AI search start a perplexity in funding talks that would double the company's valuation to $18 billion.

It's the latest signal that the focus in AI is shifting beyond the powerful models and toward the user applications.

Dear Gebosa has more on today's tech check. Morning day.

Hey, good morning, Mike. So this round, it confirms a few key developments in AI at the moment. One, the application layer has arrived. Two, model building continues to become cheaper and more efficient. Now, these two trends working together, they really are dispelling the myth of the AI wrapper, which describes an AI app that's thin, merely an

interface on top of foundational models. That now undersells the true innovation that is

happening at the app layer, which is allowing perplexity to raise yet another round at a

potential $18 billion valuation, which would represent a-fold increase in less than just one year.

Now, a lot of this has to do with the fact that so-called rapper companies, they're going deeper

and developing their own AI tech as model building becomes cheaper and more efficient.

Earlier this year, perplexity CEO, Arvin Srinivas, he posted the myth of being

a rapper will take time to go away, but eventually it will. What people don't realize, he says,

is that the team, his team, can build models and serve them on their own. This week, I also spoke

to Shiv Rao. He's the CEO of another AI at company called A Bridge that simplifies paperwork for doctors

by turning conversations with patients into notes and billing codes.

It was last valued annually $3 billion.

He said that Rapper was an outdated term when there was an impression that the only way

to compete in AI would be to raise hundreds of millions of dollars and pre-trained huge models.

But now he says people have figured out that value moves up the stack.

We are actually more akin to some of those foundation model companies in that we're also

T-shaped in our own way. Now, not at the same scale, but we also have that level of science

because we're also leveraging

proprietary data sets. We're also fine-tuning and post-training our own sort of models that

...