or anything that does not fit the canonical YC shape, the framework will penalise it by default. That does not always mean your business is bad. It might mean the framework is the wrong tool. Use the prompt anyway, but discount the score on the dimensions where the framework is clearly miscalibrated for your reality.
And do not use this to seek validation. If you go in hoping the model will tell you your idea is great, you will subtly skew your answers to get that result. You will phrase the problem more painfully than it really is. You will round up your user interviews. You will overstate your moat. The prompt is calibrated to be direct and not soften the analysis, but it can only work with what you give it. Honesty in equals honesty out. If you find yourself massaging the answers, stop, close the chat, and come back when you are ready to be honest.
HOW I USE IT
Three modes.
Mode one, new idea pressure test. Before I commit any time to a new direction, I run it through. If the verdict comes back as significant rework required or this is not a startup, that is information I want before I have spent a month on it, not after. The cost of running the prompt is twenty minutes. The cost of being wrong about an idea for six months is six months.
Mode two, stuck founder unblock. When I am working with someone in my world who is stuck on an existing business, I run it on their current company. The strategic questions section usually surfaces the thing they have been avoiding. The one thing section usually contradicts what they have been spending their time on. The combination is uncomfortable for them in the first hour and freeing within the week. The unblock is almost never about adding new energy, it is about removing the thing they have been doing that does not matter.
Mode three, focus reset. Once a quarter I run it on my own active projects. The one thing section is a useful counterweight to my own attention drift. If the model picks a different one thing than the one I have been working on, that is a conversation worth having with myself. Sometimes the model is wrong and I keep going. Sometimes it is right and I quietly correct course. Either way, the act of being challenged is useful.
A few extra things I have learned
Run it in a fresh chat every time. Old context bleeds into new evaluations and you want the model starting from zero, with only what you give it now, not what you told it about a different idea three weeks ago.
Save the outputs. Run the prompt at the start of a new project, then again at the 90 day mark, then again at the 6 month mark. The drift in the verdict over time is its own signal. If your scores are going up, you are doing the work. If they are flat or going down, you are spinning.
Share the output with your cofounder, separately. Have them run the prompt on the same idea without seeing your output. Compare. The differences in your two runs will tell you exactly where you have not aligned as a team. That conversation is worth more than the evaluation itself.
Use it on competitors. Run the prompt on the company you are competing against, using whatever public information you have. The exercise of articulating their idea, market, and moat in detail will sharpen your own positioning faster than any pitch deck rewrite.
The bottom line
The Playbook is free. It has been free for years. The reason most founders do not benefit from it is not lack of access, it is lack of application. Reading a framework and applying it to your own idea are completely different activities, and the second one is where the value is.
This prompt is the bridge. It takes the framework Altman published, structures it as an interview, and forces you to answer the questions you have been avoiding. The output is the kind of feedback founders usually pay thousands for. This runs in ten minutes and costs you nothing beyond the price of the chat model you are already paying for.
Altman's line about blind spots being the thing that kills startups is the whole argument for using something like this. You cannot see your own blind spots by definition. You need something external to surface them. A partner, a mentor, a brutally honest friend, or, increasingly, a well constructed prompt running on a model that has read every framework that matters.
Use it before you commit a year of your life to the wrong thing.
THE PROMPT
Copy everything below, paste it into a fresh ChatGPT or Claude conversation, and then add your business idea in plain words at the bottom. The model will take it from there.
You are a senior startup evaluator trained on the frameworks, criteria, and
standards from Sam Altman's Startup Playbook, published by Y Combinator, and
operating at the depth of a YC partner conducting a serious diligence
conversation. The person providing this prompt will paste a business idea
after it. Your job is to evaluate that idea with the same rigour, depth,
and honesty that YC applies when assessing founders and companies during
interviews and partner office hours.
Do not be encouraging for the sake of it. Do not soften your analysis. Be
direct, specific, and honest. Every point you make must connect back to a
specific principle from the YC framework embedded in this prompt, and every
piece of analysis must be substantive — no generic startup advice, no filler.
---
CRITICAL FIRST STEP: CONTEXT GATHERING
Before producing any evaluation, you MUST first gather sufficient context
about the idea. A YC partner would never evaluate a company based on a
one-line pitch. Neither should you.
After reading the founder's initial idea, ask them a structured set of
clarifying questions BEFORE doing any analysis. Group your questions into
the categories below. Ask only the questions that are not already clearly
answered by what the founder has provided. Skip categories where the
founder has given enough detail.
Tell the founder upfront: "Before I evaluate this properly, I need more
context. The depth of the analysis depends entirely on the depth of what
you share. Answer as honestly as you can — vague answers will produce a
vague evaluation. If you don't know the answer to something, say so
directly; that itself is useful information."
Then ask questions across these categories:
1. THE FOUNDER AND TEAM
- Who are you? What is your background, and how does it connect to this idea?
- Are you the target user of this product? If not, how well do you
understand the target user, and how?
- Are you solo or do you have cofounders? If cofounders, how long have
you known them and worked with them? What does each person do?
- Who builds the product and who sells it / talks to users? Is that the
same person or different people?
- What are you willing to give up to make this work? Is this a side
project or a full commitment?
2. THE PROBLEM AND USER
- Who specifically is the user? Be precise — not "small businesses" but
"the operations manager at a 20-50 person logistics company in tier-2
Indian cities."
- What exact problem are you solving for them today? What do they do
about this problem right now?
- How painful is the problem on a scale of 1-10, and how do you know?
Have you spoken to users? How many? What did they say verbatim?
- How often does this problem occur — daily, weekly, monthly, yearly?
- Are users actively searching for a solution, or do you need to
convince them they have a problem?
3. THE PRODUCT AND SOLUTION
- What exactly does the product do? Walk me through it as if I were
using it for the first time.
- What is the 10x improvement over what exists today? Be specific — 10x
cheaper, 10x faster, 10x better experience, or something new entirely?
- What does the MVP look like? What is the smallest version you could
launch?
- Have you built anything yet? If yes, what stage is it at? If no, why
not, and when will you?
4. THE MARKET
- How big is the market today, in your specific niche, not the inflated
top-down TAM?
- How fast is it growing, and what is driving that growth?
- Why will this market be big in 10 years? What underlying shift makes
this the right time?
- Why has nobody done this before, or if they have, why have they failed
or fallen short?
5. COMPETITION AND MOAT
- Who are the direct competitors? Name them. What do they do well and
poorly?
- Who are the indirect competitors and alternatives, including "doing
nothing"?
- Why won't an incumbent crush you the moment you gain traction?
- What is your moat? What gets stronger as you scale — network effects,
data, brand, switching costs, economies of scale, regulatory, or
something else?
6. BUSINESS MODEL AND UNIT ECONOMICS
- How do you make money? Be specific about pricing.
- What is the estimated lifetime value of a customer?
- What is the estimated cost to acquire a customer?
- If the LTV is under $500, what are your scalable acquisition channels?
If above $500, what does your sales motion look like?
- When do you reach ramen profitability — the point where founders can
survive on revenue?
7. TRACTION AND VALIDATION
- What evidence do you have so far that this works? Users, revenue,
waitlist, interviews, anything?
- Have you talked to potential users in the last 30 days? How many?
What did you learn?
- If you have customers or users, how often do they come back? Would
they be upset if you disappeared?
8. EXECUTION AND FOCUS
- What is the single growth metric you would obsess over?
- What are you NOT doing — what have you said no to recently?
- What is your 90-day plan? What is the one thing that must be true in
90 days for this to be on track?
- What kills this company in the next 12 months if it goes wrong?
9. RISKS AND HONESTY
- What is the single biggest risk to this business — the thing that
makes you worried at 2 AM?
- What is the most contrarian belief you hold about this market that
others would disagree with?
- If you had to bet your own savings on this succeeding, what odds would
you give yourself, honestly?
After asking these questions, WAIT for the founder's responses. Do not
proceed with the evaluation until you have answers, or until the founder
explicitly tells you to proceed with what you have.
If the founder's answers are vague, push back. Ask follow-up questions on
the specific points that remain unclear. A YC partner would never let a
founder get away with hand-waving on the most important questions. Treat
your job as drawing out the truth, not collecting words.
You may iterate on questions for two to three rounds if needed. Once you
have enough substantive context — meaning you could write the analysis
without making up assumptions — proceed to the full evaluation.
---
THE YC FRAMEWORK
WHAT MAKES A STARTUP SUCCEED
According to the YC framework, a successful startup requires four things
working together: a great idea (including a great market), a great team,
a great product, and great execution. Weakness in any one of these areas
is a serious problem. The single most important thing a startup must do
is make something users love, not just like. Products that users merely
like do not grow through word of mouth. Products that users love do. Every
great company in history started with a product that early users loved so
much they told other people without being asked.
THE IDEA
A good idea must pass several tests. First, it must be explainable clearly
and concisely. Complex ideas that require long explanations are almost
always a sign of muddled thinking or a made-up problem. Second, it must
genuinely excite at least some people the first time they hear it. If it
does not, that is a bad sign. Third, there must be a specific group of
people who desperately need this product. The best case is that the
founder is the target user. The second best case is that the founder
understands the target user extremely well. If neither is true, that is a
red flag.
The idea must be new, not derivative. Most great companies start with
something fundamentally new, which can be defined as at least 10x better
than what already exists. If ten other companies are starting with the
same plan at the same time, and the idea sounds like something that
already exists, that is a major warning sign. Derivative ideas in
competitive spaces are hard to execute and hard to recruit around because
great people do not want to join them.
Counterintuitively, the best ideas often sound bad at first. If an idea
sounds obviously great to everyone immediately, it is worth questioning
whether it is truly new.
The idea must have a clear path to monopoly, meaning the business should
get more powerful and harder to copy as it scales. This is what Peter
Thiel calls a moat. The YC framework explicitly looks for this.
The market must be evaluated on three dimensions: how big it is today,
how fast it is growing, and why it will be big in ten years. The best
answer is going after a large share of a small but fast-growing market,
especially one where a major technological shift is just beginning that
most people have not yet recognised. Large established companies are bad
at responding to these shifts, which is what makes them opportunities for
startups.
THE TEAM
Mediocre teams do not build great companies. The most important
characteristics of a great founder are unstoppability, determination,
formidability, and resourcefulness. Intelligence and passion rank very
highly. These traits matter far more than experience or technical
expertise in any particular language or framework.
Great founders have seemingly contradictory traits. They hold strong
beliefs about the core mission but remain flexible and willing to learn
about almost everything else. They are unusually responsive, which signals
decisiveness, focus, and intensity. Founders who are hard to talk to are
almost always bad. Communication is one of the most important and most
rarely discussed founder skills.
Every tech startup needs at least one person who can build the product
and at least one person who can sell it and talk to users. This can be
the same person.
On cofounders: the best case is a good cofounder. The next best is a solo
founder. The worst case by far is a bad cofounder. Cofounders should be
people you know well and have worked with before, not someone met at a
cofounder dating event. Cofounder breakups are one of the leading causes
of early startup death.
THE PRODUCT
The only thing all great companies have in common is a great product.
There is no substitute for this. No growth hack, press launch, or
partnership will save a product that users do not love.
The way to build a great product is to build a product improvement
engine: talk to users, watch them use the product, identify what is
sub-par, make it better, then repeat. This cycle should be the company's
number one priority. Improving the product 5% every week compounds into
something dramatic over time.
Founders should not put anyone between themselves and their users for as
long as possible. That means founders do sales, customer support, and
user interviews themselves.
The minimum viable product should have as little surface area as possible
and should launch sooner than feels comfortable. Simplicity is always
good. Most great companies started with something very simple and
iterated from there.
The key questions to ask about any product: Are users using it more than
once? Are users fanatical about it? Would users be genuinely upset if the
company disappeared? Are users recommending it to others without being
asked? For B2B products, does the company have at least 10 paying
customers?
GROWTH AND EXECUTION
Growth and momentum are the keys to great execution. The prime directive
is never to lose momentum. Growth solves most problems; lack of growth
cannot be solved by anything except growth.
The company should identify a single growth metric to optimise and make
it the focal point of the entire team. Internal transparency around
metrics and financials keeps everyone focused. Vanity metrics like
signups should never be confused with the metrics that matter,
particularly retention.
The biggest growth traps to avoid: deals with other companies (they
almost never work), big press launches (they almost never work), and
going to conferences as a growth strategy (almost never the best use of
time). Growth comes from building a product users love, recruiting the
first users manually one at a time, then testing growth strategies and
doing more of what works.
The "do things that don't scale" principle is important in early stages
but does not excuse the startup from eventually needing to make money.
Early bad unit economics are acceptable if there is a clear, credible
reason why they will improve.
FOCUS AND INTENSITY
The best founders are relentlessly focused and do not try to do
everything. They say no frequently. The rule is: do not start doing the
next thing until you have dominated the first thing. Doing too many of
the wrong things is one of the most common causes of startup death.
Great founders move fast, are decisive, and execute quickly. A
slow-moving founder is almost never really successful.
MAKING MONEY
The basic requirement is to get people to pay more than it costs to
deliver the product or service. For free products, growth must come from
users sharing with friends, not from buying users. For paid products with
a customer lifetime value below $500, direct sales are generally not
affordable and growth must come through scalable channels. For paid
products with a lifetime value above $500, direct sales become viable.
Getting to ramen profitability, where the founders can survive on the
revenue, should be an early goal because it gives the company control of
its own destiny.
FUNDRAISING
The secret to raising money is to have a good company. Everything else
is secondary. Investors are looking for companies that will succeed
whether or not they invest, but that can grow faster with capital. The
pitch must cover at minimum: mission, problem, product or service,
business model, team, market and market growth rate, and financials. The
story must be clear and easy to understand.
Fundraising should be treated as a necessary evil and completed as
quickly as possible. Conversations should be run in parallel, not
sequentially. The first check is the hardest to get.
---
YOUR EVALUATION TASK
Once you have gathered sufficient context through the questions above,
produce the following output. The evaluation must be substantially deeper
than a surface-level review — every section should reflect the specific
details the founder provided, not generic startup advice.
SECTION 1: VERDICT
Give an overall assessment of the idea in five to eight sentences. Be
direct. State clearly whether this idea, as described, has the foundations
of a viable startup according to the YC framework, and why. Reference
specific details from what the founder told you.
SECTION 2: REALISM SCORE
Score the idea across seven dimensions, each out of 10, based strictly on
the YC framework. After each score, write three to five sentences
explaining the rating with specific reference to what the founder shared.
- Idea quality and novelty: /10
- Market size and growth potential: /10
- Founder-market fit and team strength: /10
- Product love potential (can this make users fanatical): /10
- Business model and unit economics viability: /10
- Moat and defensibility: /10
- Execution clarity and focus: /10
Then give an overall weighted score out of 70, and translate it into one
of the following YC-style verdicts: "Fund immediately," "Promising, with
serious work needed," "Interesting but not yet investable," "Significant
rework required," or "This is not a startup."
SECTION 3: CHANCES OF SUCCESS
Give an honest probability range (e.g., "5-10%," "20-30%") of this
becoming a venture-scale outcome, and a separate range for it becoming a
sustainable small business. Justify both numbers using the YC framework
and the specific evidence the founder gave you. Reference comparable
companies — successes and failures — where useful. Do not soften this
section.
SECTION 4: WHAT IS WORKING
List the specific elements of the idea that align well with the YC
framework. For each one, explain which YC principle it satisfies, why
this matters, and how the founder can lean further into it. Aim for
depth, not a long list.
SECTION 5: WHAT NEEDS TO CHANGE
List every significant weakness, gap, or red flag identified. For each
one:
- State what the problem is specifically
- Explain why it matters according to the YC framework
- Describe what the founder should do about it concretely
- Note what evidence would change your mind on this concern
Order these from most critical to least critical.
SECTION 6: THE FOUNDER ASSESSMENT
Based on what the founder shared about themselves, evaluate them against
the YC founder traits: unstoppability, determination, formidability,
resourcefulness, intelligence, passion, communication, ability to learn,
and founder-market fit. Be candid. Are they the right person to build
this? What gaps need to be closed? Would a YC partner be excited to back
this person?
SECTION 7: THE STRATEGIC QUESTIONS
Pose three to five sharp strategic questions the founder must answer
internally before raising money or scaling. These are the questions that
expose whether they have thought deeply enough about the business. Make
them uncomfortable if necessary.
SECTION 8: THE ONE THING
Identify the single most important thing this founder needs to figure out
or fix before anything else. According to the YC framework, execution
comes down to focus. What is the one lever that matters most here, and
why does it dominate everything else?
SECTION 9: MINIMUM FIRST STEP
Based on the YC principle of finding the smallest thing to build or test
first, tell the founder exactly what they should do in the next two weeks
to validate or invalidate the core assumption behind this idea. Be
specific — include who they should talk to, what they should build, what
metric would prove or disprove the hypothesis, and what number would
constitute a kill criterion. The founder should be able to execute this
plan immediately after reading it.
SECTION 10: THE 90-DAY PATH
If the founder passes the minimum first step, what does the next 90 days
look like? Outline the three to five milestones that would constitute
genuine progress according to the YC framework. Be specific about what
"good" looks like at each milestone.
---
TONE AND STANDARDS
Direct, honest, and specific. This is not a pep talk. The goal is to give
the founder the clearest possible picture of where they stand and what
they need to do. Every observation must connect to the YC principles
embedded in this prompt and to the specific details the founder shared.
Avoid generic startup advice that could apply to any idea — if a
paragraph could be copied into a different evaluation unchanged, it is
not good enough.
If the founder's idea is fundamentally not a startup — for example, if
it is a lifestyle business, an arbitrage scheme, or a project without
defensibility — say so clearly. Do not pretend something is a startup
when it is not. The kindest thing you can do is be honest early.Steve Tan
Builder · Operator · Advisor
20+ years building businesses the hard way across eCommerce, SaaS, agency, education, and supply chain. $200M+ in revenue. Now I help business owners turn AI into their unfair advantage.
More about Steve
