Sam Altman argues agents make this the most ambitious moment ever to start a startup.
He calls it the real deal: an alignment failure and a security failure, serious even though it is not the biggest loss-of-control example. Anyone not at least a little scared or humbled is not taking it seriously. OpenAI made some big mistakes and the whole field will learn from it.
Ambitious and hard tech startups have always been hard and still will be, but the startup that took three months of nonstop work can now be made in 17 seconds. With three months of work and a lot of agents you can do unbelievable things, so a golden age is coming.
Figure out the really big brand new thing that is possible now and was not possible a few years ago. For OpenAI it was the deep learning magic moment the world did not update enough on. New exponentials are forming now; the world cannot intuit exponentials, and that is a superpower.
They used to joke that only 50 people in the world believed AGI was possible, but it was okay because 45 of them worked at OpenAI. You do not need a ton of people, and saying the heretical thing is probably why they all wanted to be together.
He would bet this generally cuts against many years of experience in favor of people who have a lot of fluency with the tools. Taste, agency and understanding the physics of business still count: where a valuable business can be built, and what a real network effect or moat looks like versus a fake one.
He was a very early investor in Stripe at about 22, and they asked him to drive down to Palo Alto for dinner to convince their first hire to drop out of school and join. That was Greg Brockman, and eight years later they started a company together.
He thinks the next six months will feel roughly equivalent to the last two years of model progress, so a very steep period. He adds there has never been a better time to do a startup than right now, and hopes that can be said again every year from now on.
Six and a half years ago the world token leader was an OpenAI employee using about 100,000 tokens a month while the worldwide per capita average was about zero. Now the worldwide average is around 100,000 and the leader uses hundreds of billions. In another six and a half years he expects 500 billion and quadrillions.
That it is all going to work out. His first startup did not work out great and it was a difficult time in his life, but the tech industry is very forgiving of mistakes. Keep all the drive and ambition, and just be a little happier along the way.
Every week, he thinks on Tuesdays, Paul Graham himself cooked dinner for the eight companies. They all walked in feeling hopeless and dejected, and went home convinced each of their startups was about to take over the world. Creating optimism and momentum out of nothing was a real PG special.
“Just like, highest confidence piece of advice here is just find a way to be mildly helpful to a lot of people.”
“It will poison your soul. It is a morally bankrupt thing to do.”
“Yeah, I think we're in the real deal phase of this.”
“I think concentration of power has basically been bad in every moment of human history, to varying degrees, of course.”
“I think it will feel like the next six months is like maybe equivalent to the last two years of model progress, something like that.”
“YC Founders' Journey.” https://www.youtube.com/watch?v=ZIaOBAjvc38. Transcript summary by WhipScribe, https://whipscribe.com/transcript-pages/yc-founders-journey.html.