Loading summary
A
We have some massive news out of Anthropic and right when we thought it was getting, you know, it couldn't get any crazier. With AI and agents, Anthropic is now letting their users connect more apps directly to cloud. And so I view this as kind of the precursor for where we're going to see agents take off in the future, actually integrating your AI model with real tools in the, in the real world. I'll show you some exact breakdowns of how they're of products and the apps that you're currently, currently able to connect with Claude and what you're able to do with them, some demos and stuff. So I think this is very interesting. Before we get into that, I wanted to mention, for the first time in two and a half years, AI Box AI. My startup has officially launched our first product. It's called the AI Box Playground and it allows you to test out all of the top AI models in one place for $20 a month. So you don't need subscriptions to every single platform. You get them all on one platform. 11 labs for audio, all the best. Ideogram and, and OpenAI and Flux for image generation and all of the top text models as well. Reasoning and everything else is in there. So I'm really, really excited. If you go try it out and you have any ideas, we're actively developing and adding new features for there for it. So if you're like, man, it'd be super awesome if it could do X, Y and Z, send me a Message over on LinkedIn or Twitter. We're actively adding new features to it, so I'd be super excited to have you try it out. As one of our first beta users. There's a link in the description to AI Box. AI. You can, you can go check it out. All right, let's get into what Claude is doing. So it's pretty interesting. This was just recently launched and this is kind of expanding what they're calling their quote unquote, deep research. So their deep research capability that can do a lot more than just a regular chat bot, you know, search the web and do a bunch of different things, is now getting all of these new tools. So essentially these new integrations are going to be available. They're in beta right now, but they're going to be for their subscribers to their Claude Max and then also their teams and their enterprise plans. Soon they'll be adding them to the Pro. So I think that's. Most of us that are paying for something are going to get the Pro So that soon. We don't quite get it yet, but that's okay. I'll explain what it is and you'll. You'll be prepped for it. So they also, by the way, in other news, they have added and increased the rate limits for their AI powered coding tool, Claude Code, which is amazing because my company, AI Box, uses Claude code all the time, actually, literally has been a lifesaver for us and helps helped us really rewrite and fix up and do like hundreds of lines of code. So it's very, very exciting. So all of this is pretty much part of Anthropic trying to keep up with all the other players, right? We have Google, Gemini, we have OpenAI that are rolling out a lot of these really cool features. And so they're trying to keep up with them, Anthropic says. And I mean, I would definitely take this with a grain of salt, but they said that they're trying to get to $34 billion in revenue by 2027. I hate it when these companies have forward projections multiple years into the future. Right. It's. Why are we talking about getting to $34 billion by 2027? That's because currently they're at $1.4 billion. And that's of like early March this year. That was their annualized revenue, aka like if they, for the next 12 months after March, they were projected to make that much. So if they're pretty much what's happening is they feel like they're, they're probably falling behind. OpenAI so they're trying to project big numbers. Like OpenAI has to say, look, in two years, we want to hit this $30 billion revenue target pretty much. And if we do all these cool things and we have all the integrations and we're the number one company, it's totally possible. So I'm not throwing any shade at them. I do love Anthropic. They're doing a lot of innovative things. I just, you know, the, the double did the two years in the future projections. I just think maybe give us your projections for what you think you can do this year or, or next year. Okay. So in any case, the integrations, what's going on with them? So this is all building on what Anthropic is calling their MCP protocol. This is something that they built that the entire industry has adopted. But it essentially lets different models, different AI models get data from a whole bunch of different sources. So, you know, different business tools, different content repositories, app development environments. So essentially their new Integrations, it's going to kind of tap into their MCP system and it's going to let developers create and host apps and app servers that essentially are going to help Claude's capabilities. So yeah, it's going to be pretty interesting. This is what they said specifically in kind of their, in their blog post about it. They said when you connect your tool to Claude, it gains deep context about your work and understanding project histories, task statuses and organizational knowledge and can take actions across every surface. So who actually is integrating and is going to give us a good idea of what the heck you can actually do with this thing? So there's a couple different partners that are right off the bat coming into their integrations. Number one, Atlassian, obviously huge. Number two, Zapier, obviously also huge because they have integrations with a lot of different things. Cloudflare, Cloudflare, which is going to be more of your backend stuff. Intercom, which is kind of customer support. Square, which is payments and also PayPal. So the Pay the Square and PayPal are kind of interesting integrations, actually making payments, charging for stuff. This is very interesting, getting these AIs to be able to do that. But essentially all of these are going to expand what is currently able to do. Atlassian is going to let you summarize and create pages in their Atlassian Confluence workspace, which is interesting. Zapier is going to integrate and allow you to connect your apps to your automation workflows, which I think is actually quite useful. Maybe one of the most useful integrations there would be the Zapier one, in my opinion. They have a demo where essentially you're asking it, you know, what's on my calendar? So it has a pop up from Zapier that's like, hey, you know, they want to check your calendar via Zapier. Do you allow it? You can allow it. It goes and pulls it in. So because your Zapier can be connected to a ton of different things, it's, it's actually you're connected to much more than you'd think. So anyways, in their demo, it shows you what's on your calendar. The person's like, hey, tell me more about my meeting with Lars. And it's like, okay, I'm going to look up more information on Lars. And so it like, it's like permission to access the Internet and you're like, okay, allow it. So then it's going and like searching on the Internet. And then it's like, here's what I found about Lars. And it has like this breakdown about who this guy is, what his position is, what his company size is and all this kind of information about him. So I think this is a very useful tool. And Zapier integration makes it so that beyond just like, it's not like, oh, cool, you can integrate with five softwares. Like with Zapier, you can do a lot more, right? Like it's checking your calendar, it's searching the web, it's doing a bunch of really interesting things. So with all that being said, one of the other new features specifically that they've added, and I kind of touched on it, that the searching the web thing is their advanced research. Essentially it's just going to allow Claude to go and search through the web and other things. But one of the big things is that it actually allows you to look at internal and external sources to get what they call, quote unquote, more comprehensive reports on a topic. So anyways, it's anywhere from, you know, five to 45 minutes in any case, in this advanced research that's doing these really big kind of deep dives. It's interesting because they can search through the Claude desktop app and actually see what's on your computer. So it can search through your local drives, which is really interesting. No one else is doing this. And so I think Claude was smart and built an actual app you can have on your computer. So it's looking through what's on your actual computer. You obviously have different softwares to look through what's on the Internet. Cloud has recently added that. And then what they've actually said about this is quote. With its new ability to do more complex research available when you toggle the research button, Claude breaks down your request into smaller parts, investigating each deeply before compiling a comprehensive report. When Claude incorporates information from sources, it provides clear citations with links directly to the original material. So this is pretty useful. It's, you know, they wanted to get around the problem of like grabbing information and not citing the source. People got quite mad about that when other AI models did that. So this is their solution. But kind of everybody's doing this. There's been a whole bunch of these deep research tools. The one I've used a lot has been Grok, but I know that OpenAI, Copilot and Gemini are also all doing that. So, so, and of course, you know, ChatGPT is kind of what everyone's using to do this, but it. This, you know, this isn't a brand new thing, but it's kind of cool when you start tying it together with some of their other tools like their integrations. So it's actually can, you know, beyond just looking on the web, it's like can look on your computer or your company files or your company documents. I think this is all really useful. They were, you know, previously anthropic research tool was one of the fastest. I think right now it's, people are saying or you know, okay, I'll give you a quote from TechCrunch. It's like. But the results tended to be shallower in part because research didn't employ a reasoning model. So one of the big criticisms that this is getting is beyond the deep research. It doesn't necessarily have a, it didn't at the time of this have a reasoning model that would, I mean, I guess the big difference here is like it could do research and get a lot of in depth content, but the reasoning models essentially run every single output that it generates through a series of questions, like reasoning questions to verify their accuracy. So a lot of the other models are doing this and some people are saying anthropic's not doing this. This is probably something they'll, they'll crank out here pretty soon. And I think especially with their integrations this tool is very, very useful. But it is good to know that that's not something currently baked in. So it's faster than other models. Like it can get you a research report done quicker but doesn't always mean it's like as accurate. It might be 10% less accurate than some others. And that's just a random number I'm throwing out there. But in any case, very cool. I'm super excited about the ability to connect a lot of these different apps, especially Zapier to Anthropic. There's going to be a ton of cool things coming out. And this is like I said at the beginning, a precursor to what we're going to see in agents. They put this out here first, we start using it, we start integrating with apps and once the agents come around, they're going to be able to plug into our, either our accounts or, or you'll be able to sign up, you know, with new accounts. But it's using the, essentially all of the same systems that this is built and it's able to go out and use all of these connected accounts. A lot of people are freaking out because MasterCard and Visa are building a system that essentially allows you to connect your payments to AI models and give AI models credit cards pretty much. So this is kind of one of those things that in the, in that direction, and it's. It's definitely very interesting. Hey, thanks so much for tuning into the podcast. If you learned anything about Anthropic and what they're doing doing, I would love it if you'd leave a review. And again, make sure you go check out AI Box AI, the project I've been working on for the last two years that has finally launched our beta. I'm super excited about that, and I hope that you love it. Thanks so much for tuning in and I will catch you next time.
The Mark Cuban Podcast: Episode Summary
Episode Title: Claude Gets Even Smarter with Access to Key Web Tools
Release Date: May 23, 2025
In this episode of The Mark Cuban Podcast, host Mark Cuban delves into the latest advancements in artificial intelligence, focusing on Anthropic's AI model, Claude. The discussion centers around Claude's new integrations with various web tools, its enhanced research capabilities, and the broader implications for the future of AI agents in business and technology.
Anthropic has recently unveiled significant updates to its AI model, Claude, enabling users to integrate a wider array of applications directly into the cloud. This move is perceived as a foundational step toward the proliferation of AI agents that seamlessly interact with real-world tools.
"With AI and agents, Anthropic is now letting their users connect more apps directly to cloud. And so I view this as kind of the precursor for where we're going to see agents take off in the future..."
— Mark Cuban Podcast, [00:00]
These integrations are currently in beta and available to subscribers of Claude Max, as well as teams and enterprise plans. The promise is to soon extend these capabilities to the Pro tier, ensuring that most paid users can benefit from these enhancements.
Anthropic has also increased the rate limits for Claude Code, their AI-powered coding assistant. This tool has proven invaluable for startups like AI Box AI, facilitating the rewriting and debugging of extensive codebases.
"They have added and increased the rate limits for their AI powered coding tool, Claude Code, which is amazing because my company, AI Box, uses Claude code all the time..."
— Mark Cuban Podcast, [00:03]
Claude's new functionalities are built upon Anthropic's proprietary MCP (Model Control Protocol). This protocol has gained industry-wide adoption and allows AI models to access data from diverse sources, including business tools, content repositories, and app development environments.
"This is all building on what Anthropic is calling their MCP protocol. This is something that they built that the entire industry has adopted..."
— Mark Cuban Podcast, [00:07]
Through MCP, developers can create and host applications that extend Claude's capabilities, enabling the AI to perform more complex tasks across various platforms.
Claude's enhanced research functionality now allows the AI to search both internal and external sources, compiling detailed reports on given topics. This includes accessing local drives via the Claude desktop app, a feature unique among AI models.
"With its new ability to do more complex research available when you toggle the research button, Claude breaks down your request into smaller parts, investigating each deeply before compiling a comprehensive report."
— Mark Cuban Podcast, [00:12]
Anthropic has addressed previous criticisms regarding source attribution by ensuring that Claude provides clear citations with links to original materials in its reports. However, there are ongoing discussions about the accuracy of Claude's outputs compared to other AI models that employ rigorous reasoning checks.
"When Claude incorporates information from sources, it provides clear citations with links directly to the original material."
— Mark Cuban Podcast, [00:15]
Mark Cuban notes that while Claude is faster in generating research reports, there might be a trade-off in accuracy, suggesting it could be "10% less accurate" than some competitors.
Anthropic faces stiff competition from major players like Google's Gemini and OpenAI's ChatGPT. While Anthropic is introducing innovative features, Mark Cuban emphasizes the importance of not solely relying on forward-looking revenue projections to gauge a company's success.
"They said that they're trying to get to $34 billion in revenue by 2027. I hate it when these companies have forward projections multiple years into the future."
— Mark Cuban Podcast, [00:04]
He compares Anthropic's current revenue of $1.4 billion to its ambitious goal, highlighting the challenges of making such long-term forecasts in a rapidly evolving industry.
Anthropic has partnered with several influential companies to integrate Claude's capabilities:
One of the standout integrations is with Zapier, which demonstrates Claude's ability to interact with a multitude of connected applications. In a live demo, Claude accesses a user's calendar, retrieves meeting details, and searches the internet for additional information about a contact named Lars.
"They have a demo where essentially you're asking it, you know, what's on my calendar?... And then it's like searching on the Internet... here's what I found about Lars."
— Mark Cuban Podcast, [00:10]
This integration showcases Claude's potential to streamline workflows by accessing and analyzing data across various platforms seamlessly.
The episode touches on the broader trajectory of AI agents, emphasizing their future integration with financial systems. Mention is made of credit card companies like MasterCard and Visa developing systems that allow AI models to handle payments autonomously, raising both excitement and concerns about the evolving role of AI in financial transactions.
"A lot of people are freaking out because MasterCard and Visa are building a system that essentially allows you to connect your payments to AI models and give AI models credit cards pretty much."
— Mark Cuban Podcast, [00:16]
This development points to a future where AI agents not only manage information and tasks but also handle financial operations, necessitating robust security and ethical considerations.
Mark Cuban expresses enthusiasm about Anthropic's advancements with Claude, highlighting the significance of these updates in the AI landscape. While acknowledging the ambitious revenue goals and the competitive environment, he underscores the potential of Claude's integrations and enhanced research capabilities to revolutionize how businesses interact with AI.
"I'm super excited about the ability to connect a lot of these different apps, especially Zapier to Anthropic. There's going to be a ton of cool things coming out."
— Mark Cuban Podcast, [00:17]
The episode wraps up with a forward-looking perspective on the integration of AI agents in various sectors, promising listeners an insightful glimpse into the future of AI-driven business solutions.
On Anthropic's App Integrations:
"With AI and agents, Anthropic is now letting their users connect more apps directly to cloud. And so I view this as kind of the precursor for where we're going to see agents take off in the future..."
[00:00]
On Revenue Projections:
"They said that they're trying to get to $34 billion in revenue by 2027. I hate it when these companies have forward projections multiple years into the future."
[00:04]
On Advanced Research Capabilities:
"With its new ability to do more complex research available when you toggle the research button, Claude breaks down your request into smaller parts, investigating each deeply before compiling a comprehensive report."
[00:12]
On Zapier Integration Demo:
"They have a demo where essentially you're asking it, you know, what's on my calendar?... And then it's like searching on the Internet... here's what I found about Lars."
[00:10]
On Future Financial Integrations:
"A lot of people are freaking out because MasterCard and Visa are building a system that essentially allows you to connect your payments to AI models and give AI models credit cards pretty much."
[00:16]
On Excitement for Future Developments:
"I'm super excited about the ability to connect a lot of these different apps, especially Zapier to Anthropic. There's going to be a ton of cool things coming out."
[00:17]
This episode offers a comprehensive overview of Anthropic's strategic enhancements to Claude, positioning listeners to understand the evolving capabilities of AI models and their integration into everyday business tools. For those keen on the intersection of AI and practical applications, this discussion provides valuable insights into the future trajectory of intelligent agents in the corporate world.