Episode 149: Steven Liss of OpenAds on AI start-up, AdCP, UCP, and how The Trade Desk stole his name

Summary

Marketecture: Get Smart. Fast.

Episode 149: Steven Liss of OpenAds on AI Start-up, AdCP, UCP, and how The Trade Desk stole his name
Date: November 21, 2025
Host: Ari Paparo (with Eric Franchi)
Guest: Steven Liss, CEO of OpenAds

---

Episode Overview

This episode features Steven Liss, CEO of OpenAds, who joins hosts Ari Paparo and Eric Franchi to discuss the evolution of OpenAds as an AI-native ad tech company, the latest industry standards in AI and ad tech, and what happens when a major player “borrows” your company name. The conversation explores the realities of using AI in both advertising creative and targeting, new standards like ADCP and UCP, and the challenges of interoperability in the era of embeddings.

---

Key Discussion Points & Insights

1. The Origin and Evolution of OpenAds

• Steven’s Background:
  • Originally, OpenAds set out to be the first AI-native ad network for running ads inside AI search and chat platforms.
  • [05:08 | Steven Liss]: “We built the technology… to run ads in AI search. It worked great, but we needed more supply… So, we realized… we could use AI for ads. And we are now using the same ad tech we pioneered in AI and bringing it to the open web in Programmatic Display.”
• Pivot in Strategy:
  • AI supply was dominated by OpenAI, and the market was more limited than hoped.
  • The company is now leveraging its AI for ads in programmatic display—focusing on contextual targeting and creative optimization in the open web.

---

2. The Name Game: The Trade Desk Confusion

• Issue: The Trade Desk launched a product called “Open Ads,” causing name confusion.
  • [05:45 | Ari Paparo]: “So how does it feel about the Trade Desk stealing your name?”
  • [06:21 | Steven Liss]: “It’s very confusing… it is quite likely that we OpenAds the DSP could end up bidding into the Trade Desk OpenAds.”
  • Steven jokes that he even had to explain transaction IDs to his mom because of the confusion.
  • Suggests The Trade Desk should switch their product name to “Open Auctions.”

---

3. Running Ads Inside AI: Lessons Learned

• Market Reality:
  • Few meaningful consumer AI platforms at scale beyond OpenAI.
  • [07:15 | Steven Liss]: “The bet we made two years ago was that AI was like mobile where there’s going to be an explosion of new consumer apps, when in fact… AI happens inside mobile.”
• Alternate Approach:
  • Choices for AI ad platforms: pursue long-tail AI-specific supply (e.g., chatbot companions) or apply AI technologies in existing publisher environments.

---

4. AI Companions and Contextual Ads

• AI Girlfriend Market:
  • [08:10 | Ari Paparo]: “The AI girlfriend market is equivalent to… in-app gaming—there's just attention.”
  • Steven clarifies both AI girlfriends and boyfriends are popular, and real emotional intelligence in AI companions sometimes surpasses the dating scene.
• Skepticism About Conversational Ads:
  • [09:21 | Steven Liss]: "I'm still very skeptical that people are going to talk to ads... I think the best use case of display inventory is catching someone's attention and driving recognition or a discrete action."

---

5. The Golden Age of Contextual Advertising — With AI

• Why Now:
  • AI profoundly improves contextual understanding, enabling nuanced ad targeting and creative alignment.
  • [10:33 | Steven Liss]: “What’s changing now with AI is you can derive a lot more meaning from the context on a page… we're entering a golden age for contextual advertising.”
• Industry Prediction:
  • Surge in contextual classification/verification vendors, sometimes bundled with brand safety.

---

6. Generative Creative & the Role of Purpose-Built DSPs

• How AI-Driven Creative Works:
  • Generates unique creative for each URL/context, optimizing resonance.
  • [10:50 | Ari Paparo]: “Is it like the creative says, ‘Oh, you’re interested in sports scores…’?”
  • [11:29 | Steven Liss]: “It’s this combination of contextual targeting with contextual creative… That’s why we realized we had to build our own DSP.”
• Building an AI Native DSP:
  • Not built on Beeswax; recent advances in coding models made it possible for a small team.
  • Coding models are said to have “improved the efficiency of every engineer on our team by multiples.” (12:15)

---

7. Artificial Creativity in Practice — Memorable Example

• Premium Beauty Brand Case:
  • AI generated an ad with a luxury lipstick on a messy bathroom sink, tagline: “Brand Name: The Only Essential.”
  • [13:58 | Steven Liss]: “It’s relatable and it solves the brand's problem of making it feel accessible and real to their target audience… that’s the sort of thing where… it wouldn’t hit nearly as hard as it would given that URL-specific context.”

---

8. The Standards: ADCP and UCP

A. ADCP (Advertising Data Collaboration Protocol):

• [15:28 | Steven Liss]: “ADCP is really laying out the spec for the rest of the ecosystem… closing the performance gap between open web and the walled gardens.”
• Key value is automating, scaling programmatic-direct deals, and enabling data-rich, efficient campaign planning.
• Prioritizing support for creative-focused ADCP elements due to agency/brand demand.

B. UCP (Universal Cleanroom Protocol):

• Enables privacy-safe sharing of user/audience data through embeddings.
• Technical challenge: Embeddings from different models aren’t interoperable.
• [19:59 | Steven Liss]: “Embeddings… are powerful, but the embeddings generated by one model are not at all interpretable by another model.”
• For contextual use, open-source models can work, but for audience signals, industry-specific open models are needed.
• [22:08 | Steven Liss]: “These open source models… weren’t trained on our audience signals… so you could have a data provider… donate an embeddings model with open weights.”

---

9. Notable Quotes & Moments

• Steven Liss’s Mom Reacts to Trade Desk Name Drama:
  • [05:58 | Steven Liss]: “She said that this was bullshit. Like, any DSP should be able to deduplicate incoming bid requests…”
• On Generative AI Creative:
  • [12:39 | Steven Liss]: “It’s like telling your creative team, you know, go lock yourself in a room and give me a thousand good ads… When you expose these models to all of the variety of the web, it acts like human inspiration.”
• On the Embeddings Dilemma:
  • [21:03 | Steven Liss]: “Who gets to decide which embeddings models we’re using? Are they open source or do we end up… with a branded provider?”
• Ari on AI Hype:
  • [47:56 | Ari Paparo]: "Everyone's using AI. If you're not using AI in your products ... what are you doing?"

---

Timestamps for Important Segments

| Segment | Timestamp | |-----------------------------------------------|--------------| | Introduction of Steven Liss & OpenAds | 04:54–05:39 | | The Trade Desk "Open Ads" Name Conflict | 05:39–06:43 | | Original AI Chatbot/AI Search Ad Model | 07:02–07:51 | | AI Girlfriend Market & Use Cases | 08:10–08:57 | | Contextual Advertising’s New Wave | 09:21–10:50 | | AI Creative and DSP Development | 11:29–12:39 | | Example: AI-Generated Premium Beauty Ad | 13:19–13:58 | | Standards: ADCP & UCP Explained | 15:17–22:47 | | Key Critique of Embeddings in UCP | 19:59–22:47 |

---

Tone & Style

The episode is candid, lightly irreverent, and packed with insider anecdotes—mixing high-level strategy with shop talk and occasional industry inside jokes. Ari and Eric keep the atmosphere collegial but challenge their guest’s ideas and provide their own perspectives.

---

Summary

This episode offers a deep dive into the practical and strategic implications of AI in advertising, from the limits of running ads inside conversational AI platforms to the emerging golden age of contextual and creative ad optimization powered by generative AI. It highlights the importance of new industry standards (ADCP, UCP), explores the technical challenges with AI interoperability, and provides an honest look at the evolving vendor landscape. If you’re interested in what the next wave of AI-driven ad tech looks like—and want a front-row seat to the debates shaping its standards—this episode is essential listening.

---

Transcript

[A]

This podcast is brought to you by Maloco. If you're an app marketer concentrating most of your budget in Google and Meta, you're not alone. 88% of consumer app ad spend stays inside walled gardens. Meanwhile, there are billions of daily users spending their time elsewhere in gaming apps, dating apps, productivity tools, you name it. That's where Maloco comes in. Maloco's AI powered platform connects you to those users, reaching 2 billion daily users across 3 million apps.

[B]

Ready to diversify.

[A]

Visit Malocoads.com and start scaling with confidence again. Go to Malocoads.com to learn more.

[B]

You're about to make a trade. Which u do you listen to? Is it get optioning those options.

[C]

Or.

[B]

Let'S do a little research. Learn [e-mail] TradeSmart Foreign. To the market sector podcast. This is Ari Paparo. I am with Eric Franchi and our guest today is Stephen Liss of the CEO of Open Ads. We have him here to talk about AI, which is always something we want to talk about, but also what's going on with this company and what's going on with the latest standards from the IB which I wrote about in the newsletter. And he has some interesting perspectives on Eric, you like to joke about how his, his company name was taken over.

[D]

Poor guy.

[B]

Poor guy.

[D]

The largest, you know, independent company at ad tech just decides to borrow your entire company name to, to talk about a new product and doesn't even give you the the heads up.

[B]

I'll have to ask him about that.

[D]

Yeah, let's ask him about that. But I really enjoy talking to Stephen and his co founder Michael Bishop, whatever I can. They're like just, you know, I think this next generation of founders that are so deep within AI that they're moving so quickly and they're surfacing insights that only builders can surface. I always learn something, so I'm looking forward.

[B]

Yeah. Another example of how the best way to get onto this podcast is not to have your PR agency email us, but just to shitpost on Twitter. That is that is the fast path to getting onto the market. Podcast, my guy. So next week is Thanksgiving. Do you have plans, Eric?

[C]

I do.

[D]

I do miss the. What about you?

[B]

Yeah, stay in New York. I'm doing the cooking. I saw this TikTok. I saw this TikTok about how, you know, pecan pie has been around forever, but corn syrup was only invented in the 1920s and that's the main ingredient in pecan pie. Right. I don't know if you ever made pecan pie, but corn Syrup is like the main. The thing that makes it sticky. It's what everyone uses. Not high fructose corn syrup, regular taro. Anyway, point being, like, so there's this effort to rediscover the pre1920s pecan pie that does not have corn syrup in it. And that's what I'm going to try to pull off. We'll see. It's like more of a custard based.

[D]

I mean, you got to give us a. An update on that.

[B]

Got to give an update on that.

[D]

Yeah. I so totally random. I order pecan pies from a place in Texas called the. The Berdahl Pecan Candy.

[B]

That sounds good company. But then it's. It's a pecan pie, right? Then if you're having from Texas, it's not a pecan pie.

[D]

Shout out to our Texas listeners. Anyway, what. What is Market doing for Thanksgiving?

[B]

Oh, I'm G.L. asked Eric. We are not recording with a guest because, you know, people will be. We record on Thursdays and instead we'll be watching the Detroit Lions and whatnot. So we're doing a mailbag episode. That means that your questions, the burning questions you have about this podcast about ad tech, about our opinions about Staten island pizza, about anything you want to ask us about, will be addressed next week. So that requires you to do something? What does it require them to do, Eric?

[D]

It requires them to hit us up via their favorite method to slide into DMs. So Ari and I are both available on LinkedIn and X. Do we want to put email addresses out there?

[B]

Sir, I think it's pretty easy to find our email addresses. And you in particular, as a vc, can probably get a lot of inbound. So let's not do that. We will also. Hey, look, I'm going to improvise. We'll take video submissions and audio submissions. If you want to ask the question by audio and video. Let's do it. Make it short. Don't go. That would be all length. But make it short and we'll potentially stitch it in. That would be awesome. I don't know how we'll do that technically, but we'll figure it out. All right, so you're listening to this probably on Friday or Saturday. Get those questions in right away because we'll be recording probably Monday or Tuesday. Looking forward to it. Hope everyone has a great Thanksgiving. And with that, we're gonna dive into our interview with Steven Liss for the CEO of openats. Welcome Stephen List, the CEO of Open Ads, to the show. Stephen, thanks for being here.

[C]

Yeah, great to be here.

[B]

And you were on Justify youy Existence a while back, right?

[C]

That's right. I think that was about a year and a half ago where we set out to be the first AI native advertising network running ads inside AI search and chat.

[B]

Yeah.

[C]

And what we've learned since then is really the only AI search and chat platform that matters at Adscale is OpenAI. So we built the technology as promised on Justify Our Existence to run ads in AI search. It worked great, but we needed more supply. And so we realized, well, we could be AI in ads, or we could use. Or be ads in AI, or we could use AI for ads. And we are now using the same ad tech we pioneered in AI and bring it to the open web in Programmatic Display.

[B]

Yeah. So, yeah, I want to get to that. But first, so how does it feel about the trade desk stealing your name?

[C]

So the worst part about this entire thing is I had to explain to my mom what transaction IDs are.

[B]

That is the worst part by far.

[C]

No, I think the.

[B]

Was your mom up in arms about transaction IDs? Was she. Was she upset?

[C]

You said that this was bullshit. Like, any DSP should be able to deduplicate incoming bid requests. Like she got immediately.

[B]

Let's get.

[D]

My mom said the same thing.

[B]

Let's get. Miss.

[C]

Miss.

[B]

Listen, Ms. Franchi, on the show, the Mom's episode, Mother's Day next year we'll have a Mom's episode. So reality is, the trade desk is using the phrase open ads. And it's a little confusing.

[C]

It's very confusing. And even more so because the product itself, an open auction, is not a bad idea. Which means it is quite likely that we open ads. The DSP could end up bidding into the trade desk open ads. So I would suggest they trade it in for a better name like open auctions. But we'll see where that lands.

[B]

And just wait till OpenAI launches their ads product. That's going to really help your marketing, too.

[C]

Well, I think there's a lot of draw to the concept of openness in advertising, and I think that's what we're going to get into with all these AI frameworks is how do you cooperate among different agents, different entities in the environment?

[B]

All right, let's talk about the old business model, putting ads into AI results. There were like, there are a lot of startups trying to do that. That is not a unique idea you guys had. Are they all going to come to the same conclusion you did?

[C]

So they are either going to have to now really. The bet we made two years ago was that AI was like mobile where there's going to be an explosion of new consumer apps when in fact, well, AI happens inside mobile. So it's actually competitive there. I think there will be an opportunity if AR or VR takes off like that. But the choice for every AI ad platform is either go after this long tail supply. So there is some long tail. There's a lot of student focused apps, there's companion apps that can run a lot of ads in AI Girlfriends or you take this technology to other publisher services.

[B]

Yeah, the AI girlfriend market is effectively and it's equivalent to AI gaming. Right. I'm sorry to. In app gaming there's no context, there's just attention. And the only ads that'll work will be like the applovin style, you know, video watching ads.

[C]

I think that's mostly right. No, your AI companion boyfriends or girlfriends is actually fairly split evenly.

[B]

Yeah, sorry, I didn't mean to disparage.

[C]

Show on Disney tonight. No, it's free for the first month and I just feel like that's kind of cringe. I'd rather no sell normal ads in normal places.

[B]

Sorry. A side note. What do you call a female incel. Are they also called incels or is there a separate name for them or they're Fincels.

[C]

I think Fincel works.

[B]

Fincel does work. Eric, you're offended by this line of questioning?

[D]

It's strange. It's quite strange. I'm sorry.

[B]

Okay.

[C]

I was having a conversation out here in San Francisco selling some ads and they were saying no, it's tough as women dating here. There's a lot of narcissists and no, the AIs are actually quite emotionally intelligent.

[B]

Sure, let's replace the men with AIs. It makes sense. So moving on to the new business model. So AI inside the ads. This is also somewhat crowded space, I guess. Talk about the best use cases here. I've seen demos where it's just like a banner ad as a.

[C]

Has a.

[B]

Has a. You know, ask us anything about this product thing. Is that. Is that the cutting edge or.

[C]

Or we ra about 30 million of those in production and we found that. No, I'm still very skeptical that people are going to talk to ads. I think the best use case of display inventory is catching someone's attention and either driving recognition or a discrete action right there. And I'd say the space that's about to become crowded that we're very excited about is I Think we're entering a golden age for contextual advertising for a long time, that we've had IB content taxonomies in the RTP spec for years. And well, the buy side doesn't listen, partly because you can't necessarily trust the publishers to be honest with them and partly because the context wasn't nuanced enough to really drive better performance. And what's changing now with AI is you can derive a lot more meaning from the context on a page. So I think there's going to be a lot of classification vendors, there's classify, there's Mobian. Now we've been scraping and analyzing content like this and I think you're going to have sort of a new wave of third party verification vendors possibly bundled in with brand safety that verify this contextual signal.

[B]

Yeah, I mean Mobian and other AI driven verification contextual totally makes sense because you have text on the page. Well, how does that affect the creative? Is it like the creative says, oh, you're interested in, you know, the sports scores, why don't you buy a Toyota? You know, what, what are we, what are we doing here?

[C]

Yeah, like that's the sort of broad version of it. It's this idea of generative creative optimization, let's call it Aidco gco. If we get a new TLA and it's the capacity to adapt a unique creative to every single URL. Now that's what open ads does, that's our specialty. It's this combination of contextual targeting with contextual creative that it's quite difficult to do that in the RTP pipes. And that's why we had to, why we realized we had to go out and build our own DSP to actually transact on these signals and really combine targeting and ad serving. Know the media, buying with the creative for the first time.

[B]

Build your own dsp. Oh my God, did you use Beeswax at least?

[C]

We did not. So I think you will never.

[B]

Come on, man.

[C]

Architecture of this year where you said, hey, like can you vibe code a dsp? And the answer is, well Ari, you can and maybe we can. And the few people who already know what a DSP needs to do, but it's really the coding models that came out really in the last four or five months that have allowed a small team like open ads to actually build a dsp, an AI native DSP from scratch. And there's some corners that we've been able to avoid the complexity around identity since we're going after context first. But yeah, these coding models have Improved the efficiency of every engineer on our team by multiples.

[B]

Yeah, I mean, I think we'll have achieved AGI when you can vibe code, like the logic for tracking clicks. You know, that's kind of the, that's the hardest thing in ad tech. But let's. What's an example of like, or have you seen when your AI is creating these ads, like, have you seen standouts, like where you're like, oh, wow, that ad is so much better because the AI created it than a human.

[C]

Yeah. So it's, it's really, I think ad tech can advance the frontier in AI here around artificial creativity. Because the problem with, you know, AIs just by themselves is it's like telling your creative team, you know, go lock yourself in a room and give me a thousand good ads. And they're not going to have any inspiration. They're going to give you the same ad every time. When you expose these models to all of the variety of the web, it acts like human inspiration. They're able to riff and improvise off of all this different content. And we've seen some really surprising and unintuitive ad generations. All brand suitable, all brand safe.

[B]

Yeah, like what? Like, did you have any examples?

[C]

Yeah, so let's take a premium beauty brand that has been having trouble connecting to Gen Z because it's felt almost too aspirational and. No, we identified context on the web. No, here's like 20 bougie butt cheap things to buy to improve your get ready routine.

[B]

You just say bougie butt cheeks. What did you just say?

[C]

Bougie butt cheeks, but chic.

[B]

C H I C Chic.

[D]

What is going on?

[B]

Bougie butt chic. Okay. Not bougie butt cheeks. Got it. Sorry, my bad. We might. This might require a edit of the podcast. We'll see if this stays in. Okay, tell me about the bougie butt cheeks.

[C]

So bougie butt cheap and a premium beauty brand. And it generated this ad with a picture of this luxury lipstick on a very realistic, crowded, kind of cluttered bathroom sink, saying brand name the only essential. And this was a framing that it absolutely worked. It's relatable and it solves the brand's problem of making it feel accessible and real to their target audience. And that's the sort of thing where launching that as a static creative and trafficking around the Internet, it could work, but it wouldn't hit nearly as hard as it would given that URL specific context.

[D]

And then what happened? So I'm presuming you it was generated. The brand loved it. You went live. What was the performance of this like kind of fully AI inspired and generated ad relative to, you know, what humans were created.

[C]

So that was a pre flight test where we basically take everything we know about a brand and basically back test on. All right, here's all the bitstream data we're getting. Here are the websites we would have chosen. Here are the ads we would have generated. What we've been doing with shooting, running, not ads.

[D]

Okay, yeah, that's the thing that I'm super curious about.

[B]

Let's talk about AI and some of the standards that have come out.

[C]

We.

[B]

I had the idea or the inspiration to bring you on because of some Twitter conversations you and I had. So what's your take on adcp?

[C]

So when we started open ads, our open ads, we had this idea that eventually every brand and agency is going to have an AI that knows their product and knows their goals and every publisher is going to know their context. And eventually you're going to need all of these entities to be able to talk to each other. So we've actually set out and started building a lot of that internally. And ADCP is really laying out the spec for the rest of the ecosystem. So the more that publishers understand these are the signals that buyers care about, the more that buyers understand these are the signals that are available for us to transact on. That's going to really the reason this matters is closing the performance gap between open web and the walled gardens. And what ADCP does right now is really scale and automate programmatic direct deals where it cuts down on a lot of the inefficiency of programmatic. Now you no longer need a million queries per second to actually pick out your maybe correct ID match in the haystack. It's a much more efficient way of scaling what used to be just human judgment.

[B]

Right, so which, so ADCP has a bunch in it. Are you currently developing support for it and which part is the highest priority for your company?

[C]

Yeah, so as a startup, we have to be very focused on what actually customers want right now. And where we've seen pull from the market has been from the agencies, from the brands themselves, specifically around creative. There's a lot of demand for efficiency and cost savings in creative generation. And there's a lot of context, you know, locked up inside the brands and agencies. So that's the part where I think a lot of missteps with AI startups and ad tech startups have been around building publisher monetization tools that aren't compelling to advertisers out the gate. When really the center of gravity for any innovation, I think it really starts from the buy side. So responding to what the brands want, what the agencies want right now has been that we've put out our early versions of AT CP compliant creative agents.

[B]

Right. So in that case, how would that work? So the buyer, the brand would be using AI to tell you what sort of elements to put in the creative. Is that kind of the idea?

[C]

Yeah. So we might get information from a brand saying here is our campaign idea, here's the sort of audience we want. Here is just a natural, plain language description of the kinds of creatives that we would want to see and we can go out and we're quite good at creative generation. We've been doing that live. And so it's easy for us to turn this around to the buyer side and say, yes, here are your creatives. We have your 302 50s, 300 by 600s, et cetera. Oh, you need the creative for a B2B version of your product rather than B2C. Easy for us.

[B]

And that's like separate from your DSP and ad network. You're offering like a creative generation service in that case.

[C]

Yeah. What we've found is now this is probably obvious to some, but wasn't to us is that inside a large agency or Holdco Media, buying and creative occupy separate silos. And one of the best ways over decades of advertising has been give away the creative and get them on board as media spenders. And that's really our strategy as a DSP.

[B]

Got it. Okay, I want to talk about UCP. So UCP is the proposed standard that LiveRamp donated to the IEP be recently. And it's pretty hard to get your head around. But effectively is the, the you take a bunch of information, you know about the user, their demographics, maybe their interests, etc. And then you create what are called embeddings, which are like long strings of letters and numbers that represent that user's profile, but in a very privacy protected way. There are a lot of these mathematical advantages to doing it, but there's also, and I wrote it up in last week's newsletter, sort of my description of how I understood it as sort of a moderately technical person, like maybe on the, on the waning side of technical as I get older. But, but what's interesting here is that you and actually some other people are pretty knowledgeable, have, have just privately said to me like this is there are a lot of really serious technical problems to making this work. Can you talk to me about that a little bit, what your initial criticism would be?

[C]

GP so first of all, this actually is important. Embeddings are a very powerful technology for understanding context for matching ads and audiences. It's a technology that the walled gardens use extensively. So the POS side here, adapting standards around embedding, interoperability in open web is a big thing we can do as an industry to close the performance gap between open web and walled gardens. Now, the challenge with embeddings is they're powerful, but the embeddings generated by one model are not at all interpretable by another model. So if we use even OpenAI embeddings, version 1, version 2 isn't going to understand it at all. And so as Speaking from a DSP's perspective, if we can get trustworthy embedding signals from a publisher, say about their logged in audience or about their content that's valuable for us to consume. The challenge here is, well, who gets to decide which embeddings models we're using? And now, are they open source or do we end up having to invoke them from a branded provider here?

[B]

Right, right. It's just a string of letters and numbers. And if you have two different code bases generating those letters and numbers, they're just nonsense to each other. You would need either everyone in the entire industry to adopt the same model and then upgrade at the same cycle, or you would need kind of the standard to pass along the version and have a fairly limited number of versions available, which are really pretty enormous practical problems giving the speed of model development right now.

[C]

Right. So what I can see working pretty effectively is let's talk about two categories of embeddings. There's semantic contextual meaning of content, and then there's embeddings about the audience for contextual embeddings. This is something where there's a lot of open source, very good models out there available and they were built to do this sort of semantic contextual understanding. I think it makes a lot of sense for the industry to pick one of those and you can use these for 6 months, 12 months. It's a lot better than nothing. And I think that sort of rotating standard would work, right?

[B]

Yeah.

[C]

These open source models, or even the ones coming out of labs like Google and OpenAI, well, they weren't trained on our audience signals, they weren't trained on our contextual signals or performance signals. And that's where the power and complexity comes from. Industry specific embeddings models. So you could have a data provider, say Liveramp, who's donated the spec. They could also donate an embeddings model and if it was open source, open weights, you can share the training data around the industry. Anyone can run it inside their own hardware, inside their own containers. Perhaps that would work a lot better than going down the road of a proprietary embeddings model.

[B]

Yeah, perfect explanation. I appreciate that. And that's what some other folks have said to me, which you know, in my entrepreneurial brain I say, well that sounds like an opportunity, you know, if you it's actually a better opportunity for entrepreneur is like to have a locked in embedding model than to have an open source one. But someone would actually have to do the work and it sounds very complicated. So if you're working on that, let me know and we'll have you on the pod. All right, Stephen, let's call it there. I really appreciate you coming in and giving us the latest on AI super interesting open ads soon to be renamed or something like that. We'll see. We'll keep an eye on the trademark situation.

[C]

We have not acquired the Trade Desk in some confusion.

[B]

That's fine. It's not like the Trade Desk is a very unique brand name either. They probably stepped on some people's toes when they named the company that. So it's all good, it's all fair.

[C]

Great to be here.

[B]

Thanks for joining us.

[A]

By now you've probably heard about Market Live. We put on two sold out events jam packed with the most insightful advertising content around. Speakers included Eric Seoufer, James Borrow from Universal Ads, Mark Grether from PayPal, Olivia Corey from Houzz, and of course me and the Market Extra crew. Well, this spring we're coming back bigger and better with a two day that's shaping up to be a must attend event. March 10th and 11th in New York. We're putting on the new Tentpole event in collaboration with Adweek and TV Red.

[B]

And you absolutely need to be there.

[A]

Early bird tickets are 25% off and qualified brands and agencies can be comped. Go to marketlive.com right now. That's marketlive.com to get the early bird discount.

[D]

All right, we're back with the refresh on this week's docket. We have some M and A, some investments, some regulatory stuff, some presentations, something for everybody. But before we begin, shout out to you, Ari, you join the board of VidMob alongside longtime exec and leader Molly Spillman and friend of the pod. Friend of ours, fellow Staten island boy Mark Manito was promoted to CEO. A lot of stuff here. So first of all, tell us about VidMob. Tell us why you joined the board and the whole story of how this came together.

[B]

Yeah, the short version is that VidMob is, was originally known for something totally different. They did like crowdsourced and freelance finding for creating ads. They pivoted a couple years back to be video analytics because they had built this tech that understands video ads at a really deep level. Company's doing really well and I'm always, I've always in my career been super interested in creative. That's where I got my start. So seeing the tool and understanding what it could do and the opportunity in the age of AI, it's sort of like this company was chugging along with its, with its product around video understanding. And then generative AI comes up. It's the huge opportunity because basically you not only have this intelligence you can give the AIs to generate better ads, but also when the AI creates hundreds of variations of your, of your commercials, it can now understand which ones are compliant, which ones are not, which ones don't show the brand for enough seconds, which, which models are more, you know, human models are more appealing to the consumer. That's what this tech does. So I was pretty excited to join the and, and especially because Mark and I go back, you know, over 15 years of working together.

[D]

Yeah, that is awesome. We'll talk about creative a little bit later on. We kind of dive into, to the new AI Lumascape. But yeah, it's pretty fascinating to, to see all of the momentum around creative in the age of AI. And then to your point, and you always had this conversation a bunch of times, these standalone creative generative companies, you know, like they're a dime a dozen. But when you start thinking what you can do, when you connect creative capabilities, creative generation capabilities to other things, whether that's measurement, analytics, you know, just like straight up activation becomes really, really exciting. So this is, this is, I think square in that, in that, you know, kind of way of thinking.

[B]

Yeah, I think the ability to naively create 1530 second commercials, we all saw that with the original video models maybe 18 months ago. Right. And everyone said the same thing, which is like this isn't really ready for real ads, but, but it's coming and it's shaking out where it is, starting to get ready for real ads. It's not easy. And there are parts of the puzzle that still need to snap into place around brand safety, analytics Things like that. And I think this company is well positioned. But there are other innovative companies doing really cool stuff here. There's no doubt that the majority of ads are going to be AI created at some point a couple years from now, maybe sooner, maybe. So the infra for making that possible is a pretty interesting space to be.

[D]

Absolutely. All right, well, congrats, man.

[B]

Thanks.

[D]

All right, let's talk M and A. Adobe to buy SEMrush for 1.9 billion. Did you see this one coming?

[B]

I did not. I didn't even know Semrush was a public company. I thought they were way smaller than they were. I think we had them on architecture once, but I wasn't the interviewer, so I wasn't that familiar with them. Could you give me your take on this? I'm like having a hard time. I'm scratching my head on this one.

[D]

Yeah, you know, Yes, I have a take. Right. So Adobe has been somewhat quiet in the, like, marketing and advertising space, you know, just in like ad tech and in our world. But I mean, everybody is a customer of Adobe. Every enterprise is a customer of Adobe. So whether it's for creative, whether it's for their site experience, they're deeply embedded in there. So the thinking around Semrush is can you get deeper within the site and what they call the kind of like customer experience and start to develop some capabilities in this age of not just SEO, but. But CEO, Right. Generative engine optimization. Answer Engine optimization. And that's the bet here. The real question is how much of Semrush today is powering Geo versus are the elements that power SEO going to be the elements that power geo? And does Adobe have like a giant leap forward by having this asset?

[B]

Yeah, I think these are all good questions. I mean, a big part of this move from SEO Geo is going to be the customer relations. And so you're naturally going to call your SEO provider to do the geo. And so SEMrush is, I think, the leader in the space. I think if after 20 years of the web, this is the first substantial exit for an SEO company. SEO is a category, is generally considered to be a bad investment category. So this is the first and only real significant exit.

[D]

Yeah, but it's called Semrush. It's not called SEO Rush.

[B]

Yeah, I know, right?

[D]

It's, it's, you know, I think, you know, the revenue is, is not just SEO, but it's also, you know, just like the, you know, they do a.

[B]

Bunch of buying is. That's a big Part of the business.

[D]

Yeah, that's my, my impression is. Yeah, yeah.

[B]

Well, buying on search has also been historically a really bad business. You know, both, you know, Kenshu renamed to sky and now they're almost all retail media, not search. Marin went out of business. There's an. Actually. Okay, this is a little bit of a side quest here. Do you know what's going on with Marin Software?

[C]

Not at all.

[B]

Marin went bankrupt, went out of business.

[C]

I remember that.

[B]

And they've been acquired by a Vibe coding studio that is attempting to rebuild the company entirely using Vibe coding as an experiment. They're like have these young engineers in training who are part of their program who pay for training and they're systematically rebuilding Marin like a Frankenstein's monster from dead parts and trying to make a go of it as a new ad tech company built entirely 100% vibe coded. And I've been following it. The, the guy who does it, he's very vocal on Twitter. His name is Austin. Austin. Hold on one second. Austin Allred. Austin Allred. And his.

[D]

Yeah, yeah, yeah.

[B]

He, he, he runs a, he's, he's run several of these like trained boot camp things. I actually, if anyone knows him, I asked him to be on the, on the POD because I'm so fascinated by this, but he hasn't responded. So if anyone knows Austin, give him a shout out that he's being talked about on pod and we'd love to hear about this Marin software revival effort.

[D]

Yeah, just add him on the socials. Super cool. As a related aside, Business Insider is doing a five part series on geo. So like, if you had any questions around. Is this a category? Is this a high category? Is this a category that, you know, people are trying to figure out like a five part series on GEO that they just dropped this week.

[B]

That's, that's, it's sort of a little bit of a proving the point. Like the more text they could write about geo, the more higher they're going to be ranked about their understanding of geo, which is going to produce more traffic for geo. Why is it called. Can we just call something. I hate the GEO so much. I like aeo, although it's a little hard to say. Or aio, I liked aio, but that didn't catch on at all.

[D]

Yeah, for us, GEO becomes, it becomes difficult because when we think of geo, we think about something completely different.

[B]

Completely different.

[D]

Forevermore think about it completely different. But it's, I can. The volume of GEO startups. It's interesting. Like there was the wave of The Monster series A a few months ago and then the volume of raw startups like just pinging me saying we're a new Geo entrant. It's extremely high. So it's interesting. It's a space that the barriers to entry are quite low.

[B]

Yeah.

[D]

Just to be able to spin up something and provide some insights. So it's going to be a head scratch of a category but without doubt this needs to exist.

[B]

Yeah. I'm invested in this category in a company called Gumshoe. I'm a believer in analytics, not a believer in optimization. A lot of the optimization is is is advice like you know, post about your products on Reddit and I I think that's pretty non sustainable.

[D]

Yeah, yeah. Shout out to Todd Sawicki from Gumshoe Architecture.

[B]

One presenter yes, exactly.

[D]

And then the other M and a rumor is WPP is apparently a potential takeover target. Allegedly a potential takeover target. And the names that have been floated were Havas as well as some of the usual suspects in pe. We'll link to it in the newsletter last week. But the thing that stood out was the market cap has gone from 21 billion to 3 billion, which is wild crazy and it's gone down significantly even with Cindy Rose being the new CEO, although she's just brand new in the seat. Be hard for her to turn things around in this short a time. Consumer I never thought I'd see the day that you'd see something like this where Havas would allegedly potentially even though they said they're not be looking at acquiring wpp.

[B]

Yeah, I don't really have any intelligent opinion about this except that having however many five majors plus a couple minors, it is ripe for consolidation. There's no real reason these companies shouldn't combine in some way and obviously the cost savings would be a big part of that.

[D]

Yeah, some public markets investor once, in asking me about the future of the demand side space once observed that almost every category has a max of three.

[B]

There's a very important Harvard I think it was a Harvard Business School study from the 1960s called like the Rule of Two or Three that almost all industries consolidate to either two or three depending on some exogenous factors. And that seems to be seems to be in play here. Avaas also was aggressive. They just did a deal with Horizon a couple of weeks ago it was announced that they I think if I'm remembering this correctly, that Avast is kind of becoming the international arm of Horizon, which is Horizon's a big indie run by Bob Lord.

[D]

Yeah, yeah, yeah. That makes a lot of sense. All right, we'll pay attention to this one. Couple of investments were announced. So Agentio raised a big Series V. Series B. Sorry, 40, $40 million Series B. Are you familiar with Agentio?

[B]

Sort of. I've heard of them and I read this article, but I'm not. It's not in my radar very much.

[D]

Yeah, yeah. It reminds me of that company, Influential, that was acquired by Publicist, which is basically like a platform to discover and execute buys with creators for brands. It's a big Series B. Sounds like there's a strong AI angle here. And I think the reason why. It's interesting and perhaps because it was perhaps the reason why it was picked up by ad exchangers, they're positioning it as the TTD of creators, kind of making sense of a fragmented space and having a system of record for buying.

[B]

Yeah, I think the creator market is fully mainstream at this point. You know, it's kind of gone through its evolution. It was early on it was sort of this weird thing you would do, sort of an adjunct to PR or product placement. But with the declining ability to reach young consumers through traditional media, it's gone front and center. There are platforms like Influential were largely matching platforms that had a lot of humans involved. You had to negotiate the deals and had to, like, you know, be careful about who you picked and stuff like that. And from my limited conversations with folks in the industry, they've. There's a lot of concern on the brand side that it's not. Not very automated, not very safe. They're not sure they're getting their money's worth. So there's a lot of opportunity here. Whether this platform is able to capitalize or not, I think it's a pretty important area for marketers to get better tooling.

[D]

Yeah, agreed.

[C]

It's this.

[D]

It's a lot of the same things that were areas of problems for, like, early digital ads and an entire ecosystem called ad tech propped up to.

[C]

To.

[D]

To solve it.

[B]

So I wonder if we'll see.

[D]

Yeah, yeah, we'll see if there's, you know, kind of people from ad tech moving into creator. You know, Rembrandt kind of started there and then expanded out. But it's an area where I'm trying to pay more attention because there's just so much growth, so much spent.

[B]

It's very similar to podcasting, where if you have host red ads. So podcasting has two different modes. There's an ad network model, but there's. The more valuable part is the host read stuff and the host read stuff is largely an influential game. You want that host, that Joe Rogan read, and you want to give him the script, and you want to approve the script. And the pricing is very arbitrary. That's really a very good analogy. And there are companies in podcasting like Acast, that do that sort of thing. Not big businesses, though. They've struggled with scale.

[D]

Yeah, I spoke to Acast last week. Greg Glenday, who's the former CRO from Undertone, is the CEO there. And it's actually total sidebar. We should have them on architecture because there's so many interesting things about podcasting in terms of the concentration of large shows and the ads opportunity. It's very interesting. But. But it's a small space. We've talked about this for some time. Right. It's like a $2 billion space generally. And it sounds like they're. They're. They're the leader.

[B]

So we. We did have them on Mark. Not on this podcast, but on the Monday. The Monday. We had them on the Monday.

[C]

Yeah, of course.

[B]

Yeah, Yeah.

[D]

I meant the weekly where we can.

[B]

We can dive in.

[C]

Okay.

[D]

Two other deals. Curved, Raised a Series B. Curve does like Brie does, add formats and measurement in ctv. They did not disclose the size of the Series B. Are you familiar with Curve?

[B]

Yeah, yeah, pretty familiar. You know, custom, creative, custom targeting in ctv. I think they do a lot of stuff around live and shopping, but I'm sort of beyond that. Interactive.

[D]

Interactive format.

[C]

Yeah. Yeah.

[D]

And then. And then another new custom algo company is born. Biddable Assets, founded by John Donahue. Raised a seed round with investment from TTD or TD7.

[B]

I like the name Biddable Assets. It's got some. Got some moxie I like.

[C]

Does. It does.

[D]

John's a cool guy too, and he knows the space quite well. Love to see it.

[B]

Yep. And the trade desk had previously invested in Chalice, which is a direct competitor. I think the trade desk's M and A. I'm sorry. Investment strategy is like very corporate. They only want to invest in things that are really tight to their objectives.

[D]

Yeah. Having spent a bunch of time with them, they're strategic investors, so it makes a bunch of sense. All right, regulatory two things here. Meta wins an antitrust suit regarding acquisitions of Instagram and WhatsApp. So they are, unlike Google, not declared a monopolist. Talk about this one.

[B]

Yeah. This is a tortured case. It's got started in the Trump administration. It got thrown out of court for. By the judge. And then I believe under Lina Khan, under Biden it was re submitted and the case largely was about the acquisitions of Instagram and WhatsApp. And there were various emails from like Zuck and other people saying like, you know, things that courts don't like to hear about dominating the markets and stuff like that. But the fact is this was really old. The judge was pretty convinced that TikTok was part of the competitive set and that it's a very competitive market. And you know, the, the FTC's argument that there was kind of a separate market for communicating with your friends versus, you know, social was rejected. So this was a win across the board by Meta. It was not unexpected. Most court watchers felt that that was going to be the case. And now the real question is, does this reopen the floodgates for acquisitions from Big Tech? The they were largely on hold during the Biden administration. This case kind of gives Matt a little bit of a green flag to move ahead, although probably not like a big one, like a Snapchat or something like that. More of a, you know, more tuck ins. I would guess so. So I think this was expected. I think for tech people, this is welcome, that this, this was not an antitrust suit that a lot of tech people supported because it was really M and A oriented. It wasn't about abuse of the marketplace.

[D]

Two questions for you. So do you agree with the judge's finding?

[B]

I do. I think that at the time these acquisitions happened, if the FTC had wanted to investigate and block it, that might have made sense because they do have the rights to do that. But to wait five years or more after they had been acquired and then have a suit, it seemed like a perversion of their authority and, and a effectively just a way to strike against Big tech without any really rationale behind it.

[D]

That makes sense. And then the TikTok argument resonated with me, which is like, hey, you know, it's like he's not dominating, right? Or Meta is not dominating when you have TikTok, you know, just, you know, creeping up at their heels every day. And then on the other side, do you believe that this does sort of give the green light, so to speak, for Big M and A by Big Tech?

[B]

Well, the red light was definitely on during the Biden administration. Like it was very clear that nothing was going to get through. And under Trump, it's been more permissive, but with the overhang of sort of crony capitalism and the requirement that you make someone in Trump's orbit rich to do a deal. So I'm not sure that this case Changes that much. Maybe it does in a future administration as a precedent.

[D]

Got it. Got to make sense. And then on another thing on the regulatory front, so the EU is pulling back on privacy regulations, including the GDPR of making it super difficult and threatening of fines or actual fines and all sorts of consent pop ups. What's going to happen?

[B]

Yeah, it's not official. It's been mostly leaked documents and conversations, but they're definitely looking at some revisions to both the GDPR and the E Privacy Directive. And the E Privacy is why you have the cookie banners. And some pretty reasonable stuff like the reason you have cookie banners is because under E Privacy, just the setting of a cookie, an anonymous cookie, is considered something that requires consent and that doesn't really make sense to anybody. It's the identification of the user that really is a privacy issue. There's also a bunch of stuff in GDPR about sort of loosening the consent, like enabling you to identify a user anonymously as part of legitimate interest instead of requiring consent. I think for most business people and Americans this sounds really great, these changes. The European privacy folks are freaking out. They're not happy at all about it. And there's some uncertainty about how the actual enforcement will work because in Europe the enforcement is at the country level and those country regulators may not be copacetic with these changes. I think the best coverage of this so far has been in our newsletter, the monopoly report that came out this past Wednesday. Alan Chappelle did a really good job of breaking it down. So I would recommend reading that. And that'll be in the notes too.

[D]

Yeah, absolutely. All right, cool. Two big presentations dropped. Let's talk about them. So the first was the, I guess long awaited. Said they would never do it.

[C]

Did it?

[D]

AI Lumascape.

[B]

This is new grounds for the podcast. We've never had audio visual before. All right, so I can see it. The. Once again, podcast listeners cannot. But you'll have to trust us on that.

[C]

This. Yeah.

[D]

So we've got the AI Lumascape here. Let's like talk about it. Let's go through it. You know, again, maybe this will encourage some folks to, to pull up the. The YouTube. Here's a few observations for me to start. Number one, credit to Luma to try to make sense of a very fast moving space with a lot of moving parts. But also it, it, it is, it's kind of fitting things into boxes that arguably kind of like already exist.

[C]

Is thought number one.

[D]

Thought number two is the display lumascape. Basically all the other lumascapes are filled with these red boxes around logos, meaning acquired companies. There's so few acquired companies on here. I think it's very interesting. And then number three, up to this whole kind of creative topic on the upper left, you have the customer experience, basically the AI generation tools. There's so many creative generations. There's so many companies that they have a little line at the bottom because they ran out of room, quote, unquote. Many more companies, you know, a nod to this space, you know, being so, so large. What do you see when. When you look at this?

[B]

So there are pretty big, somewhat vague categories. There's content creation tools, there's creative intelligence, there is experience intelligence. I don't know what that means.

[D]

Right. Which is Adobe.

[B]

Yeah, Adobe is there. So is Braze. These are pretty big companies. I think it's pretty challenging to say who's AI and who is using AI in their products. Everyone's using AI. If you're not using AI in your products and you're selling like a SaaS product at this point, what are you doing? Like, you know, it doesn't make any sense that your competitors are certainly doing it.

[C]

It.

[B]

So this is sort of like a reshuffle of the other lumascapes more than something that's telling me very much here. I would kind of prefer. I'd love to see a take of this, which was Pure Play. How many. I'd love to see the pure play AIs that didn't exist before LLMs came around and what they're doing, you know, having, you know, Index Exchange. No offense to Andrew. Love Andrew, but like, why is it helpful for me to know the Index Exchange is in the AI Media category?

[D]

Yeah, yeah, I think that's. I think that's right. And the teaser that they put together prior to publishing this, which was, you know, basically three categories. AI First, AI Adapted and posers. I think actually might be the way to do this, which is just have a clean lumascape of AI first companies, maybe with the qualification of you need to have, you know, started the business or like fully pivoted into the business beginning in 2023, right when you saw the LLM Wave kickoff, maybe you raise some funding level such that you're actually a company. And I think that might be interesting to maybe request for another one.

[B]

The main use of these Lumascapes is that you print them out and then your BD guys start circling which ones you've had conversations with and which ones you haven't. And then you go, huh, I Never heard of those guys. You pull them up on the webpage, then you hit the. Hit the contact us button, which is pretty useful. But like, so. So they serve their purpose. And I've personally been guilty of doing that in the past.

[D]

Yeah. They also help to point to the fact that the market is crowded.

[C]

Yeah.

[D]

So, you know, kind of like if you are in the investment banking business, I think you can kind of use it to show that consolidation is needed. And then for companies, you know, like, the one thing that these things have done a good job with, and again, I give, I give Terran team a lot of credit, is, you know, they help to organize a space and help buyers of technology just like figure out, like where you fit in and who the competitors are. So if you agree with the, you know, kind of sub box that you are in, you can say like, hey, this is what we do. Here's how you can think about us. Here's where we appear on the. On the AI Lumiscape. But I would imagine that companies like Scope 3 and Swivel, both of which we're invested in, both of which are kind of deeply in trying to change everything when it comes to Gentec buying and selling, might not like the idea that they're in optimization or adopts respectively.

[B]

Yeah. At some point the logos get so small, we start hitting the laws of physics. Like you can't have pixels that are even smaller than these. So I think we should probably wrap up on that. I think that's kind of the main thing we wanted to talk about. So thank you everyone for listening. Thank you, Steve Liss for joining us and also a reminder to send in emails to both Eric and I for our Thanksgiving Mailbox Mailbag episode. No question. Too dumb. Too smart. You can hit us by email, by LinkedIn by DM. There's a lot of ways to reach us and we're looking, looking for the juicy question. So please reach out.

[D]

Absolutely. Look forward to it. See you next week, everybody.

[B]

See ya.

[D]

Thank you for subscribing to marketecture.

[B]

New interviews are added every week at.

[D]

Marketing and your favorite podcasting app.

[B]

Thank you for listening to the Market podcast. New episodes come out every Friday and an insightful vendor interview is published each Monday. You can subscribe to our library of hundreds of executive interviews at Markitecture tv. You can also sign up for free for our weekly newsletter with my original strategic insights on the week's news at News Market tv. And if you're feeling social, we operate a vibrant Slack community that you can apply to [e-mail].