Dylan Ratcliffe

through like hundreds and hundreds of interdependent pieces of infrastructure, 400 things and 2000 relationships.

And it's doing it perfectly.

And the magic definitely hasn't been lost on me.

Every time I test it, I'm like, God, this kicks ass.

It's fantastic.

And it continues to not be lost on it.

I'm always excited about the product, watching the product do what it does because it is awesome.

And the other thing is the team.

It's really important to me that the people who work for me leave in a better situation than they came in.

And I've been able to watch the team grow in terms of size, but also individuals stepping up and taking on new responsibilities and learning and growing and getting better at their jobs and taking on new jobs.

And that has been extremely satisfying.

It's probably 50-50, but it's really important to me.

The thing that has surprised me most, actually, in terms of things that I've been wrong about is, and it's more the team saving me from a mistake, but so that first trip in Greece when GPT-3 first came out, we were trying two things.

I wanted to see whether it was possible for us to determine the routineness of a change so that we could say, hey, you do this all the time.

Don't worry about it.

Or, hey, you've never done this before.

This is a problem.

The other thing we wanted to see whether we could do was to assess the risks of a change, of if you do this could go wrong.

I was totally convinced that the risks would not work.

I was like, there's no way that's going to work.