The best ad Andrea Tortella has ever seen inside an AI conversation is his own face.

He knows how that sounds. "I'm gonna be selfish here," he said, laughing, when I asked him for it in the rapid-fire round. Thrad was among the first companies with access to ChatGPT ads, back when the only other advertisers in the room were, as he put it, "the biggest enterprises in the world," running the kind of careful, logo-forward creative that a legal department can sign off on without waking up. Tortella ran a headshot. "It performed extremely well," he said. "It had this, like, very surprising shock effect." Everybody else showed up as a logo. He showed up as a person.

That is the most Andrea Tortella thing I can tell you, and it is also, accidentally, his whole thesis about the medium. In a channel where nobody knows the rules yet, the human thing beats the safe thing. He is betting a company on that.

Tortella is the CEO and co-founder of Thrad, which calls itself the world's first supply-side platform for large language models. The company started as a university dissertation with a question that sounded, two years ago, like a provocation for a seminar room: what if ChatGPT had ads?

Then ChatGPT got ads. The seminar question became a market.

There was no single eureka moment, and he refuses to invent one. "It's hard to define one specific moment," he told me. "I think it was more something that grew into me, right? You know, when you're spending time in the shower and you just have like ideas coming to you." He and his co-founder, Marco Visentin, who left an Imperial College PhD to build Thrad with him, were already making advertising apps together. "We started flirting with this idea, which grew more into us. We built more conviction, and then that is really when we decided to go full time."

I asked him to explain Thrad to a media buyer who has never once thought about an ad inside a chatbot. He did it in two sentences, which is two fewer than most ad tech CEOs need to clear their throat. "Twenty twenty-five was all about GEO and AEO, right? Organic AI advertising. And so twenty twenty-six is the year that ads, paid advertising, enters LLMs." Thrad puts "actual sponsored paid suggestions inside of AI."

If you lived through the last twenty years of programmatic, the map Tortella draws will feel familiar, and that is on purpose.

On one side are what he calls the AI walled gardens, ChatGPT chief among them, building their own ad businesses. On the other is the AI open web: the long tail of independent chat apps, assistants and AI tools with real users and real token bills. Thrad sits across both. "We're very much building on top of all the conversational surfaces in the world," he said. "We both aggregate the AI walled gardens as well as the AI open web, enabling the long-tail AI applications to monetize." The pitch to developers is simple: let us sell your chat surface, and ads will pay for your tokens.

So I asked the obvious question. His Yoda is Brian O'Kelley, the man who built AppNexus and, with it, a good chunk of the plumbing that programmatic still runs on. Is Thrad to AI what AppNexus was to programmatic?

"Probably," he said, and then, to his credit, he added the honest part. Thrad was early. When it started, a lot of AI developers simply did not want ads, and he understood why. Nobody wanted to be the first app with a sponsored card in the answer while the competitor down the street stayed clean. "They would essentially have a first mover disadvantage," he said. "If I have ads and somebody else doesn't have ads, then, you know, people are gonna switch platform. And so people were scared, right, to have ads."

What changed was OpenAI. When the biggest chatbot in the world said it would run ads, it handed everyone else a permission slip. "This granted the legitimacy to independent AI developers to now monetize with ads," Tortella said. The stigma, he figured, would fade the way it faded on the web, where ads became the business model and everyone pretended they had always been fine with it.

The second thing that changed is growth, or the lack of it. Early on, the hot AI apps were growing so fast that nobody wanted to risk a single user on a sponsored suggestion. Then the curves bent. "We're even seeing ChatGPT itself," he said. "Now they have a billion users, but it's not like it's grown from like a billion to two billion very rapidly. They were at nine hundred million for a few months. Now they've reached a billion." When growth plateaus, monetization starts looking a lot less like betrayal and a lot more like rent.

I asked Tortella about Anthropic, which spent Super Bowl money this year promising that ads are coming to AI but not to Claude. A previous guest on this show, Phylicia Co, told me she thought that promise was a lie, either one Anthropic tells itself or one it tells everyone else. I wanted to know whether the man building the pipes agrees.

His answer was the most fun two minutes of the episode, because he did not call anyone a liar. He called it a math problem.

First, he loved the ad. "I actually find the Super Bowl ad that they did hilarious," he said. "I think it's super funny." Then he flipped it. "Paradoxically, they might have done a disservice to themselves by running that ad." A Super Bowl spot makes Claude more popular. More popular means more users. More users means a bigger bill for every free conversation. "What if that ad made them 10x Claude usage?" he asked. "Well, they would actually be a victim of their own success. And at half a billion users, you really need to have ads."

He was careful not to oversell it. In the short term, he thinks Anthropic is fine without ads: well funded, heavy on enterprise revenue, nowhere near forced. But he lined up the precedents anyway. Google said it was against ads in the early days. Meta got there. Netflix said never, and then built an ad tier. "They do have this positioning around safety," Tortella said, "but again, I think those can live together. Eventually, I do believe they have ads if growth keeps at the same pace."

I will be watching that one with more personal interest than most, for reasons the disclosure at the bottom of this issue makes clear.

Below the email line, Tortella explains why he thinks AI ads can be more private than the web ads you already hate, why the brightest college graduates no longer pick advertising, what Brian O'Kelley did with a Thrad baseball cap, and why he thinks the industry's favorite new phrase, "ads to agents," is nonsense.

He also explains why the only people he would take to a desert island are the people who already love him, which is either the most grounded answer in three years of this show or the least imaginative, and I have decided it is the first one.