Winston Weinberg

I think just moving fast enough on product.

Like that is always, I think, the biggest existential threat for all the application layer companies.

And it's not necessarily that, you know, Anthropic or OpenAI are, you know, tomorrow going to put 50% of their resources after the legal vertical or tax vertical or anything like that.

But they're just improving their product and models.

And the value of your product is going to go down unless there is a massive delta between what your product does and what you could get from an enterprise GPT license, right?

And so it's just a constant existential threat of how do you make sure you get to like escape velocity on product so you have enough of a product mode for them to not run you over.

And I think about that daily.

Like when I'm thinking about competitors, the main thing I think about is just, I'm more bullish on these labs than most people, I think.

I mean, a lot of people are very bullish.

They have incredible talent.

And I think more about what are the frontier problems that our customers have that they're gonna solve later

Not entirely, but there were definitely... So we route basically based off of the use case, we route to the best combination of models and definitely our traffic went up to Opus 4.5 significantly.

Well, we aren't routing the majority yet, but even if we were, there's no conflict.

Basically, they want us to win and they want us to use the best model, right?

There's nothing in our agreement or our relationship or anything like that that you have to use OpenAI models.

And if anything, I think that for them, getting the feedback from application layer companies on this is where your models aren't doing as well, this is where your models perform super well, and this is where you need to improve is super valuable to them.

I think that we're seeing a plateau in performance for consumer use cases.

And the reason why I think this is a misnomer or something that people actually shouldn't pay attention to is we don't need them to be better for consumer use cases.

I feel like this is something that folks don't quite understand is a lot of the consumer use cases before were done.

You don't need better reasoning to solve these problems.