Apple Price Hikes, Adobe AI, and MCP Connectors (Full Transcript)

[00:00:00] Speaker 1: This week on the podcast, we're talking about Apple and their recent price increases. Fable 5 is finally coming out. Also, Adobe recently released a whole bunch of new AI updates, and we'll dig into what some of those are. And then MCP connectors, what are they and should you be using them? Hi everyone, welcome to the Kevin Stratford podcast. My name is Kevin, and joining me today we have my co-host, Nick Brasi. How's it going, Nick? Hey, Kevin. It's going good. Yeah, we have, there's a lot of news to discuss, and it seems like one of the ones that I've been hearing about the most is Apple's raising all of their prices.

[00:00:42] Speaker 2: Yeah, that was, I guess, a little over a week ago they announced that. And it's not all the way across the board, but computers, iPads, and I think I had somewhere in here the average boost was around $250, $290. Some devices are $200 more, some are $300 more. I think that the Mac Studio is significantly more, like an extra $500 on top. So we're really starting to see those prices creep up, even with Apple.

[00:01:15] Speaker 1: Yeah, and Apple has historically always been a more premium brand. So you're taking what were high prices to begin with, and now they're becoming, I guess, even more premium and exclusive. But I think here I'm looking at some of the show notes, and the MacBook Pro is increasing $300 to, I guess, just about $2,000. I like how it's $1,999, so kind of in your mind you round down, but it's actually about $2,000. We have the iPad Air going up by $150, so that'll be $750. I mean, percentage-wise, that's a massive price increase.

[00:01:49] Speaker 2: Yeah, I think the one to really look at is the MacBook Neo, because that was the one, you know, that recent release that was, wow, Apple is finally making something affordable, and it sold very well. People were crazy about it because it was finally something from Apple that was fairly affordable, and it's gone up an extra $100. So, I mean, in the context of everything else that's getting more expensive, I guess it's still affordable, you know, if everything's more expensive.

[00:02:22] Speaker 1: You know, the interesting thing, yeah, I was looking at that. So it was, I think, $599, which is, you know, entry-level kind of PC pricing. Now it's $699, which, you know, $100 extra. But, yeah, if you look at the percent increase, it's about a 17% price increase. I mean, that's a big percentage increase.

[00:02:42] Speaker 2: Yeah, and when it's that low to begin with, $100 is a big difference. And there is a student discount, and I would assume it's just, because it was $599 regular, and I think $499 for students or four. It was a little bit less, and I wonder, I didn't check this, but it's probably an extra $100 on top of that, even for the student discount.

[00:03:06] Speaker 1: Yeah, and one thing that was interesting, too, is not only have the prices gone up, but then Apple stock, I guess, on this announcement dropped by about 6%. So the market reacted negatively to this news.

[00:03:20] Speaker 2: But I looked at it because this was, you know, we were looking at this about a week after the announcement, and we saw that stock drop. And I just checked it again today, and we didn't talk about this, but it looks like it's rebounded. And that's how the stock market works, of course, up, down, up, down. You know, it's not the highest it's ever been, but it looks like it's a little bit higher than it was the day they announced that price drop. Now, in the meantime, I think that Microsoft announced their second price increase on the Xbox. You know, similar devices like video game consoles are getting wildly expensive. And so it's sort of become this is the standard. Everybody's, I guess, kind of making peace with everything getting more expensive. Everything that's got RAM in it is now more expensive.

[00:04:08] Speaker 1: And what's interesting is, like, you say this is now the new standard, but I feel like in technology, the old standard used to be things get faster and cheaper over time. And it seems like that's kind of been flipped on its head now, and it's really heading in the opposite direction. And I think to your point, it's not just Apple where we're seeing these price increases, but I think Microsoft, other device manufacturers are increasing prices, and not just tech products, but you go to the grocery store and prices have gone up. It seems like prices just kind of across the board have gone up, but especially now with tech products. And I feel like this might have something to do with all the AI infrastructure build out that's happening.

[00:04:47] Speaker 2: That's the general consensus. Yeah. Do you think that's the case that, you know, they're building out AI data centers, so RAM is expensive for everybody and it's just driving up those costs?

[00:04:57] Speaker 1: Yeah. So, like, the one thing I think of is you have, I mean, like, it's been in the news all the time now, you have all these massive data centers being built, you know, in communities across the world. You have all these data centers going up. And then, you know, you think about what goes into a data center. Well, you know, you have servers, you have lots of RAM, you have the hard drives. So you have all these kind of traditional PC components ending up in these servers. And so I think what's happening is, you know, you have your consumer electronics, which used to be some of the biggest purchasers of, you know, RAM and hard drives and, you know, processors. And now there's another customer in town, a big customer in town. And I think they're taking a lot of that supply and they're shifting it towards these data centers. So basically, we as consumers, we're now competing against data centers for these same components.

[00:05:49] Speaker 2: And there's that concept in any economy like this that you're voting with your dollar. We as the consumer tell the product producers what we want by what we buy or what we don't buy. But I don't feel like that really holds in this case because we don't get a say. No individual consumer has said, I would prefer to have more data centers than an affordable computer. That's not an opportunity that a consumer has been given. In this case of that voting with your dollars almost being taken away.

[00:06:22] Speaker 1: Yeah. You know, I think I think what's happening is these these A.I. companies are raising so much money. You have this massive, you know, build outs of A.I. data center infrastructure. And these companies are willing to pay a lot of money for memory, for for GPUs. And in many cases are probably willing to pay more than what the consumer is willing to pay. And so, you know, supply and demand. You know, there is massive demand for data center operators. There's only so much supply. So what happens? The prices start to elevate and just increase and consumers then have to decide, am I willing to pay these higher prices or do I sit out?

[00:06:59] Speaker 2: I heard a thing and I don't know if it's true because I can't track the source. But somebody was saying Steam, who they have a marketplace for video games and they make some of their own hardware, Steam machines, Steam Deck. And they were saying, we have our RAM distributors. They give us a price. And if we argue, they just say, OK, we're done and they walk away. So to a certain extent, these these device manufacturers, they just have to deal with it. OK, that's the price for RAM. We'll take it because there's no other option. And then they've got to bump up all of their other prices.

[00:07:35] Speaker 1: Now, what I what I think is interesting, too, is you have a lot of companies like I think Anthropic now, the maker of the Claude models and Fable 5. They're actually looking at building their own chips. So I think what's happening is the supply has been so constrained. There's the prices are so high that, you know, you have Microsoft, you have Google, you have Anthropic. All these companies are getting into the chip business because the prices have gone so high and they're so astronomical. So I wonder I wonder if this is more of a temporary thing where we see, you know, supply is really constrained. We see prices skyrocket. I wonder if that invites more people to compete in this space. And, you know, maybe maybe maybe not in the short term here, but maybe, you know, several years out as more entrants come in and more supply emerges. Do prices start to come down again?

[00:08:22] Speaker 2: Yeah. And when will that supply even out? I mean, at some point they have to finish building data centers. I'd like to believe, you know, I'd like to see this sort of even out. I think an interesting angle to this is, you know, we look at that MacBook Neo and that was their loss leader. But we didn't see any of these price increases on the iPhone. And I wonder if this is it'll it'll be very good to watch what happens in September when Apple traditionally releases their new iPhones. Will new iPhones this year be significantly more expensive or will they have a way of absorbing that cost to make iPhones as if not that they've ever been super affordable, but as affordable as they were just to keep people in the Apple ecosystem?

[00:09:13] Speaker 1: Yeah, it's been interesting because like what happens is right. Prices go up. You're probably going to get less demand. And and so as Apple raises their prices, you're probably going to lose some consumers. And I think maybe part of the reason they kept iPhone pricing where it is, is they're going to lose a little bit of market share there as the pricing goes up. But I, I would guess if all these component costs are increasing and you're, you know, you have to pay a lot more for RAM and for all the different components. My guess is the iPhone will have to follow suit and the prices will also have to go up to match all these, you know, all the other price increases.

[00:09:51] Speaker 2: Or does Apple as a trillion dollar company somehow absorb that for one specific product line?

[00:09:59] Speaker 1: Right, right. Maybe, maybe you want to protect market share and you're willing to take, you know, smaller margins on it. I don't know. Yeah.

[00:10:07] Speaker 2: And it may suggest that so far that's what they've been doing because we haven't seen any price increases on the iPhone when pretty much every other product they make has increased.

[00:10:17] Speaker 1: Right, right. Yeah. But I think one thing's clear, you know, I read stories about companies wanting to build data centers in space. I think there's still a lot of runway on building out AI infrastructure. Think so. My, my, my only guess would be that prices continue to rise, at least for the foreseeable future. So the takeaway is if you want to buy a new iPhone, now's the time. Yeah, maybe, or, or a computer or anything else.

[00:10:46] Speaker 2: I did see, I didn't really dig too deeply into this and put it into show notes or anything. But I was seeing there's going to be a big investment in Korea in building out AI infrastructure, a big investment in China. We're talking about hundreds of billions of dollars of investments. And yeah, they're building a lot of data centers here, but that's happening around the world. And that's going to affect the market as well.

[00:11:09] Speaker 1: Yeah. We might be in it for a while. Yeah. More, more demands coming on board. So probably, probably higher prices as a result of all that. Yeah. Good. Like everybody wants to use a computer. So, so maybe, you know, you heard it here. If you want to, if you want to buy, you know, new electronics, now's the time to do it. Yeah.

[00:11:30] Speaker 2: They're not going to get less expensive, at least not for a few years. Right.

[00:11:35] Speaker 1: Right. And I guess a lot of, a lot of this infrastructure that's being built out is, is being, you know, built to support all these different AI workflows. And I know a big, a big announcement that came out of Adobe is a lot of the different Adobe products are going to support the, what's called the fly or fly AI assistants that'll live inside various Adobe products like Photoshop, Premiere, Illustrator. And Nick, I know you live in that ecosystem. So what are you excited about? What are you most looking forward to with this announcement?

[00:12:09] Speaker 2: Well, I think there's a tremendous opportunity to streamline workflows. I think that firefly is, is fairly well known and people use it for generating content, which is not something I'm so interested in. I'm, I, we've talked about that before. I like people making content, but there are opportunities for these AI tools to streamline the workflow. I think they announced that the AI assistant was coming in mid-June, somewhere around there, they made the announcement. And then just this past week, they started putting it out in the betas. And so I played around with Premiere beta and a little bit with Photoshop beta with the new AI assistant. And it feels like to me, you got to be careful when you're judging a beta, right? You don't want to pass, you know, give it a bad grade when it's the whole point of a beta is they're still figuring it out. But to me, it felt like the bones were there. The structure is there. But what I asked it to do didn't really work very well. And it, it feels like, you know, you go into Google documents and you have your Gemini panel. You go into Microsoft Word and you've got your co-pilot panel. That's becoming the standard to have a panel on the side of your interface where you can ask questions. And that's basically what it boils down to. And I did something like, okay, AI assistant, label all of my live action clips as yellow and label all of my screen capture clips as red. And it did that. You know, so as far as like organizing your, your data or organizing your media, your raw media, I think it's going to be so great for that. I did another thing where I asked it because I had a shot where I didn't have the right headroom. And I, I punch in on my shots, like I crop in a little bit. So I told it, okay, just typing in the app, make all of my AI clips have more headroom. And it did it, except it moved the frame down so far that there was empty black space at the top of the frame. So it was smart enough to do what I asked it to do, but not smart enough to recognize that it was screwing up the framing. Which seems like an important thing for a graphic or editing or media production app to do. So there are a lot of little things like that where, yeah, the structure's there, the tool's there, but I think it's good that it's still in beta. I don't know how useful it's going to be yet.

[00:14:36] Speaker 1: Yeah. And I wonder if that's the type of thing where, you know, once you have the foundational piece in place, then the model can continue improving. And then you start, you know, you get better and better at those different tasks. I think we have to extend that assumption. Yeah.

[00:14:51] Speaker 2: I think that we have to assume that they're going to build it to be far more capable. So I think it's very, very promising. But if anybody's interested, you know, I wouldn't recommend just install the beta and go have fun because I don't think it's quite there yet. I think it's still going through that growing process.

[00:15:11] Speaker 1: Right. But I think the vision then, so you mentioned that you work in Premiere. I guess the vision would be that eventually you could upload your footage or load your footage into the app. It could identify what all the different footage is, maybe even pull together an initial edit of the video. And so I think Adobe's promise is that it removes or eliminates a lot of those repetitive steps like, you know, organizing footage, classifying things, maybe trimming the, you know, the silence at the beginning or at the end of the clip. So kind of those repetitive things that, you know, maybe don't add that much value, but can help kind of streamline your workflow. Is that kind of what they're, it sounds like that's what they're promising with these tools initially, at least. I think that's the plan. I think that's the idea.

[00:15:57] Speaker 2: Yeah. I produce a good amount of videos for your YouTube channel, which are instructional videos. And you probably do the same thing where you'll record a stretch of recording the computer screen. But there'll be a lot of pauses there when we record, especially when we type a prompt into an AI assistant, we have to wait a minute for the response to come. So in our raw, unedited video, there'll be these pauses. And so, you know, I was experimenting with that. I asked the AI assistant, put all of my screen capture clips into the timeline. And it could actually tell the difference between live action and screen capture. And that was great. Put them in the timeline. I said, OK, edit out all the pauses. And then I said, well, I can't analyze the audio. I'm going to have to do this with the transcript. And so it took a while to run the transcript, transcribe the clips, and then analyze the transcript. And then in the end, it more or less did that. It wasn't perfect, but it more or less did that. So I think the things you're describing are the things that ultimately it will do. But it's almost like skills in copilot. It needs to build out, Adobe needs to build out skills, abilities that it has, ways it can interact with your material that don't feel like they're quite there yet in the beta.

[00:17:15] Speaker 1: Right. And I've also heard it described more as, I guess what they're calling agentic AI, where it's not just, you know, ask one question, get a result. But I think the idea is, you know, let's say you're in Photoshop and you have a photo and you want to remove the background. It'll, you know, identify the subject. It'll get the back. So basically what would have been a multi-step process, you can now have this agentic AI go through and complete all the steps involved in that. Yeah. I think the way I've heard it described is you kind of, you define the goal that you want to accomplish, and then it'll work through multiple steps to get there.

[00:17:50] Speaker 2: Yeah. In fact, I tried something like that in Photoshop because I had a graphic where it was more or less a photo with some text on it. But I couldn't find my original PSD file. I just had a flattened image. So I asked it, can you extract this text and give me just the text? And it did that, like you're describing, multiple steps. And ultimately it redid the text and didn't match the font and the style that I had, but it did give me something like what I asked for. So like you're saying, it goes through those multiple steps to a result. And once it's finished with beta, maybe it will do that in a very powerful, effective way. Yeah.

[00:18:30] Speaker 1: So yeah, it'll be interesting to see with all these AI tools, will the so-called creators who use these different apps, will they end up spending more time creating or will you spend more time directing the AI to do the creation for you?

[00:18:47] Speaker 2: Well, ultimately, I don't think the user will choose to waste that time. You know, I kind of sat down and said, well, can I do my normal workflow just with AI prompts and found it was taking way too much time. And my conclusion was, OK, I'll just do it the traditional way. But if the AI gets to a point where it is saving time, and I think that's the goal, then people will make that choice. I don't think people are going to suffer the AI taking more time, but I think they'll keep refining it and making it better and better.

[00:19:23] Speaker 1: And I wonder, I wonder too, like even today, if you have a new user coming in who might not know all the shortcuts and all the menus, like you've been using, you know, Premiere and the Adobe Suite for a long time. But if you have a new person come in and maybe they're not as familiar, maybe it is already saving time for some users. And then, you know, we'll see if it, you know, helps more and more over time.

[00:19:43] Speaker 2: Yeah, it's a mixed blessing. It democratizes the technology. It lets people who would not otherwise be able to do a job do that job. But it also doesn't particularly breed great editors. You know, it doesn't force somebody to learn how to be a good, effective editor. So I think we gain something over here and we lose something over here. And hopefully the people who want to dedicate themselves to being great editors will find those paths and maybe we'll just benefit from the democratizing effects.

[00:20:15] Speaker 1: And I think this goes back to a conversation we had last week on the last episode where, you know, you don't want to just, you know, you have all your video files and you ask AI, hey, pull together a video for me. Like that's one approach. The other approach is like it can assist me with some of the steps, but I'm still, I'm still driving it. I'm still making the creative decisions. And, and this AI is assisting me making my job easier, but ultimately it's still kind of my vision that's coming to life. It feels like that's more kind of the better outcome. But like I think of the example, like, you know, writing an email, like, yeah, you could have AI write it, but you probably want to give your own input. You probably want to, you know, refine it and make it yours as opposed to just, you know, completely relying on the AI.

[00:20:59] Speaker 2: Yeah. And as we transition, will we start spotting and be able to tell, oh, that video was edited by AI? You know, did they just run the quick tools to cut out pauses and then hit publish rather than going back and doing that human pass and doing the things that are, you know, a real editor would do? I don't know if that's quite AI slop, but it's, it's something to look out for.

[00:21:24] Speaker 1: Yeah. Yeah. And so, so speaking of new things, so not only do we have, you know, more AI capabilities in Adobe, but we've been talking about this for the last few podcast episodes. Fable 5, Anthropix, you know, super powerful model is finally here. It's out.

[00:21:43] Speaker 2: Or at least in a different form, but it's finally here. Yeah. I was reading a bit about, and we suspected this, that they'd have to put some safeguards in place. And I haven't used it. Clearly, I didn't use it before it was delayed. So I can't really compare, you know, whether a user will feel a difference, but apparently there are some significant safeguards in place, specifically in the world of cybersecurity. So that, you know, Claude can't go out and hack other people's systems. Seems like an important requirement.

[00:22:17] Speaker 1: I think one example I've seen, let's say you're a software developer and you ask, you know, Fable 5 to identify software vulnerabilities in your app that you're developing. It'll present a message that says, sorry, you know, it says that it can't help you with it. And then it refers you down to Opus 4.8 as an alternative. Right. And those safeguards, kind of the number of potential queries that could hit those safeguards has been expanded. So, but I mean, it's back and, you know, for lots of queries, it will be active and it will assist. So it seems like they were able to find a compromise with the U.S. government and get the model back out.

[00:23:01] Speaker 2: From what I was reading, and this was an anthropic statement, they've got these classifiers, like you're describing, where eventually somebody will ask for something and Claude will just say, I can't do that. Or it'll roll over to, like you said, Opus 4.8 and then it'll do Opus 4.8's version of it. But they say that any one individual classifier is probably not enough, which is why they have multiple overlapping classifiers. And at some point, if you are using Opus 5, and this will be an interesting thing to look out for, at some point it will say, sorry, can't help you with that. And I want to see how often that pops up. And theoretically, that's a good thing when it pops up.

[00:23:47] Speaker 1: Yeah. And I guess the question, too, is, like, at least moving forward, how involved will government be in the release of new models? Because here, this was kind of one of the first examples where the government stepped in, stopped the release, they wanted additional safeguards. Will companies now, say when OpenAI or Anthropic wants to release additional updates, does the government step in and say, hey, we need to review test cases, we need to check the vulnerabilities of this? I wonder, is this now becoming the norm? Or what's it going to look like?

[00:24:19] Speaker 2: Well, I found a story, and I threw this in our notes as well, but there's only one story. And the story came from TechCrunch. And so I don't know, you know, since the story didn't get picked up too wide, how big of a growing concern this is. But OpenAI is rolling out GPT 5.6. And apparently, they just kind of talked to the government oversight agencies and decided, we're going to limit that rollout. And we're only going to roll it out to a few trusted organizations, and we're going to slow roll this and analyze that. But then OpenAI was saying, this cannot be the standard. We're going to do it now. But government oversight on every model rollout can't be the standard. So I think they've got to figure it out. And I think both Anthropic and OpenAI are kind of calling for that. Like, let's get some sort of standard process here.

[00:25:15] Speaker 1: Because they're just going to get more powerful and more risky. And what was interesting is, you know, you could argue the influence of the organization, but the U.N. I was looking at this as I was kind of preparing with the show notes. But they've actually warned that AI innovation and development is outpacing the ability to regulate it. So these companies are moving so fast. Models are improving so rapidly that, you know, perhaps at least they're claiming that government's ability to, you know, kind of, I guess, regulate and oversee what's being done is just not keeping pace. So that's interesting. And so they're warning that, you know, especially, you know, these AI agents are becoming more and more capable. Models are able to have deceptive behavior. You can also, you know, you think of some of the cybersecurity risks or, you know, potentially, let's say you use AI to develop, you know, biological weapons. Like, there are all these, you know, really terrible ways that you could use AI as well. And I think their concern is, you know, government's just not in a good position right now to keep up with all of that.

[00:26:21] Speaker 2: You said something a moment ago that I hadn't really clocked into. Clearly, we don't want these AI models to be able to exploit vulnerabilities in secure software. Clearly, we don't want that to happen. But you mentioned, well, as a software developer, don't you want to be able to ask the AI to identify vulnerabilities so you can fix them? And so what's that threshold between let's let it identify problems so we can protect ourselves to let's give it the ability to actually exploit those problems? And like you say, the government is not really in a position to write those rules yet. Yeah.

[00:27:01] Speaker 1: And it, you know, it feels very much like a race, right? Like in the past, you know, software companies would develop their software. They would test it, you know, try to find all the vulnerabilities, all the bugs, fix it. In some cases, people would report bugs to the company and then they would fix it. I think what this AI paradigm shift has done is everything is just so much more accelerated. And I guess, too, like you look across, you know, the entire tech industry and there's a lot of legacy code that's out there. And if you could just deploy these AI models to start analyzing all this code to find whatever, you know, vulnerabilities exist, it's hard to stay on top of that. But on the other hand, you know, maybe you have these companies, too, where, you know, they can they can use the same tech to look at their code and say, hey, we need to fix things before other people find these vulnerabilities. So it's very much an arms race where you have the bad actors and the good actors trying to stay ahead of the other.

[00:27:58] Speaker 2: And accelerated so profoundly, you know, it's a classic battle, but it's just going so fast. How can anybody keep up with it?

[00:28:07] Speaker 1: But then you look at like I also like I look at government and like a lot of these tech companies like Anthropic, OpenAI, they're hiring, you know, the best and brightest people in this field. And then I don't I don't know how the government, if they're able to recruit as many people to be able to oversee these models and, you know, make the appropriate decisions. Do you rely more on the company to do that or do you need some of that in kind of a centralized role? And I don't know if the government, do they have the ability to stay on top of what all these companies are developing?

[00:28:39] Speaker 2: Yeah. Can they tempt people skilled enough to come into the oversight organization and do that job that has to be done? Yeah. Have you heard this? This was something I I saw somewhere and again, I can't really cite it, but somebody was saying there's going to be some sort of 9-11 like catastrophic event around AI that is going to change the entire landscape and could even be the thing that turns everybody against AI. And that that's kind of scary to think of. Are we building towards some actual specific event that's going to define all of this moving forward?

[00:29:20] Speaker 1: I mean, I always one thing I always think back to is, you know, just think on a day to day basis how much technology you're using and how much the systems you're using are connected to others. You know, even even if you think you don't use technology that much, you know, just go into the grocery store for your groceries to get there. You know, you have a supply chain and they have systems that, you know, talk to the grocery store, that talk to, you know, the talk to all the different intermediaries in that entire process. And there are all these, you know, you know, how do you track the inventory and all these other things? If if some tool or, you know, bad actor is able to manipulate that or disrupt some of these, you know, technology processes, and that's going to have catastrophic impacts. You know, I just think of like supply chains, if somehow you could disrupt that. I mean, I always think back to what was it Y2K back in 2000. And I remember there was I think it was a meat packer where their system basically they evaluated whether to keep or destroy the meat based on the expiration date. And so for the year, they just had the two digits and you went from 99 down to zero because they didn't track four digits for the year. And so they their system made the decision to destroy all the inventory because it's too old. And that's just a, you know, purely that's not a bad actor coming in and manipulating that. That's simply, you know, an honest mistake. And they made a decision based on a, you know, a bug. But but now if you could have these super powerful, you know, let's say AI models with, you know, certain prompts that are directed to come in and do certain actions. I mean, I don't even want to think about what type of disruption it could cause. And so there is there is a lot of concern that, wow, what if this is used in a bad way?

[00:31:08] Speaker 2: Yeah, I don't have much to add to that except except fear. Yeah, right. Just hope hopefully the good guys innovate faster than the bad guys. Yeah. Yeah. And hopefully there are people there, you know, instituting some sort of oversight. And if not the U.S. government, then perhaps the European Union. You know, often we've seen the EU step in and make policies that then affects, you know, global policy. But these are the things we definitely need to be looking out for. If if the unsupervised version of Fable 5 has the ability to exploit cybersecurity, you know, implement cybersecurity exploits, then we seem to be pretty close to that horrifying outcome already.

[00:31:59] Speaker 1: Yeah. Yeah. I think that's the yeah, the risk with with with every new technology, it can be used in good ways and bad ways. And you just hope that that gets used in more good ways. And I think more than hoping, I think there needs to be some oversight, some regulation. But it's trying to figure out what is that balance. So you continue innovating, but then, you know, you don't bog things down too much. But at the same time, you still want to be safe.

[00:32:23] Speaker 2: Yeah. Yeah. And we're getting tremendous benefits all over the place from from AI anyway.

[00:32:30] Speaker 1: Yeah. And I was going to say it. I don't know if this this is a there's a lot of benefit from this. But I guess it also kind of plays into risk. But you have these MCP connectors that are becoming more and more prominent. And I think if anything, this is kind of interconnecting all the systems even more. Maybe just just as background for for people who haven't heard of this before. You have you have something called MCP. And just at a very kind of simplistic level, it allows different AI tools to communicate with other tools that you're using. And I think a really basic example is, you know, let's say you're using chat GPT so that, you know, you ask a question, you get a response with MCP. You could connect, say, chat GPT to your Gmail or your Google calendar. So this way, when you go into chat GPT, you could say, hey, like, you know, what is my highest priority email sitting in my inbox right now? And it'll look across all your emails. It'll read them. It'll understand which ones are important, which ones aren't important. And it will surface the most important one. So it's a it's a way for different AI tools to communicate and interact with all these other tools and services that you're using. So whether it's, you know, Gmail or Google Calendar or maybe it's Salesforce or maybe it's HubSpot or whatever the tool happens to be. But I think one thing we're seeing is these tools are becoming more and more interconnected now with these connectors.

[00:33:51] Speaker 2: Yeah, I don't think it's a particularly new thing, the MCP connectors. But I really noticed them a lot recently because a lot of the videos that you've been putting out on your channel and that some of us additional contributors been putting out on your channel. Just suddenly, it seems like MCP is all over the place with these with these different apps and tools that we're covering. You recently did a Cloud MCP video.

[00:34:18] Speaker 1: Yeah, so the way I think of it is, you know, when people first start using AI, kind of that first step is, you know, I ask a question and in here it gives me a response. So I think like when you when you when you're a beginner with AI, that's kind of your first step into, you know, first step in the you're dipping your toe in the water. I think step two then is like, oh, like AI is really powerful, gives me these good responses. It'd be nice if it can now take action and do things for me. And I think that's really where, you know, that this is really where AI takes that next step, where, you know, now, you know, let's say you work in, you know, maybe you're you're a mechanic and you have a new customer. You know, in the past, you'd go into your CRM system to add that customer. You're just talking with AI and you say, hey, can you add this new customer? It'll just go and do it for you. So now it's taking action versus just giving you a response. So I think I think a lot more people are becoming aware of the fact that you can connect AI to all these other tools and that it can now take action on your behalf.

[00:35:14] Speaker 2: Yeah, where this really clicked with me, I do a lot of training for Microsoft Copilot and a bit of the pitch for Microsoft Copilot in an enterprise situation has been, you know, if you use Copilot, then it can access your Outlook email. It can access your documents on one drive, access all of your meeting transcripts and teams. Isn't it great that you can leverage all of that stuff? And if your business is completely in the Microsoft 365 environment, it's all there. And it really clicked with me recently, and I should have clicked with this earlier, but MCP connectors almost give you that ability to build that foundation with other tools. So, you know, if you do your mail over in Gmail, connect the Gmail MCP connector, but you also have files on one drive, connect the Microsoft 365 connector. Or you could also make the argument if somebody doesn't want to use Copilot, they prefer to use Cloud. Well, then you can connect Microsoft 365 into Cloud and then you have that data foundation inside of Cloud. And then the ability to build this foundation with multiple apps, all of my meetings are here and all of my documents are here and all of this other stuff. It's just now this knowledge base within the AI tool that you use.

[00:36:35] Speaker 1: Yeah, like I think ideally, I guess the vision is, you know, you choose some AI tool to use and then you connect that AI tool to every other tool that you use. And then, you know, your Cloud or your chat GBT or your Copilot, that's your central location you go to. And if you have questions about your meetings, questions about your documents, it knows kind of which tool to go to to get that information. So I guess in a sense, you almost abstract away all these other tools and you interact with the AI and the AI interacts with every other tool that you've been part of.

[00:37:10] Speaker 2: Yeah, and I think it is worth mentioning that MCP connectors are available in Cloud and in chat GPT, and it's not really fully implemented in Copilot. And that's something that I'm very much keeping an eye on. And in something like Copilot Cowork, you can use plugins, which is a set of tools that Microsoft has sort of white listed. And these consist of, you know, an MCP connector with some skills mixed in that almost a recipe that Microsoft has whipped up. But the standardized MCP ecosystem is not really there in Copilot. And I'm wondering, is Microsoft working on that? Are we going to see that implementation in Copilot? Will we see it in Gemini the way we see it in Cloud and in chat GPT?

[00:38:02] Speaker 1: You know, what's interesting is I was reading about MCP that because it's basically a standard that all these other companies have agreed to, that this is kind of the standardized way that your AI communicates with other tools and systems. And it's been one of the fastest adopted standards, you know, in tech. Like usually it's hard to get agreement between all these different tech companies. And, you know, here this is, you know, got an agreement from so many different companies. They've gotten on board, they've been, you know, developing support for MCP. I think like the way I think of it is it's kind of like how USB-C has become one of the dominant ways of, you know, connecting devices. You have a standard, other companies have agreed to it. And I think MCP is like that. I wonder if, you know, Microsoft, I'm sure Microsoft, Google, all these other companies are going to support it. In a sense, I guess Microsoft had a little bit of an advantage where, you know, they have a lot of your data, Copilot's connected to your data. And so maybe the need to kind of support MCP at the start wasn't as strong because, you know, they already had all these connectors. But it feels now like if you're a company and you want your data to be used, there's almost an expectation now from consumers that you offer connectors.

[00:39:17] Speaker 2: Yeah, every app developer wants to have their app available as an MCP connector so that, you know, they'll have more and more subscribers and purchasers and just be part of the ecosystem. And I think you make a good point. I mean, the part of the appeal of Microsoft 365 for an enterprise is that's your entire solution. And it's also all contained within the same secure environment. So once you're in the Microsoft ecosystem, you don't have to worry about the security risks of going outside of it. Maybe that's why Microsoft isn't pushing so hard on MCP.

[00:39:51] Speaker 1: Yeah, no, it's interesting because I think a lot of times, what is it, like the network effect, the more people who have it, the better off you are. And I think like I think with an ecosystem, if you're offering an AI tool or an AI model, if you don't have, you know, connectors in place, I think that's going to disadvantage you. I think like people are going to look at it and be like, hey, like what's your ecosystem of connectors you have available? You know, if you're and I think, Nick, you made a video on Granola. But if you know, if you can't connect Granola to your chat GBT or your Claude, do people say like, well, I'm not going to use this solution then. Like I need it to fit into the ecosystem of the AI tools that I'm using.

[00:40:31] Speaker 2: Yeah, I think Granola is a perfect example, because if you're not familiar with Granola, it can transcribe any meeting from any meeting tool. So I'm not loyal to any meeting app. Sometimes I'll use Zoom, sometimes I'll use Google Meet and I'll bounce around, but Granola will transcribe all of them. So with that MCP connector, if you are using Granola as designed, then your AI assistant now has an insight into every single one of your meetings. And that's incredibly powerful. And, you know, if for those people who are watching the video version of this, I've got caught up here. And what you want to do, it's actually really easy if you look for the customize option on the sidebar on the left, then go to connectors. And you can see I've got a few connected here. But then you can hit the plus button, browse connectors, and you'll see this is a huge library of different apps and tools that you can work with. So if you use, you know, Monday.com to organize your projects, or, you know, if you use Canva to design graphics, like you can connect all those things together and then give your AI tool insight. And I believe ChatGPT works basically the same way. Connect your connectors like this. And then I think sometimes you should say, like, you know, review my Granola meetings, you know, actually put that in your prompt. But a lot of the times you don't need that actual language. A lot of times you can just say, you know, mention something about your meeting and it knows to check, you know, the right connector for that meeting information or whatever. But then you're just having those natural language conversations with the AI assistant as normal.

[00:42:13] Speaker 1: You know what, I think this leads to a future. I think it's Satya Nadella said this, where today we have this paradigm where like me as a user, when I interact with tools, I do so through the user interface. Right. So, you know, if you're buying a flight, maybe you go to Expedia and then you research flights and then you purchase it. But I think he's been hinting at, I guess, more of an agentic future where instead, you know, you ask your AI or your tool, maybe you say, hey, I want to plan a vacation, you know, find the, you know, the cheapest ticket and then, you know, book it for me to this destination. And I guess leveraging like MCP and these different tools, your AI could then go off to this other experience, you know, book the ticket. And then, you know, so you're no longer, you know, interfacing with the UX. You're just kind of, you know, writing out or describing your end goal and then it goes and accomplishes it for you. But I think like here with the granola example that you were sharing, you know, you don't necessarily now have to go into granola and get your meeting transcript. Instead, you just ask questions about it and, you know, the AI already has access to that. And so I think it's almost, it's like questions and kind of the text interface becomes almost your main way of interacting with the computer. And then kind of the agents or the AI kind of handles everything else kind of on the back end to get you what you need.

[00:43:39] Speaker 2: You know what I find fascinating? There's a template for this that we've seen. It's Star Trek The Next Generation. And far too often I see technology that we have these days like building towards what we saw in Star Trek as far as they were predicting what we would need. And if you watch some Star Trek, you'll see they just talk to the computer. You know, I need to know the orbital velocity of this comet so the ship doesn't hit it. And they just go to a computer terminal and they say that and the computer figures it out. And it always seemed ridiculous. But now I know, you know, in the context of the AI that we have, well, there's probably an agent under the hood that has the skills for orbital mechanics. And it has, you know, the skills related to the sensor array. And it, you know, understands natural language prompts. So a lot of this stuff we've seen a template for. We're just kind of working towards the sci-fi we've seen.

[00:44:35] Speaker 1: So Nick, what you're saying is in just a few years, they'll be able to beam me up to other location. Is that coming next?

[00:44:43] Speaker 2: Yeah, that'd be nice. Let's get anthropic on that one. But let's have some safeguards in place to make sure that we don't have any transmission errors there.

[00:44:54] Speaker 1: Right. If you get beamed up, your entire self shows up in that new place. Yeah. There's risks.

[00:45:02] Speaker 2: Yeah.

[00:45:03] Speaker 1: Let's be cautious. No. So, yeah, I think, yeah, MCP is, it's really interesting. It gives your AI kind of a lot more capabilities. It gives it access to a lot more information. It can now take action. But I think, like, people still have concerns that, hey, if I'm connecting my AI to everything else I use, what if that AI doesn't behave in the best way? And now it has access to, you know, everything I touch. Yeah, you've got to consider the security. I think some people have been a little scared about that.

[00:45:31] Speaker 2: Yeah. Yeah. And people who are in an organization, like in a company, there are almost always people in your IT department who are responsible for AI governance and setting those rules. So I would always defer you to them. They know the risks. They know the specific data they're trying to protect. But people as individuals or sole proprietors of their own business, you know, they've got to be careful and understand the security risks behind it. You know, do you want to give this tool all this permission to access all of your emails? It's an important question to consider each time you do that.

[00:46:12] Speaker 1: And I think what's interesting is we've worked with Zapier before, but they also, you know, offer MCP, which allows you to, you know, connect to all these, you know, different services. But they follow the approach where you can assign specific permissions as part of the MCP connection. So, you know, if you want to give it access to your Gmail, like, you know, maybe you give it read access, but you don't give it the ability to write an email. And so, you know, they give you granular control over, okay, I'm giving it access, but let me specify exactly what it has access to versus like, hey, you get everything. So, and I think I've seen Claude do that too, where, you know, there's some granular control. So, you know, you have a little bit more ability to set, you know, what its access is.

[00:46:57] Speaker 2: Yeah. And I think it's worth pointing out with Zapier, they also use a system of API keys. So it's different from MCP where each app developer can give another app a connection into it using basically software development APIs. And I'm not sure, maybe you do know, maybe Zapier does a combination of MCP and API connections, or maybe it's only API connections. But they were playing that game before MCP was on my radar. And Zapier has been doing that automation across multiple apps for a while. So it's definitely a good one to consider.

[00:47:35] Speaker 1: Yeah, what's interesting about them is like if you go into Claude and look at all their connectors, I think they have about 200 connectors. So if you find your service on there, great. But, you know, if you're trying to connect to some other tool that's not supported yet, then, you know, you're kind of out of luck. But with Zapier, I think they have like 8,000 plus integrations. So this way, like you said, they've been working on it, you know, for many, many years. So chances are, if you want to connect, say, Claude or ChatGPT to some other tool, you could do it. Yeah.

[00:48:06] Speaker 2: And I do also want to underscore what you're saying. After you've connected a MCP connector inside of Claude, absolutely. There are granular, you can say, allow, ask every time or allow every time on each and every action that an MCP connector can use. And some people would just say, allow all, you know, but you might want to take that moment to actually go through and say, I want to give it this permission, but not this permission. And you can do that granular control.

[00:48:35] Speaker 1: Yeah, no, that's a good point. So like in the case of, you know, maybe you give Claude permission to send an email, but you want to approve before it actually sends it out. So it would say, hey, can I send this? And you click yes, and then it does. Yeah. And Nick, I know kind of outside of, yeah, outside of MCP, you know, we've been working on a lot of videos. What are some of the kind of recent videos you've been working on that you're excited about?

[00:49:00] Speaker 2: Well, I did do, back to MCP, I did a video on Granola and also a video on Gamma, which is this awesome tool for making slideshow presentations. You know, way easier and way better looking than you would have in something like PowerPoint. But it was fun to use MCP connectors because you can build a Gamma's presentation that then pulls information from my email and from my documents as actual real information in these presentations it builds. And also last week we were talking about CoPilot CoWork. And so I have finished recording my CoPilot CoWork video. So new information about CoPilot CoWork now that it's available, kind of clarifying what that billing system is all about, comparing it to Cloud CoWork. And I think that will probably launch on Kevin's channel next week.

[00:49:53] Speaker 1: Nice. Yeah. Now, at least me personally, so two videos coming up soon, and they're actually focused on the video creation portion of kind of work. The first one is how to create AI videos using Comfy UI. And so I recently had a video go out on Comfy UI, and it's a free open source tool that allows you to run these different AI models on your computer. And as part of that, you could also generate AI video directly on your computer. And it doesn't cost anything aside from having a somewhat beefy computer to be able to run these and the electricity and all that. But that's kind of interesting where it makes AI video accessible to whoever wants to run it just at home. So that's one video that's coming up soon, and I'm excited about that one. And then the other one, also on the topic of video, I dug into Google Flow. I don't know if you've heard of that. But so recently, I think it was on our last call, we talked about some of the, or a few calls back, we talked about the different conferences going on. But Google released a new model called the Omni model. And with Google Flow, it uses this Omni model to generate video clips. But it's almost, it's kind of what an AI video creation suite could look like. So you go in there and you share your concept of what you're trying to create. And it'll pull together a storyboard, at first a text-based storyboard, then you get a visual storyboard. And then you can review that, you can give feedback. And then you can say, okay, let's proceed and let's create a video based on this. And then it generates all the individual clips that match the storyboard. So it really walks you through the creation process of making an AI video. And what's interesting then is you get all the clips, and then there's a kind of an AI-first editing interface where you could put together the timeline, you could move clips around, you can make edits using AI. So that was a really neat one to see, just how Google is envisioning kind of AI video creation kind of evolving. But yeah, that one should be landing soon. So this is part of the reason all these Apple devices are getting more expensive, to be able to use these different tools.

[00:52:05] Speaker 2: That's where the investment is going. Yeah.

[00:52:09] Speaker 1: But yeah, no, I think we, a lot of good videos in the pipeline. I think we have another Cloud Code video coming up that David's working on. And I think Elizabeth is looking at another chat GBT video. So there's a lot of good stuff coming soon. Indeed, yeah. And again, if you're watching this, we'd love to hear your feedback. What do you want to see in future episodes? What do you think of this format? Let us know right down below in the comments. You know, this is still a new series. And so we're looking to your feedback to help improve this and just kind of make it so it's more interesting for all of you tuning in. But with that, thanks for tuning in, and we hope to see you next week.