Frighteningly Ambitious Startup Ideas

Frighteningly Ambitious Startup Ideas. An essay by Paul Graham, March 2012.

Ideas that scare you

Graham noticed something odd at Y Combinator: the boldest startup ideas scare people, though any of them could make its founder a billionaire.

If you shrink back, that's not weakness; it's arguably sane. A giant idea makes you ask: am I ambitious enough?

Like the woman in Being John Malkovich who warns the nerdy hero he'd be out of his depth with her. Big ideas say that to us.

Invisible ideas

Graham counts this among the most important things about startups: you'd expect huge ideas to attract you, but they repel.

So when most people hunt for startup ideas, these never show up. The mind quietly screens them out.

Even very ambitious founders may do best to come at them sideways; more on that later.

A new search engine

Idea one: a new search engine. In Graham's words, “The best ideas are just on the right side of impossible.” In 2012 he saw cracks in Google.

Fear of Google pulled Microsoft into search, letting a rival set its agenda. Google, he hints, risks the same with Facebook. And its results felt cluttered.

To win, be the search engine the top 10,000 hackers use: a small base, but a powerful one, as Google itself once was.

You're one of them, so build the one you want. If you and your friends prefer it to Google, you're a tenth of the way to going public.

Replace email

Idea two: replace email. Your inbox is really a to-do list, and email is how tasks land on it. A terrible to-do list.

So build a protocol for tasks that gives the recipient more power: senders must say what they want, how much it matters, and by when.

Scary: entrenched protocols are nearly impossible to replace. But will email hell last another century? The powerful suffer from it too, so they may switch first.

Speed alone could win users; Gmail had grown painfully slow. And people would pay: Graham would gladly pay $50 a month.

Replace universities

Idea three: replace universities. They've lasted a thousand years, but Graham thinks American ones took a wrong turn lately. You could do far better for far less.

They won't vanish. They'll lose their monopoly on some kinds of learning as new ways to learn appear. Y Combinator is arguably one.

If learning splinters, credentials may split off, and campus life may need a stand-in. High schools come later: too much red tape.

Drama on the Internet

Idea four: drama on the Internet, which Graham says had beaten cable by 2012. TVs were so bad his family swapped theirs for an iMac.

Apps and games will steal some viewing time, but people will still want to sit back and watch a story: a series or a film, not clips.

Who delivers it? Maybe Netflix or Apple becomes the app store for shows. Or they overreach, and new firms sell payment and streaming to producers.

The next Steve Jobs

Idea five: the next Steve Jobs. Could Apple still invent without him? Someone who knew it well told Graham: no. Revenue may keep rising, but it lags.

So who makes the next iPad? No giant is run by a product visionary, and you can't hire one. You get one when a founder doesn't get fired.

Too ambitious? Apple did it, and a newcomer has Apple's example. Once Roger Bannister ran the first four-minute mile, others soon followed.

With Jobs gone, there's a vacuum. Users would follow a bold new hardware company, and its leader needn't match Jobs, just beat Samsung, HP and Nokia.

Bring back Moore's law

Idea six: bring back Moore's law. Until about 2002, chips kept getting faster. Now its denser circuits arrive as more processors, not faster ones.

So slow software no longer fixes itself. You used to just wait. Now you must rewrite it to run in parallel, which is far more work.

The prize: make many processors look like one fast one. The boldest way is automatic: the sufficiently smart compiler, a byword for impossible.

The impossible compiler

Is it truly impossible, though? Could no pattern of bits in a computer be that compiler? If you're sure, prove it; that would be news.

If it's merely hard, long odds are worth it. Sold as a web service, it would be like owning the only chips that still get faster.

Gentler routes: give programmers parallel building blocks, like Hadoop and MapReduce. Or hide people with good tools inside what looks like that compiler.

Graham admits it sounds crazy, and loves how many ways it's wrong. His inner forum troll can't decide where to object. That's a startup idea.

Ongoing diagnosis

Idea seven may meet even more resistance: ongoing, automatic diagnosis. Graham asks what about us will look backward a century from now.

Like waiting for symptoms. In 2004 Bill Clinton felt short of breath; his arteries turned out more than 90% blocked. Even he learned that late.

Doctors resist fishing for problems; it brings false alarms. A scan once left a friend of Graham's terrified for days, over a harmless cyst.

Graham calls that a limit of today's tools. Scan often, read scans well, and she'd know that cyst like a birthmark. Someday, people will pity us.

Start small

Graham's tactics: don't attack a problem this big head-on. Announce you'll replace email, and you get impatient backers and haters hoping you fail.

Just say you're making to-do list software. It sounds harmless. People can find out you've replaced email once it's done.

Big things tend to start deceptively small: Microsoft with Basic for a rare computer, Facebook with a site for Harvard students.

Head west

Starting small helps you too. Gates and Zuckerberg didn't know how big they'd get, and long-range plans are likelier to be wrong.

So don't plot a route to an exact spot. Be like Columbus, and just head west: any blueprint you draw now is almost surely wrong.

We picture visionaries seeing the future clearly. Graham suspects a blurry view is better: in the biggest successes, the first plan least resembles the result.

In short

Huge startup ideas frighten us, so most never see them.

Graham's seven all look close to impossible.

Don't charge in head-on. Start small, and seem harmless.

Then head west: a blurry view can beat a sharp one.

The original is longer, and better than this. Read it: paulgraham.com/ambitious.html