Garry Tan

It's like in your body, do you feel more excited to come to work and why?

And for me, that answer is like when I get to talk to new people about new things instead of talking about the same people to the same things, right?

And so that's where AI is super good news for people who are super creative and highly agentic.

Like if you have agency, you can design your day to be exactly like that.

And then the second thing that I think is really important that we learned at YC is you have to write the prompts.

So we almost did this to ourselves where we use AI to like pick people.

And we always, we have for like many, many years, but the old AI and then the old AI was not smart.

It was, we call it feature-based.

It was literally just matrix math.

It was before the transformer.

It was before all of the crazy stuff that happened.

And that stuff was so dumb that it required like an incredible amount of data labeling.

And then what we found was the 15 partners started becoming data labelers under the API.

There's this very crazy idea that kind of Venkatesh Rao talks about.

He's this internet thinker that talked about all of the world is being recreated into these like Uber or DoorDash sort of API lines.

With AI, the danger is like everyone in the organization is either like above the API line or below it.

You want to be above it.

And then the ideal to me is as many people in the org as possible are above the API line.

And what that concretely means is actually from my partner, Pete Kuman, who figured out, actually, we need to allow every person to own their own and change their own prompt.

So when you use Google, for instance, and they have their AI product in your Gmail, it's kind of dumb and doesn't really do what you want.