Harry Stebbings

And then I would say that the last one is... Can you train models on ambiguous externalities? And what I mean by that is like, yes, traditionally speaking, it takes 15 minutes to cook a pizza at Pronto Pizza. But they've got a new person this week and they're 10 minutes late because they're new. That's the challenge of real world, right? Or it's raining.

And so it takes longer for the driver to get there because it's raining. And so they're going to be five minutes late to pick up. Does data actually help in a highly ambiguous world?

And so it takes longer for the driver to get there because it's raining. And so they're going to be five minutes late to pick up. Does data actually help in a highly ambiguous world?

And so it takes longer for the driver to get there because it's raining. And so they're going to be five minutes late to pick up. Does data actually help in a highly ambiguous world?

One area for me that's so optimized is when you can actually stack orders with the driver and one can take six. How do you think about like, increasing unit economic, like unit efficiency, or whatever you want to call it, unit economics, improving them, whatever, but actually a slight deterioration in customer experience.

One area for me that's so optimized is when you can actually stack orders with the driver and one can take six. How do you think about like, increasing unit economic, like unit efficiency, or whatever you want to call it, unit economics, improving them, whatever, but actually a slight deterioration in customer experience.

One area for me that's so optimized is when you can actually stack orders with the driver and one can take six. How do you think about like, increasing unit economic, like unit efficiency, or whatever you want to call it, unit economics, improving them, whatever, but actually a slight deterioration in customer experience.

They're going to wait a little bit longer because the driver's doing five orders.

They're going to wait a little bit longer because the driver's doing five orders.

They're going to wait a little bit longer because the driver's doing five orders.

When we were chatting before, you said about speeding up market expansion to capture time.

When we were chatting before, you said about speeding up market expansion to capture time.

When we were chatting before, you said about speeding up market expansion to capture time.

What was the test?

What was the test?

What was the test?

With McDonald's.

With McDonald's.

With McDonald's.

So a lot of founders are told early on when they meet a big customer, like a McDonald's say, actually don't let it influence your product strategy. Don't be too concentrated in your customer dominance, whatever you want to call it. Would you say that's wrong advice?