Paul Graham’s core argument is deceptively simple: the way to get startup ideas is not to try to think of startup ideas. Instead, you look for problems, preferably problems you have yourself. This essay, originally published in 2012 and still widely referenced by founders, lays out a framework that Y Combinator has used to evaluate thousands of startups. The principles hold up because they address something fundamental about why most startups fail: they build things nobody wants.
Why “Thinking Up” Ideas Backfires
Graham draws a sharp distinction between two approaches. The first is sitting down and brainstorming startup ideas. The second is noticing problems organically as you go about your life and work. He argues the first approach is “doubly dangerous” because it doesn’t just produce few good ideas, it produces bad ideas that sound plausible enough to fool you into working on them.
The verb he wants founders to use is not “think up” but “notice.” When you force yourself to generate ideas, you tend to land on concepts that pattern-match to what a startup should look like without solving a real problem anyone urgently needs solved. These ideas pass a surface-level smell test. They sound reasonable in a pitch deck. But they don’t correspond to genuine demand, and you discover that only after months or years of building.
The Tarpit: Ideas That Trap Beginners
Certain categories of ideas attract first-time founders like magnets. Graham and Y Combinator have observed this pattern across thousands of applications. When YC first started reviewing proposals, they expected the most common pitch to be multiplayer games. It was actually the second most common. The most common was some combination of a blog, a calendar, a dating site, and a social network. These categories feel promising because they’re familiar to everyone, but that familiarity is exactly the problem. The ideas seem validated by the fact that you personally use similar products, when in reality they’re overcrowded spaces where the winners have already been decided by network effects and scale.
Graham has pointed to his own early startup as a cautionary example. He co-founded a company called Artix to put art galleries online. It seemed like an obvious opportunity, connecting galleries with a wider audience. But galleries didn’t actually want their inventory visible to random visitors, much like an antique store wouldn’t. The idea made sense from the outside but fell apart on contact with the actual customers. That disconnect, between what seems like it should be a problem and what actually is one, is what makes tarpit ideas so dangerous.
Organic Ideas and Living in the Future
The most successful startups almost all begin with what YC calls “organic” ideas: ones that grow naturally out of the founders’ own experiences. You encounter a problem in your work, your industry, or your daily life. You look for a solution. None exists, or the existing ones are terrible. You build something to fix it. That sequence produces ideas grounded in real demand because you are the demand.
To put yourself in a position where this happens, Graham recommends a specific approach: get yourself to the leading edge of some technology. Paul Buchheit, the creator of Gmail, described this as “living in the future.” When you’re working at the frontier of a field, whether that’s machine learning, biotech, hardware engineering, or something else, you encounter problems that most people don’t even know exist yet. Ideas that seem uncannily prescient to outsiders will feel obvious to you because you’re experiencing the gap between what’s possible and what’s available.
The practical formula is straightforward. Learn a lot about things that matter. Work on problems that interest you. Do it with people you like and respect. This isn’t motivational advice. It’s a method for turning your mind into the kind that generates startup ideas unconsciously, so that when you do notice a problem worth solving, you recognize it for what it is.
Schlep Blindness
One of Graham’s most useful concepts is “schlep blindness.” A schlep is a tedious, unpleasant task, a word borrowed from Yiddish that’s entered general American English. Schlep blindness is the phenomenon where your unconscious mind won’t even let you see ideas that involve painful, complicated work. You skip right past them without realizing you’ve done it.
Graham uses payments as his primary example. Before Stripe, the payment infrastructure underlying most internet commerce was clunky, outdated, and universally disliked by developers. The problem was right there in plain sight. But people’s unconscious minds shrank from the complications involved: dealing with banks, navigating regulations, handling fraud, integrating with legacy financial systems. It’s far more intimidating to start a company like that than to build a recipe site. So almost nobody tried, even though fixing payments was one of the most valuable things anyone could do for the internet economy.
The takeaway is to actively interrogate your own aversion. When you catch yourself thinking “that would be too hard” or “someone must already be working on that,” pause. Those reactions are often schlep blindness in action, and the difficulty is precisely what keeps competitors away. The most valuable startup ideas tend to look like work that nobody wants to do.
What Makes an Idea Good
Graham’s framework gives you a few filters to evaluate whether a problem is worth building a company around. First, is this a problem you personally have? If so, you understand the need at a level that’s hard to fake through market research. Second, is the problem something a small number of people want desperately, or something a large number of people want only mildly? He argues that you should strongly prefer the first scenario. A small group of users who love your product will tell others, give you feedback, and tolerate early-stage roughness. A large group of lukewarm users will churn before you’ve had time to improve.
Third, is this a problem that’s getting worse or becoming more common? The best startup ideas sit at the intersection of a real problem and a growing wave, whether that’s a new technology, a regulatory shift, or a demographic change. You want to be building something the world is moving toward, not something that requires you to push the world in a new direction through sheer willpower.
Where YC Sees Opportunity Now
Graham’s principles aren’t just theoretical. Y Combinator publishes a “Request for Startups” list that reflects the same logic: real problems, at the frontier, that most people shy away from because the work is hard. The current list leans heavily into AI-native applications, not chatbot wrappers but deep integrations. They’re looking for startups that rebuild entire software categories for AI agents as first-class users, replacing the bolted-on AI features that incumbents are shipping.
Other areas on the list include AI-powered personalized medicine, hardware supply chain optimization, counter-drone defense systems, semiconductor supply chains, and industrial capabilities in space like extracting raw materials from lunar regolith. They’re also interested in founders willing to take on entrenched enterprise software: chip design tools, ERP systems, industrial control platforms. These are exactly the kinds of schlep-heavy, technically demanding problems that Graham’s essay predicts will be underexplored by founders. The difficulty is the opportunity.
Putting the Framework Into Practice
If you take Graham’s essay seriously, the implication is that the best thing you can do right now is not to brainstorm. It’s to get deeper into a field you care about, work on hard problems within it, and pay close attention to the friction you encounter. Keep a running list of things that annoy you, processes that feel broken, tools that should exist but don’t. Talk to other people working in the same area and ask what frustrates them.
When something on that list keeps coming back, when you find yourself building a quick hack to solve it, when other people ask if they can use your hack, that’s an organic idea forming. The founders who do best at YC aren’t the ones who had a flash of genius. They’re the ones who were so close to a problem that the solution was almost inevitable.