Hard Startups

TL;DR: An easy startup is a headwind; a hard startup is a tailwind. This sentence is the most counterintuitive thing in startup advice and also one of the most reliably true. Ambition attracts everything — talent, advice, partnerships, capital — and easy startups attract none of those forces because nobody is excited about them.

What it means

Altman’s argument inverts the conventional risk-aversion logic: hard startups are actually easier than easy ones. When you’re working on something significant, people care about your success and want to help. It becomes a background gravitational force accelerating you toward the goal. Easy startups offer no such pull — every employee, every investor, every partnership has to be earned on the merits of the immediate transaction, with no ambient enthusiasm doing the work for you.

This connects to Altman’s broader thesis in How to Be Successful: compound growth solves almost all internal problems, and long-term thinking is the biggest competitive advantage available in modern markets. “I will fail many times, and I will be really right once” is the entrepreneur’s way, and it only works if the really right one is big enough to pay for all the failures plus the opportunity cost of trying.

The argument

Difficulty as moat. Hard problems deter competitors the same way counter-positioning deters incumbents — the perceived risk keeps others out. The startup’s willingness to tackle what others won’t is itself a form of defense (moats). If you’re starting a company in a space where the work is visibly hard, you’ve already filtered out the casual competitors before you’ve shipped anything.

Ambition attracts talent. “People want to be part of something exciting and feel that their work matters.” If everyone else is starting meme companies and you want to start a gene-editing company, do the gene-editing one. The recruiting advantage alone is worth the difficulty — you’ll pull engineers and scientists who would never have looked at the meme company, and they’ll work harder because the mission is real.

Optimism is a prerequisite. “It is very hard to do good work without being optimistic, exceptionally determined, and intellectually curious.” This isn’t soft advice — it’s a filter. The hard startup selects for founders with these traits, which in turn makes the company more likely to succeed. The easy startup selects for founders who are mostly just trying to make a living, which is fine but not the same thing.

The OpenAI version

Altman built OpenAI on this exact thesis, and the result is the most empirical validation of his own essay he could possibly have produced. OpenAI was unambiguously a hard startup — train models nobody had trained, on infrastructure that didn’t exist, with a business model nobody had figured out, in a regulatory environment that didn’t yet have rules. The hardness was the entire reason it could attract Ilya, Greg, Mira, and the rest of the founding research crew. An “easier” version of OpenAI would have lost the talent war before it started, and the company that eventually won would not have looked anything like OpenAI.