Paul Graham Just Published 'Making Startups Powerful.' Here's the Test Your YC W27 Application Has to Pass

Paul Graham published a new essay this weekend, Making Startups Powerful, and it hit the Hacker News front page twice in one day. If you're applying to YC Winter 2027, you should read it as something more specific than startup philosophy: it is a rare, direct look at the actual questions YC partners ask themselves when they evaluate a company in office hours. Which means it's also a preview of how your application will be read.

Here's the test the essay describes, and how to run your W27 application through it before you submit.

What does "powerful" mean in PG's new essay?

PG's core heuristic is a question he asks startups in office hours: what would make this company more powerful? Not more profitable next quarter. More powerful. He argues that asking about money tends to yield incremental improvements, while asking about power can make a company orders of magnitude more valuable.

The essay then catalogs the transformations he probes for. Can you go from component supplier to owning the customer relationship? Can money (or tokens) flow through you? Can other companies build on top of you, app-store style? Can you add network effects to something that shouldn't have them? Can you go full stack and compete with your would-be customers instead of selling to them?

None of this is abstract for a YC applicant. These are, quite literally, the follow-up questions a YC partner is trained by culture and habit to ask about your idea. The application that anticipates them reads very differently from the one that doesn't.

Why this matters for your W27 application specifically

The YC application asks what you're making, who needs it, and how you'll grow. Most applicants answer those questions at the size of the product they've already built. PG's essay is a reminder that the reader is running a different program: they're mentally stretching your idea to see what it could become.

PayPal is his historical example: a demo of handheld-device security software that eBay sellers started "misusing" for payments, until the peripheral thing became the company. Rippling is his modern one: they started with onboarding software not because onboarding was the prize, but because that's where employee data enters a company, and being upstream of the data made everything else possible.

So when a partner reads your W27 application, the question isn't only "is this a good product?" It's "is there a bigger company hiding inside this idea, and do these founders see it?" You want the answer to the second half to be visibly yes.

How do you apply the "more powerful" test to your application?

Take your one-line description and run it through the essay's transformations. Concretely:

Are you a component, or do you own the customer?

If your answer to "what are you making?" is an API, a model, a data layer, or a tool that lives inside someone else's product, the partner reading it will wonder whether the companies you sell to will eventually own you. You don't have to change the business, but your application should show you've thought about the path from component to relationship. One sentence is enough.

Could money flow through you?

PG is blunt that it's always good when money flows through you. If you're building agent infrastructure, the question he says he'd ask first is whether the agents could pay one another through you, because that turns a service into a marketplace. If there's a marketplace version of your idea, name it in your application, even if you're not building it yet.

Where are your network effects?

The essay treats it as a challenge to find network effects in ideas that shouldn't have them, including the AI-native version: users opting in to let their interactions improve the model everyone else gets. If your product gets better with each additional user and your application doesn't say so, you're leaving your best growth answer on the table.

What's the tail that could wag the dog?

If users are doing something unexpected with your product, that anomaly belongs in your application. PG explicitly says misuse is a signal of desperate demand. A W27 application that says "we built X, but we noticed users doing Y with it, and we think Y might be the company" demonstrates exactly the kind of perception the essay describes.

The trap: ambition without the customer constraint

There's a failure mode here, and the essay names it. Every one of these transformations has to make things better for the customer. You can't bolt a marketplace or an app store onto your application because it sounds bigger. Partners read thousands of applications per batch cycle; inflated ambition without a customer reason reads as exactly what it is.

The other failure mode is the opposite one, and PG says it's more common: startups holding themselves back. A surprising amount of his advice, he writes, amounts to telling founders they don't need the complicated part, they can just do the simple, more ambitious thing. If your application spends three sentences explaining a workaround, ask whether the workaround exists because of a real constraint or because you're cowering.

One practical note on scope

PG's essay also argues that selling to early-stage companies makes you more powerful, because they decide fast and grow with you, the way Stripe signed up startups at inception. If your W27 application says your customers are enterprises, expect the interview question: who is the smallest, fastest-deciding customer who could use this today? Have an answer that isn't "we'll figure it out post-batch."

Run the test before YC does

The uncomfortable truth about this essay is that it hands every W27 applicant the grading rubric, which means the bar just went up for everyone. The founders who read it and revise will submit applications that already contain the answers to the stretching questions; everyone else will get those questions cold in a 10-minute interview.

This is where outside eyes help. It's hard to see the bigger company hiding inside your own idea, the same way it's hard to proofread your own writing. Before you submit, have someone who has been through YC run the "more powerful" test on your draft. YC Roaster exists for exactly this: YC alumni read your actual application and tell you, honestly, whether your idea is standing up straight or cowering.

PG ends the essay with a generator you can run in reverse: imagine the perfect world from your customer's point of view, and ask which component of it your startup could become. That's a good final edit for your application too. Describe the world where your customer wins, then show YC you're the piece of it that matters.