Y Combinator CEO on How AI is Changing What Startups Build

Asks how long until AI is as good at Gary's job as he is

Speculates AI coding agents won't need him in about 6-18 months, but knowing what to build stays human

AI will produce more winners than losers in aggregate and could raise living standards toward a near-utopian outcome

AI lets tiny teams reach tens of millions in revenue on small seed rounds, an unprecedented shift

This is great for founders but not necessarily for people who want to be employees at those companies

Those aspiring employees should start their own companies instead

AI-driven cost reductions now make previously unfundable hard-tech ideas viable

Raises whether the same drone logic could be misapplied to human targets

Any harmful application of the technology would simply be illegal

YC uses watcher AIs to monitor other agents and restricts write access to avoid agents acting outside intent

Notes rising backlash and a specific act of violence (Molotov cocktail at Altman's house) tied to AI opposition

Finds the violence personally concerning but believes it's premature for new AI laws and calls anti-AI scaremongering unproductive

Startups need open access to LLMs and platforms, including Chinese open-weight models, to preserve competition and choice

Questions why the origin of the model (American vs. Chinese) matters if access and safety can be managed

Slower corporate AI adoption is a beneficial safety valve for society, not a flaw

Notes that past technological innovations have not historically reduced working hours, questioning why AI would

AI-pilled people currently work more, not less, but he hopes AI ultimately enables more entrepreneurship, leisure, and 'third spaces'