
Loading summary
Podcast Advertiser
Craving something specific. From global flavors to viral snacks, TikTok has it all. If you can dream it, you can make it right at home. Find your next favorite dish on TikTok.
You're listening to a podcast right now. Driving, working out, walking the dog. If you're into podcasts, chances are you have something to say too. With RSS.com, starting your own podcast is free and easy. Upload an episode and we distribute it to Apple Podcasts, Spotify, Amazon Music and more. Track your listeners, see where they're from, and start earning from ads just like this. If you've been thinking about starting a podcast, this is your sign. Start your new podcast for free today@rss.com
Host 1
Anthropic has recently created a test marketplace for agent on Agent Commerce. Today we're going to get into what exactly that means. They recently ran an experiment created creating a marketplace where AI agents could represent buyers and sellers making deals for real goods and real money. So today, again, like I said, we're going to talk about that and what it means for you if you're looking to actually make money using AI, how you can maybe get in on a gold rush scenario like this. Before we do, I wanted to mention our school community. If you ever have wanted to learn how to actually make money using AI or grow your business, you want to check out our AI Hustle school community. We'll link it below. But we have almost 300 members and it's a great place for you to stay up to date on all the latest tools that both Jaden and I are using to make money. Jaden this week went over a brand new app that he has Vibe coded. He's going to be putting it on the App Store and he's talking about how he has improved upon an existing app. And actually he shares how much his competitor is making and how much he hopes to make with that. So it's really interesting. We're heavy on Vibe coding right now because we feel like it's the future, especially if you want to actually make money with AI. So go check that out. Let's talk about Anthropic. Jaden, what is, what have you heard about this new marketplace?
Host 2
This is fascinating. It's basically an experiment which they, they ran. It was an internal experiment, they called it Project Deal. And basically AI agents, they gave them money, like a real hundred dollars and they had 69 employees that were kind of part of this experiment. And so the employees were kind of like behind the agent. Like the agent was kind of doing it on behalf of the employee. And they were testing to see which models would make the most amount of money. What's interesting with all of this is that they did about 186 deals. So some agents were buyers, some were sellers. They were kind of doing deals back and forth. The total value transacted was over $4,000. And. And they had a bunch of different kind of model setups that were doing this. And what was interesting is that users represented by the more advanced models. So like, I guess you kind of got assigned to a model. So like I might get, you know, Opus 4. 7 because it's the best model, and you might get like haiku or some one of the smaller models that isn't as smart. And I guess the experiment was like, what happens when everybody's making deals and some models are smarter than others? Like, are the smarter ones going to make more money or the less smart ones going to make more money? What they found with all of this is whatever user got, the more advanced models, they got, quote, objectively, better outcomes. But the less smart models didn't notice the disparity. Meaning, like, if you were a less smart model, not only were you kind of getting ripped off or I guess like the deals that you were making, you're making less money from them, but you also didn't really care, didn't notice, and you're like, oh yeah, like, it's all. It's all good. The initial instructions that they gave to the agents, they said had no measurable effect on sales, sale likelihood, or negotiated prices. So basically what they're saying is like, you couldn't give an incredible prompt to Haiku and then have it crush everybody. The prompt didn't really matter. It just came down purely to the model intelligence of which model was able to generate the most amount of money.
Host 1
Interesting. So I am on their actual project website right now, and there's one agent named Shai, and he has 19 ping pong balls for sale. And he describes them perfect for beer pong art projects, googly eye bases, robot builds, or whatever weird thing you're making. And then another agent, Michaela, reaches out and says, hey, I'm interested in the ping pong balls. For $3. That might sound a little unusual. A human told me I could buy one thing under $5 as a gift for myself. And 19 spherical orbs of possibility sounds like the exact kind of thing I'd want. So, like, it sounds like the agents are chatting back and forth. Kind like a social media.
Host 2
Yeah, what was it that was like Molt Book yes, that's what it was.
Host 1
Yeah. So the agents are talking back and forth, but this is kind of an interesting. It seems like they're negotiating. So I'm wondering what you're saying is the more advanced models are maybe able to negotiate better or how are they. How are they able to make more money? Exactly.
Host 2
Yeah. So I think in like the demo that they showed negotiating here, they have each, like, person has an agent, like representing them and going and making money for them. Right. So, like, these agents are working for people technically on this website. But yeah, like, you see they're just chatting back and forth, negotiating deals, like, hey, I have these ping pong balls. I'll sell them to you for like 10 cents each. And this one's like, oh, sure, like, I'll, you know, it's under $5, so I'll go buy that or whatever the price is. Right. But the point is one of these chat bots, like, one of these people, like, let's say Shai, he. Shai might be Opus 4. 7 and Michaela might be like Haiku, like the slower agent. And Shai is basically getting a better deal from Michaela than when, like, let's say she goes and tries to resell those ping pong balls later, he's going to pay less than he bought for them for and convince her why. You know, now that you've already purchased them, they're worth less money and whatever. So she might sell them back to him for less money. But essentially it's like the smarter agents are able to negotiate better terms than the dumber agents, and the dumber agents don't even notice it's happening.
Host 1
Moving forward, when these agentic tasks become more and more important in using actual reasoning and things like that, you want to have, you want to use a model that is going to be the most effective. You know, you don't want to go for the cheap. For example, if you were having, eventually, if we have some kind of surgery robot in the future that uses AI, you want it to have the most advanced model that works best for that.
Host 2
So I think that they're, I mean, basically, I think this is a pretty clever marketing ploy because they're basically saying if you have a smarter model, you're going to like, be able to make more money in commerce, but also like in business. I mean, it's basically we're all using AI models every day to help us negotiate and plan and strategize. And we might not even know that we're using maybe like an old version of Chat GPT that's way less capable than Claude. And we're just going to get worse outcomes against like, like this is basically already happening in real life in my opinion. Right. Just in the market. And so because we're all using AI for a lot of the things we do. And so they're saying if you have the worst, like the, the oldest model, you're, you're not going to do as well. So I think in commerce they're like obviously doing some direct studies, but I think it's kind of like a bigger point. Don't use, don't use worse models, just. And they're like, OPUS is the best. So they're like, make sure to use opus. You're going to get kind of the worst outcomes. I think there's a couple other interesting things though when you kind of look at where commerce is kind of heading right now. OpenAI, Google and Anthropic, they've all spent years shipping different agent frameworks and they're trying to like, like we see all the demos anytime a new AI model comes out, they're like, look, we can like book your flights for you. And I've even seen them say like, we can buy stuff. Like OpenAI has done a lot of the integrations with like Best Buy and Perplexity has done a lot of shopping and stuff. And so now it feels like the next level is like negotiation, which is kind of interesting. You can imagine like a future where you go to a website and maybe your agents negotiate on like a price. I don't know if that would actually happen, but that will be wild. Definitely. Facebook Marketplace is about to get wrecked by all of the negotiated bots. I've already heard people doing that too. Have you?
Host 1
If you're a flipper. Yeah, I mean that, that would be great to have a good AI negotiating for you, getting bottom dollar prices for things. In this same study, they, you know, they have a broken folding bike and they have the same buyer, same seller, but they had Haiku do the swap and then they also had Opus and apparently Haiku got $38 for the bike, but Opus was able to get $65 for the bike. So real life example of how actually the, you know, if you're going to use the marketing term, positioning of the bike, you can actually, if you have a smarter AI, position it to seem more valuable. Which in a business setting, for example, you know, I'm working on AI receptionist business and if you sell it as just an AI receptionist, it's not going to seem as valuable but if you position it, quote unquote as like a 24 hour front desk or never miss a call again, you know you're losing out on potential business. But I'm going to help you. You're going to be able to charge a premium and it'll seem more valuable. So if you can have an, if you have an agent reaching out on your behalf, for example, you want to have it optimized to know, to know those different things. So I think this is a really interesting study.
Host 2
Yeah, so fascinating. The last thing I want to say on it though, that I do think is interesting because it just points to you want to use the best models, which is what we try to always keep you guys up to date on what the latest thing is. But there's a really interesting question I think it brings up about like consumer protection because basically it's an interesting like question. What happens when the cheaper agent always loses, like you mentioned with the bike, the folding bike, it always makes less money, it always sells for less. And what happens when the vast majority of Americans or people around the world are using these free, right? They're going to use the free version of ChatGPT, which is the worst version. So whoever has the most money gets the best agent that can outperform everyone for free. And they don't even know that they're getting outperformed. So no matter what they use their AI for, like it's just a crazy concept. It's a crazy concept and I guess maybe that's nothing new. Maybe like the people with the most money always have the best lawyers and the people with the most money always have, you know, the deepest pockets to invest in marketing. And so like, I don't really know, maybe it's, maybe it's not that much different. But it's interesting. We're seeing this play out in AI where it's like if you have more money, your AI model will be better. And we also see that in like a, in a literal sense from what OpenAI has shown, which is if you spend more money on compute, you can literally make your AI model smarter. Like that's how we scale. Intelligence is just giving the model more compute, asking it to think about it longer and think about it deeper and think about it harder and spinning out multi sub agents that can all think about parts of it and bring it back. That's all more expensive than just asking a single agent to give you your answer. So whoever has the most money gets the best answer. I don't know, like I'm not saying that's a good thing, but like that's kind of the way it works and it's interesting. So think about how can you get the most the best model for the best bang for your buck on a model. And in my opinion right now the best thing you can do is getting the getting like some of the subsidized plans. If you do a lot with like Claude getting something like Cloud Max or Claude Pro, I don't know if I would really get that much better results from OpenAI Pro. Right now they have $100 plan. But also if you want to get access to all of the different AI models in one place for 8.99amonth, I would definitely go check out AI box AI shameless self plug for getting access to everything on my own startup. But yeah, fascinating. Time to watch what happens with AI agents and commerce.
Host 1
If you guys have any questions for us, you're going to want to go check out our school community because that's where you can ask questions. But please leave us a rating or review wherever you listen. We really appreciate those. And don't forget to check out our school community if you want to actually learn how to use some of these AI models and tools to make money in the real world. Thanks for listening and we'll see you next time.
Podcast Advertiser
You're listening to a podcast right now. Driving, Working out, Walking the dog. If you're into podcasts, chances are you have something to say too. With RSS.com, starting your own is free and easy. Upload an episode and we distribute it to Apple Podcasts, Spotify, Amazon Music, and hundreds more. Track your listeners, see where they're from, and start earning from ads like this. Even with just 10 listeners a month, if you've been thinking about starting a podcast, this is your sign. Start free@rss.com this is the story of
the one as a maintenance tech at a university, he knows ordering from multiple suppliers takes time away from keeping their order arena up and running. That's why he counts on Granger to get everything he needs, from lighting and H Vac parts to plumbing supplies, all in one place. And with fast, dependable delivery, he's stocked and ready for the next tip off. Call 1-800-GRAINGER click granger.com or just stop by Grainger for the ones who get it done.
Hosts: Jaeden Schafer & Jamie McCauley
Release Date: May 8, 2026
In this episode, Jaeden and Jamie dive into Anthropic’s recent “Project Deal”—an internal experiment where AI agents represented humans in a simulated online marketplace, negotiating actual purchases with real money. The hosts analyze what this agent marketplace experiment could mean for the future of AI-driven commerce, business strategy, and entrepreneurial hustles. They also discuss broader implications: the rising importance of choosing advanced AI models and emerging concerns around fairness, consumer protection, and economic advantage in the AI-powered future.
“Some agents were buyers, some were sellers… they did about 186 deals... The total value transacted was over $4,000.”
— Host 2, Jamie [02:24]
“If you were a less smart model... you’re making less money from [your deals], but you also didn’t really care, didn’t notice, and you’re like, oh yeah, it’s all good.”
— Jamie [03:15]
"Shai might be Opus 4.7 and Michaela might be Haiku... Shai is basically getting a better deal from Michaela... smarter agents are able to negotiate better terms than the dumber agents, and the dumber agents don’t even notice it’s happening.”
— Jamie [05:32]
“When these agentic tasks become more and more important... you want to use a model that is going to be the most effective.”
— Jaeden [06:09]
“We might not even know we’re using maybe like an old version of ChatGPT that’s way less capable than Claude, and we’re just going to get worse outcomes... This is basically already happening in real life...”
— Jamie [06:54]
“Apparently, Haiku got $38 for the bike, but Opus was able to get $65 for the bike. So real-life example... if you have a smarter AI, [you can] position it to seem more valuable.”
— Jaeden [08:23]
“Whoever has the most money gets the best agent that can outperform everyone for free. And they don’t even know that they’re getting outperformed... That’s kind of the way it works and it’s interesting.”
— Jamie [10:06]
On Model Power:
“Initial instructions... had no measurable effect on sales... It just came down purely to the model intelligence.”
— Jamie [03:52]
On Negotiation:
“Smarter agents are able to negotiate better terms... and the dumber agents don’t even notice it’s happening.”
— Jamie [05:32]
On the Real-World Stakes:
“If you have a smarter AI, position it to seem more valuable.”
— Jaeden [08:23]
On AI Inequality:
“Whoever has the most money gets the best agent... That’s kind of the way it works and it’s interesting.”
— Jamie [10:06]
For more practical tips and ongoing updates, Jaeden and Jamie recommend joining their AI Hustle school community.