Do We Need the Internet Anymore?

On GEO Day 12 LinkedIn Live, Paris Childress and Simon Young ask whether humans still need websites when personal AI harnesses can research, book, and buy. They cover LinkedIn agent lockouts, Meta Muse's free-token push, Amazon blocking Muse, walled gardens versus a single chief-of-staff agent, Simon's Stitch WhatsApp compliance video hitting Google #1, and whether the open web is becoming a machine-readable library.

[00:00:00] We're live back again. How are you doing today, Simon? Busy as always. I presume you are as well. Um- Yeah, I have been busy

[00:00:10] the, the news keeps coming, doesn't it? Literally every, every second. Well, obviously being in AI search, you and I will see our feeds filled to the brim with all of the stuff that's coming through on a daily basis, but, um- Yeah ... keeping up with that seems to be hard as it is. Yeah, there's a lot, there's a lot going on.

[00:00:32] Um, Grok 4.7 just dropped, so if you liked Grok bot before when it was on 4.6- Oof ... I think it's getting, it's getting even better, and it's the same price. Oh, is it? Not more expensive, yeah. So if, uh... You actually do need to go into your cursor project settings and select Grok, uh, the latest model, 4.7, if you want to upgrade.

[00:00:57] You can ask Grok bot to walk you through [00:01:00] that. But, um, I can't say that I've done anything too difficult where I can assess whether it's much better than 4.6. I don't know yet. But according to some of the videos I've watched, um, my YouTube... By the way, the YouTube algorithm right now is maxing, maxing me out on Grok bot.

[00:01:18] I mean, it's all over my YouTube feed, and I can't resist. So, so 4.7 has dropped. Yeah, but it's getting busy. I think that there's this feeling that now that because I'm pretty much doing all of my work with AI, and it's so efficient, and I could, I could do, I don't know, five to 10 times more in a week than I used to be able to do.

[00:01:45] I, even though I, I should be enjoying the productivity, but I'm actually feeling even more frustrated sometimes when I don't get... 'Cause I, I give myself now extremely ambitious targets, like, um- Mm ... to do some [00:02:00] major things in, in a day that used to take a week or more, and if I don't get to them, I feel disappointed and frustrated.

[00:02:07] Um- Oh, I had, I had that feeling yesterday. Didn't, yeah- Yeah ... didn't quite achieve what I wanted to, but yeah, you've still done an incredible amount of work. Yeah, and we should put it into perspective. And anytime that the, a bot or, um, wh- whatever, anytime an agent is sitting idle, I'm, I'm very consciously aware of it now.

[00:02:29] Right now- ... Claude, Claude is not working on anything as far as I know. I think it's done. Grok bot is on hold waiting for me to, to give it some thumbnails, because I don't like the thumbnails it gave me, and I've gotta do a bit of work in ChatGPT images. And, and then of course, ChatGPT, Astra and ChatGPT is not...

[00:02:51] So all of my agents are sitting idle, and now I have Instinct as well, which is pretty cool. But everything is sitting idle, and I feel like I have a bunch [00:03:00] of people- ... that I've hired and that are on my payroll And they're all sitting around waiting for me to tell them what to do, and I- Uh, yeah, but you've, you've gotta remember they work through the night as well, you know, if you want them to.

[00:03:11] Yeah. So you can, you can put them on a second shift. Yeah, I'll do that. And I usually try to give them... I usually try to assign a big task before I, right before I leave the office And then I feel much better when I'm walking out of the door knowing that they're all, they're all working and they're all ramping up and getting started on, on something.

[00:03:33] I think one of, one of the huge things that we're all gonna come up against is, like, obviously you come up against consumption of token limits. And we're getting into a world where the optimization of how you use your credits across different engines is... That's definitely a thing, 'cause I'm constantly swapping across from Claude to give, you know, maybe you've harvested some data or you've done some research or you're...

[00:03:59] Well, my, [00:04:00] you know, my Claude at the moment is actually pulling the top 200, or what it considers to be the top 200 marketing directors that I'm connected to within my first-degree connections out of LinkedIn- Yeah ... who could well be interested in AEO, GEO. Um, it's then crafting personal messages to them.

[00:04:20] But I don't wanna do the outreach piece in Claude. I'm gonna do that through Grok. So- Yeah ... you know, but I am also gonna test, uh, voice messaging as a, an outreach piece, which I don't see happening an awful lot on LinkedIn. I don't know, do you get many voice messages coming to you? No. N- never. I used it, uh, years ago I tried doing that, and I think I had decent success where you'd...

[00:04:47] Yeah, you drop a little bit of a voice note. Yeah. And then they know. I think I did that pre-AI, but now doing it post-AI a- adds another element to it, which is proving that it's [00:05:00] authentic and it, that it's really you. It's not an AI agent, it's not a bot. Yeah, and there's definitely a- Definitely an

[00:05:06] added benefit ... I think there is definitely a pushback at the moment in terms of LinkedIn itself, um, stopping, you know, people using all these tools or trying to, but- Oh, yeah ... there's, there, there's no way. I don't see how if... You know, imagine if LinkedIn banned interaction with any of the, you know, any of the assistants.

[00:05:27] It would collapse maybe. Yeah. Well, I'm, I'm living through this right now, Simon. I... As we speak, we're streaming to LinkedIn right now, but I can't even access my own LinkedIn account right now, and I've been locked out since Sunday. And I got a message. I mean, I've, I've been locked out, and then they give me a time window when they, when they, uh, open it back up.

[00:05:50] But then I wake up the next day and I'm locked out again, and that's happened three days in a row. And the message that I'm getting is that there have been high... There was a... It was detected that there was a high [00:06:00] volume of LinkedIn profile data that I was accessing, and I know exactly what this was.

[00:06:05] Right. It's not Grok Bot, and it's not, um- any of the extensions that I have. I've tested different tools and extensions. I had something called Linked Nav, which is something I tested, and I left the extension sitting in there in Chrome. Yeah. What it was was it was ChatGPT Astra. It was a project... Because ChatGPT Astra is so good and so fast at browser use, I gave it a task to go and research all of the Brighton SEO speakers, 'cause, um- Mm-hmm

[00:06:36] um, as you know, I'll be speaking in Brighton SEO in a couple of weeks. And I said, "Go out and research all of their... I want, I wanna find, of course, uh, I want their LinkedIn profiles. I want some research, and I would like a, a whole outreach plan from you." But that's not, that's not hundreds and hundreds of people, is it?

[00:06:58] Maybe what- It was 128. [00:07:00] Yeah, so I would've thought- Yeah ... well, depends on the pace at which it went. I don't know. Speed. I think it was the speed, because it, it completed that task pretty quickly, in under an hour. I guess that's quickly, but it must have been going through... I couldn't see, I couldn't see what it was doing.

[00:07:17] I couldn't watch its screen, but it must have been cycling through these profiles and then doing something to extract data from the profiles. Maybe, maybe it was copy-pasting or doing something that detect, LinkedIn detected was bot activity, and I think that's what got me in trouble. But man, that's a scary, it's a scary- But these-

[00:07:36] moment when you, when you get locked out of LinkedIn, and, uh- A lot of those would've been second or third connections. Are you connected to all those people or not? No, majority I'm not. Second or third, yeah. Yeah, maybe that's an issue. Yeah. And, uh, the next step naturally is that I wanna start to reach out and make connections there.[00:08:00]

[00:08:00] And, you know, hopefully even, you know, line up some meetings, but, I mean, that's, that's a great audience for me to network with the other speakers, and I think the other speakers are probably thinking the same thing. Um- But you're only, you're only really allowed maybe two, 250 connection requests a week as it stands, right?

[00:08:21] Um, and- Is, is that the number? I, I, I don't know. I haven't- Yeah, it is ... I thought it was more like 100 a week. No, 200. I think 200 would be a good number if you're on, depending if you're on Sales Navigator. But if you're on Sales Nav, you're a paid user, and therefore- Mm ... you should be allowed to get away with a little bit more than you do if you don't, right?

[00:08:40] Yeah. Um, but- Yeah ... 100%, they don't allow the new or non-paid for accounts to do anything along those lines. Yeah. Ironically, today is my last day of LinkedIn Premium, and then I'm gonna let it expire, 'cause I, I don't know, I'm just not getting value, and I, I [00:09:00] generally think LinkedIn's Uh, products are overpriced as com- like, terribly overpriced It's horrendously expensive, isn't it, for Sales Navigator?

[00:09:09] Yeah. So what- Yeah ... I think here for me it's 95 quid a month, which- Yeah ... you know, the best part of 1,200 quid a year for, okay, you can get a discount if you buy annual, but when they start banning people or your account is unaccessible for, you know, a number of days or whatever, it's all a bit tight if, you know, if you're getting charged for that.

[00:09:34] Yeah. I, I've always, I, I've been in and out of LinkedIn Sales Nav many times over the years, on and off and on and off, and I can't really say that it's ever given me any real value. I mean, it's, it's kind of a, an attempt of a CRM inside of LinkedIn, but I've n- never really, uh, found it to be extremely valuable.

[00:09:55] And right now I'm not, uh, I'm off of it and I'm [00:10:00] gonna be letting LinkedIn Premium go. But, um, I just wonder if I would've gotten the same restriction if I had been a Sales Nav paying Sales Nav customer or not. Are you n- are you not a Sales Nav c- customer then? Not at the moment, no. Right. Well, I would imagine that would be the issue Yeah.

[00:10:21] Well, I think I agree with you that LinkedIn, LinkedIn is, is, is running up against a real problem. If they start... I think everybody's moving to personal agents now. I mean, the Grok Bot, Muse, uh, it, um, Instinct from, that I've, I've, I've checked out, and it, it's gonna be... OpenAI's gonna have one I'm sure soon.

[00:10:45] Um, Google has to have some, some answer to this that, that can be a, like an AI personal assistant over your whole Google Workspace account. And, and I'm sure Claude's gonna get in the game. But everybody more and more is gonna be using their AI [00:11:00] assistants to go and navigate the web, and also do some sort of curation or...

[00:11:06] I mean, people will still go to social media, but I think I would send my agent to social media first to say, "Look around. Tell me who's talking about these topics that I care about. Point me in the direction of conversations that I should get engaged with. Um, show me who I should be trying to connect with.

[00:11:23] Give me some suggested outreach messages." I'm gonna want my agent to manage my LinkedIn presence, and I'm gonna- But the problem- ... I, I have to send- Yeah, but the problem- I've gotta send them to LinkedIn ... the absolute problem then is, and just follow the logic, if we were all doing that, and excuse the language, surely the entirety of LinkedIn then just becomes a shit storm.

[00:11:47] Like, if it all- Yeah. 100% ... if it's agent to agent to everybody to everybody to talk to this to that, you know, like surely there's so many messages bouncing around that- Yeah ... maybe, maybe we're already at that. But that's... [00:12:00] Well, I mean, that's, that's naturally bringing us to the topic of this live discussion today, because you suggested do we even need websites anymore?

[00:12:09] And of course you have LinkedIn as a, as one example of this. We are moving towards agent to agent. I'm gonna, when I shop, when I book flight tickets, when I make a reservation, or even when I do a checkout, tools and whatever for business use, I'm gonna be more and more just sending my agents, "These are my needs.

[00:12:29] These are my requirements. I want something that does this. Go out, do the research for me, come back, make a recommendation." And then after that I might even tell you to go ahead and buy it or to go ahead and book it or to fill out the form, set up the call. Um, be, be my personal assistant. Be my, my chief of staff, my EA.

[00:12:48] My EA for work and my EA for life. And I think that this, we're right on the cusp of this wave. I, I feel it. I, I see it in the news. Meta Muse right now is the hottest AI [00:13:00] product. Meta s- uh, Meta stock jumped by like 17% overnight because of Meta Muse. They finally have, Meta finally has a legitimate AI story to tell, and Me- Meta Muse is a, is a massive hit.

[00:13:17] And it's for personal AI assistance. It's not in Europe yet. I, I guess we can't get our hands on it yet, but it's taking the US by s- firestorm right now. It's doing what Grok Bot did a few weeks ago. Grok Bot is still red hot too. Um, but this is gonna be... Everybody's gonna do this. Uh, OpenAI is gonna have one, Claude's gonna have one, Google's gonna have one, and then you're just gonna try to have to choose which one do you wanna go with, 'cause you shouldn't have, I don't think you should have multiple.

[00:13:46] But once everybody has their own personal AI agents- Why, why would I go to a website if, I mean, if I trust my agent and I trust them enough to even purchase [00:14:00] at my direction, purchase something, um, I don't need to, I don't need to go to that website, and I don't even need to go to LinkedIn. If I treat LinkedIn purely as a, a network place for networking, and I wanna be as efficient as possible and as productive as possible, then the dumbest way for me to spend my time is to go myself to my LinkedIn feed and start doom scrolling my, my LinkedIn feed.

[00:14:24] But that's what people- Even though it's a good feed ... are doing with websites, right? Like, why- Yeah ... why would you go and s- scrape the web yourself? Why would you go and spend time looking through 20 different websites to find out some information when you can just ask the question? Yeah. So yeah, I think genuinely, I do think, um, well, the internet is...

[00:14:46] I saw the inception of the internet. I don't know whether you're quite old enough to have seen. I am, yeah. Yeah. I saw it. Right. But, um, working through that, this feels, this moment feels very much [00:15:00] like that. Like, when things changed at such a great pace, and then, not obviously not at the pace back then, it didn't move anywhere near as quickly as it's moving now because it feels like day- Mm-hmm

[00:15:11] to day to day, it's the, the, you know, it's a tsunami of new opportunities, and the whole, um, space is changing. I don't know whether you saw Sam Altman. I tagged you in the other day to his interview that he gave at Sanford, and in front of a, a class where he's... The, the lecture, and they were talking about this being a generational...

[00:15:35] Well, I think he, he likened it to the invention of electricity. It was that important. Mm-hmm. This, what's happening to us now as a, as a world. Yeah. If you tagged me on LinkedIn, then I, i- if it was in the next, in the last few days, then then I missed it. I, I didn't, I didn't catch that. But I do agree that it's, it's on that scale.

[00:15:57] Electricity or the Industrial Revolution, [00:16:00] certainly the dawn of the internet, where suddenly with the whole, the whole world's knowledge was getting, was going online and becoming accessible to anyone with a connection. And in the beginning, these were really flaky dial-up connections. I, I remember that, where you would- Yeah

[00:16:19] I mean, I would, I would actually be online, and then if somebody picked u- if somebody called or if somebody started using the phone, then I would... It would intercept and block, and I would, I would... It would drop my connection. Well, it was like a filing cabinet, right? It was literally FTP sites where you placed up a bunch of text and documents and it...

[00:16:36] There was no real... There was... You could put a few images up and it was- Mm-hmm ... yeah, it was fun, right? To get involved, and then it very quickly turned into, "Well, actually, we can get this found by, you know, meta tagging it, and putting it in the right places, and linking it," and all the good old- Yeah ... SEO stuff that we...

[00:16:52] And the Black Hat stuff. Yeah, well then, yeah, search engines came, came right on the heels of the, the, the dawn of the internet, and if it [00:17:00] wasn't for search engines, I, I think the internet just wouldn't be usable. A- as a large directory, the way Yahoo did it in the beginning with all the d- the categories and sub-categories and sub-sub-categories- And Lycos-

[00:17:11] it would be- ... and Ask Jeeves and all of those guys. Yeah, yeah. But search, search definitely, um, made the internet accessible. But we're... So I think what, what I, what I'm experiencing is not necessarily the dawn of the internet, but the dawn of search as a layer over the internet, which actually opened up the internet and made the internet accessible.

[00:17:33] And now what we're finding is that, I would say the equivalent is that AI is, is the internet, and search, the equivalent of search is, is these AI assistants, the harnesses so to speak. Because the Groq bot, Muse, th- they're harnesses actually. Th- they're not the underlying models themselves. Yeah. And the harnesses are what is now opening up all the [00:18:00] opportunity.

[00:18:00] The, uh, the ability to chat and, um, in the case of Groq, you know, to have a, a whole team with cl- defined roles that you can give to different bots who can do work 24/7 on their own computers. Um, with, with Muse it seems to be going more of the consumer route where people are, um, using it to go out and shop and order things and book reservations or book, uh, book travel, things like that.

[00:18:29] Interestingly, um, Amazon blocked Muse yesterday. Blocked it? That's the big news. Yeah, Amazon blocked Muse because it was not playing by its rules of engagement and it was not, uh, clearly identifying itself properly or something. But you can imagine, uh, how many people were sending their Muse agents to, to go to buy things on Amazon.

[00:18:53] I mean, it's, uh, natural. So- Yeah, find- I guess- ... find me the cheapest, find me the [00:19:00] cheapest and quickest to get it here, all of that Yeah, but Amazon must be really shooting themselves in the foot here, because how many of those purchases are incremental that would not have happened otherwise by the human directly going to Amazon to take the time to find the item and do the shopping and bu- and, and check out.

[00:19:19] But if you go to Muse and say, "I'm in the market for new headphones. This is what I'm looking for. This is the price range. Go out and find me the best deal," or even go out on Amazon and just monitor the best deals. I don't need to buy right away. But when you spot a deal, could be three days from now- Yeah

[00:19:38] if it's something, if, if you, if you spot a great deal, just buy it immediately. And that, I don't know, Amazon must be... I think, I think they're throwing the baby out with the bathwater, as they say. I, I don't, I don't think it was... It may not even be the fact that, yes, I could set it up to look for deals and, you know, I could get discounts and I can, you know, find out more [00:20:00] information.

[00:20:00] I think more, more so maybe it's Jeff is sat there going, "Well, I don't want, um, Mark scraping all my data." Like, I don't want to let the, you know, the wolf into the henhouse, so to speak, you know, to come into the back of the system. I guess so But think of what that can do for incremental sales. How many of those purchases would have happened anyway from the human going on their own to buy?

[00:20:29] Or how many of them are more spontaneous or, or that, you know, I'm, I'm doing this because now Muse is doing it for me, but I otherwise wouldn't do it myself. Which is- Mm ... a lot of, a lot of my AI use cases is, is doing things that I didn't do before and that I would never have done before AI because it takes too long and it's too much of a headache.

[00:20:51] So I think Amazon is, uh... This is costing them, uh, a lot of, uh, incremental revenue if you're, if you're blocking [00:21:00] these Muse agents who are coming to Amazon to buy things for their human masters, and you're saying, "No, get out. You can't, you can't come in." You're not- But, but if you're, if you're Amazon, surely you say, "Well, we should have our own AI, and you should use our AI for that purpose."

[00:21:17] I guess, yeah. The- And we're big enough ... they had something called Rufus, I think. Yeah, Rufus. Yeah. But that never took off. Yeah. No, but we're b- we're big enough and, you know, we're big enough and ug- ugly enough that we should have our own AI. And if you wanna use an AI to run through the Amazon system, Rufus can help- Mm-hmm

[00:21:31] you do all of that, right? So Rufus and Muse then therefore need to- Yeah ... be friends and... Well, side story here. I think Amazon, uh, f- just s- f- fumbled in the worst way. I mean, this is gonna be, this is gonna go down as one of the, the greatest missed opportunities. Amazon Alexa should have been the first voice AI assistant, and it should've been good.

[00:21:57] Just... This is, this goes in the category with Siri, Apple [00:22:00] Siri. Apple Siri, Amazon Alexa, both completely spoiled, ruined, missed opportunity. They had these devices that were, that were good at voice, and, um, what Muse is doing now, that, that, that should be happening with Amazon Alexa and with Siri, where you, uh, you can have a conversation about anything.

[00:22:22] You can say, "Go out, go do this on my calendar, book this, check out this, make sure we have, uh, g- restock us with toilet paper and get it ordered for tomorrow delivery." And Amazon should know I, you know, I always get this type, I always... And there's enough of the recurring purchases where, uh, I'm, I'm sure a lot of groceries are like this, a lot of household items.

[00:22:44] Yeah, Amazon was, uh, Alexa was supposed to be, to be that, but it never panned out. It did Well, that's, that, I was talk- I was... That's my original thesis back in 2019, talking about answer engines. And- Yeah ... would we then quickly move [00:23:00] into a world of, right, we will just speak what we want. And I don't think I ever imagined we'd get to where we are now.

[00:23:07] Now I think- Mm ... you can, if you could imagine where we can be in a year, or imagine where we'll be in 10 years. What on earth will be going on then? It's all gonna be agent to agent. I think in, within a year, the agent layer is, it's gonna be clearly in place. The, um, the whole web, to bring it back to our topic, humans will very rarely visit websites because we won't have a need, we won't have a need to.

[00:23:36] We'll- So what about- By the way, Simon, you're, you're frozen. Are you still there? It looks like- Oh, can you, can you hear me? I can hear you, but the video looks a little bit choppy. I'm not sure. Oh. I... It looks okay for me, but- Okay. Um- We still have the voice going, so we're, we're good. Yeah. But imagine a scenario then if you take that next step where I have [00:24:00] my own business and my personal profile within my own agent.

[00:24:05] So rather than having a webpage, I have, I don't know what we're g- you're gonna define that as, then my knowledge base and my offer, or whatever you wanna call it, uh, within my agent, which then sits there as a resource, as maybe part of the internet, um, if that's what you wanna still call it. Um, but the fundamental problem as I see it at the moment, and I don't know how you, you feel it's going in terms of, um, a lot of these agents are now trying to ring-fence themselves and keep their own knowledge bases.

[00:24:39] It, and it, it feels like we- we're sort of splitting out. You know, like you've already seen that you don't see YouTube videos in ChatGPT, and Muse will pull its own data from its own source, and you'll have, you know- Mm-hmm ... Google now trying to stop certain other AIs coming in and scraping data from [00:25:00] it. Um, whereas fundamentally, Google is not the internet.

[00:25:04] I think Google thinks it owns the internet, right? Or is search. Whereas there's- Yeah. Yeah ... so many... You know, that's... I think we're entering an age where it's all gonna suddenly splinter out into trying to- No, that's not good ... different- All these, the walled gardens, huh? Yeah. That, that's a bad thing. That would be a, a bad development as opposed to the, the open internet.

[00:25:28] I mean, Google was always supporting an open internet and without any walled gardens, and it could because it, it had that vantage point of sitting at the top of the mountain and being able to see everything th- with Search. But yeah, imagine if I had to have, uh, one agent for Amazon because Amazon wouldn't let any of the other agents in, and then I needed another agent for LinkedIn, let's say, because LinkedIn, uh, is gonna decide, "I'm gonna block all these bots, but we...

[00:25:56] Because people are demanding to use bots, uh, we'll [00:26:00] create our own bot and we'll give a, we'll give that to people." Yeah. But I need my LinkedIn, I need my LinkedIn agent to go out and do my LinkedIn work. I need my, uh, my Google agent to operate in my Google workspace with, with my emails, chats, calendar, G Drive, and then maybe, I don't know, what else?

[00:26:21] Um- I don't, I don't want... I wouldn't wanna live in that kind of world. Oh, I think you don't. I, I think I wanna have just one agent. But that's where... That is definitely, I think, where we're heading. It feels like it. Um- Mm-hmm ... Sa- Sam Altman, and the irony within that guy's name, I don't know, you know, like Altman at the end of it, that maybe- Yeah

[00:26:41] that's just a, a freak of nature. He was, you know... But then maybe he can just phone up Elon and they can work together, 'cause they look or feel very similar, um, to me anyway. That's just a personal obs- observation. But if we- Mm-hmm ... we end up in a world where he was talking about, uh, token prices, [00:27:00] and talking about the cost to run your life or whatever, all these agents you want to undertake all these different tasks for you.

[00:27:10] Mm-hmm. The way it looks, you could get to a scenario where it becomes extremely expensive within certain models to run what you wanna run. Yeah. Now, he then started talking about the next development will be people running or trying to create agents which are affordable for the mass, you know, the mass users, rather than the enterprise user who's prepared to pay the- Yeah

[00:27:34] $1,000 a month to run everything on autopilot. Where does it sit with Joe Public, who, you know, wants to do a certain level, but should be ma- Is that a mobile phone subscription rate of 40 bucks a month, and it'll do almost- Mm. Yeah ... I don't know. This, this is what fascinates me. Yeah. And what's gonna, what's gonna democratize this.

[00:27:57] I can tell you that Meta's Muse is [00:28:00] free, entirely free, and from what I've heard, it has extremely generous token limits, and that's because Meta has the advertising cash cow that can subsidize that. I think there's only really two companies that can do that in, in this race, uh, the AI frontier race, and it's Meta and Google, because they're both such crazy profitable businesses with their advertising.

[00:28:27] Well, Google's revenue now is more, even more diversified. Meta i- has a cash cow advertising business which allows it to then offer free, free tokens. So I think Meta is making a very aggressive play now, making this free, m- giving very generous token limits, and getting, getting lock-in, getting... I, I, I wouldn't be surprised if there are a billion users on Meta Muse by the end of this year, really.

[00:28:56] Because Meta has two billion, or more than [00:29:00] two billion. I, I don't even know what the number is. I guess if you count WhatsApp. Yeah. They have more than two billion global users. Could they get half of their global user base onto their leading AI product, onto Muse? And it's free, completely free, and they're gonna cross-sell the hell out of this inside of WhatsApp, inside of Facebook, Instagram.

[00:29:20] So- They, they could get to a billion users with a free product with very, very, uh, generous token limits. What's that gonna do to, to the Grok bot and the other products that are coming? And that's gonna put a lot of downward pressure on prices, and I think that's gonna democratize it. Um, not to mention the o- the open source models are still there, and the gap between where the best open source models are and the best closed source, I think that gap is shrinking.

[00:29:50] I mean, it's- But then the, the open so- will the open source models have the compute power that... That's the difference, isn't it? Yeah. Yeah. [00:30:00] Right. But, um, we'll see. I guess, uh, I think Meta, uh, Meta's making a really interesting play. I mean, they were a laughing stock until very recently with, with AI. Well, they were, they were a laughing stock over trying to, trying to run a metaverse.

[00:30:17] How much money did they pump- Oh, yeah ... into a metaverse- Oh, yeah ... which nobody used. Oh, that was a, that was a big mistake. But even their early... They stumbled out of the gates a lot with AI, and, um, but now I think they're, they're firmly in the, in the game. But you saw, did you see- I would have guessed, if you, if you would've asked me three months ago, "Who has the, uh, which, which company has the, have the two best, hottest products in AI right now?"

[00:30:45] And you would've told me... If, if, if you would've asked me that in, say, April, May, and you, and you would say, "Well, by Oc- by September or October, those two companies are, are not gonna be OpenAI, not gonna be Claude, not gonna be Google." [00:31:00] Yeah. No, you, you would say no chance. I'd be shocked. It, it's X, it's SpaceX AI, and it's Meta with the two, with the two best, two hottest AI products because of the harness, because it, it's an assistant that is proactive and, uh, it actually allows you to really give it work, offload work.

[00:31:18] Instead of just going task by task and coming in and saying, "I have this task. Here's the context. Here's the attachments. Here's the, here's the long prompt that I have to come up with." That's a, that's a bit of a cognitive load every time you go to Claude. I have to think about it. Every Claude conversation is a mini project, you know?

[00:31:35] But man, with Grok Bot, and especially now with voice, I just hit the voice and I just say, "All right, Sandy, uh, I just recorded episode 41 today. Um, let's get to work." She knows exactly, and then there's a nine-step process. I don't have to re-explain it. I don't have to upload the, the Descript MP- MP3 file. I don't, because they, she's gonna go to the browser [00:32:00] and, and do everything.

[00:32:00] Actually she's gonna, she's gonna, um, delegate that to Connie, who's my content lead. And, um- I like the way you've named them all. My, mine are, mine haven't got names. Yeah. I'm, I'm trying to, yeah, make it more, more fun. But Sandy is my chief of staff. Connie is my, my content lead. And 80% of my work is with those two now, and, um, I'm, I'm taking incremental steps 'cause I, I wanna, I wanna really nail this podcast workflow.

[00:32:31] And I'm still tweaking a lot of things. Like the YouTube thumbnails are not coming out the way I like right now. And- You should, you should let me, let me build you a studio like I've got on Lovable. Yeah. Like, because- That's your offering. I'm, I'll, I'll use that then ... then... Yeah, yeah, yeah. I'll, I'll do it.

[00:32:46] And then what you can do is just tell your bots to go and go to the studio and grab all the collateral from there. Um, and- Yeah ... but the other thing- A creative studio ... yeah, the other thing with the studio was all the images that are produced are from [00:33:00] vectors, so they don't come out as AI, which is, I think, maybe important going forwards.

[00:33:05] Mm-hmm. But we'll see. Did you see- Yeah ... did you see Google starting to pay people for, uh, contributions to the knowledge base? Yeah. Yeah, you sent me that. What does that mean? How, how much, how much are they paying, and for what exactly? Well, no one knows what they're paying yet, but that feels like, um... So they're saying they've signed contributions with major publishers, whatever that means.

[00:33:29] So the archives of- Mm-hmm ... you know, imagine- New York Times and stuff ... the, the World Library or whatever it is, I don't know. They've got all these publishers, and they've signed up all their data into that, so we can use whatever your content is in our model. Um- Mm-hmm ... but how on earth... Well, at the moment, you don't get access, right?

[00:33:48] Unless you're invited. Whereas, you know- Mm-hmm ... if you or I were invited, we would... And if there was significant revenue for contribution, and your- Mm-hmm ... your work is being judged as [00:34:00] filling gaps, people will... Maybe that is a... Well, that is obviously intelligence, and knowledge is a currency now. Yeah. So that then becomes what's that worth, and which model wants to buy it?

[00:34:15] Yeah. Do you know these companies, um, I think the leading company's called Scale AI. Let me just check this out. And they basically hire professionals to train AI. Scale AI. Um, is that the name? Yeah. Yeah, based in annotation large language model. Is this one of these adverts that I saw where it's... And it's very well paid, where they're looking for people.

[00:34:42] It is, it's extremely well paid. They're hiring experts, they're hiring PhDs across all fields, and they're paying them by the hour, and it's... I, I don't know how much, but it is, uh, it's very high paid. They're paying them by the hour to train, to train the AI, basically, to impart their [00:35:00] knowledge. And I don't know how that's done.

[00:35:01] It could be through, maybe it could be through interviews, it could be through written content, it might be through some video sh- sh- uh, shared screen activity. But whatever it is, the goal is to get the expert knowledge out of these PhDs' heads and give it to AI for training purposes. And, uh, uh, that sounds like what you just told me with...

[00:35:21] Is Google's doing some form of that. They're basically, um- There, there are other things for But it's supposed to, it's supposed to pop up in your search console as a, you know, as an offer within the search console to say, you know, "Can we train AI using your knowledge?" And at that point, I don't know what you're gonna get paid, but- Yeah.

[00:35:43] Yeah, that would be interesting Ja

[00:35:51] Scale AI But this, it, it does... That's another confirmation of what's making me feel like we're at a point where [00:36:00] each of the models is throwing up their walls and saying, "Well, we'll keep..." You know, let's say, say I put my knowledge... I don't- I'm not saying I'm that valuable, but I put my knowledge up for sale, and they're like, "Right, okay.

[00:36:10] Come and join our club," or, "You come and..." Like, well, what do I sign away? My publishing rights for the next 20 years, or infinitum? I don't know. Like- Yeah ... the mo- model buys the person. Yeah. I think it better to be a free agent and you can get paid from all, all of them. Well, or they... Or, or then you work with the open source models, and you're- Yeah.

[00:36:36] Yeah But this, this might ruin the whole movement of an AI assistant if you have all these walled gardens, and you would need multiple AI assistants to, um, to do different things. I mean, you'd need an Amazon AI assistant just for your Amazon shopping. You'd need Meta Muse for, uh, I guess social media, but that really is, let's just say Facebook, and Facebook [00:37:00] Marketplace, and Instagram.

[00:37:02] And then you'd need another one for, for your work life, maybe Google or if Microsoft. I don't know. I, I think once you get even to two or definitely three, it's, it's too many. Too many- Yeah, and then you do, you get to, well, this is an, this is an aggregator, or this is a facilitator, and this one works with all the models, and that talks to that, and it does all that for you.

[00:37:23] And there's another- Yeah ... layer on top. Oh, you mean there's a layer that would be your, your real chief of staff who's not affiliated with any of these walled gardens- Yeah ... but they can, they can go to my Amazon AI assistant and say, uh, "Go, and Paris wants to buy this, this, and this. Go and do it." They can go to my social media AI assistant and say, "He, he, uh, he wants to do this kind of research in LinkedIn, and he wants to engage in this way on Twitter or whatever," scraping.

[00:37:51] But, uh, I don't know. That's a... Then you have, then you have two layers. You have a... No, a whole nother layer, which I don't know [00:38:00] To me that would, that would kind of spoil it, I think. But what, the one thing that you said that was really important earlier, um, was Google was supposed to be, you know, we are not going to wall or gate, and we are the gateway to everything.

[00:38:16] And- Yeah ... that is definitely not feeling like it is the case at the moment. It feels like they've changed tact. Yeah. And they are, they are 100% search changed very recently because the evidence, you know, and I know I showed you to it. Mm-hmm. The, the video that I ranked for Stitch, we put that video up last Friday live during the session.

[00:38:42] Well, I put it up about an hour before the session. Mm. We looked at it during the session, and that video now ranks number one on Google Right? Yeah. And, and, and I tested it with a friend in the US yesterday to make sure it wasn't me sat here in the UK being the person who [00:39:00] published the video and etc, etc.

[00:39:01] A friend in the US searched- Mm-hmm ... um, WhatsApp regulations on bulk messaging, uh, in the UK, and he was sat in the US, and that video came up number one. And it... I then set Gr- I set Grok to work and said, "Okay, I can see that video is coming up for this search term. Can you, can you find some search terms and phrases that are around that and are relevant, um, and- Mm-hmm

[00:39:29] go and test for me whether the video shows for those as well?" And it does, and it came back- Oh, really? ... saying that... Oh, yeah. So this was really interesting to say not just for the phrase that we optimized it for, because, um, typically if you were gonna rank something, be it a webpage, in Google, you would, you would se- you would use a key phrase or a, or a keyword, right?

[00:39:54] Well, the question that we put up, and it was quite a long [00:40:00] tail- Mm-hmm ... it then related. It was coming back for WhatsApp and, uh, PECR re- regulation or P-E-C-R regulation in the UK. It was coming back for all sorts of terms. Mm-hmm. So not just the title, not just what was spoken in the start of it, stuff that- And these were terms that you hadn't even put into the description of the YouTube video?

[00:40:21] No. No. Wow. So it was, it was assigning relevance to the content- Mm-hmm ... and expanding that, whatever you would call, key chain of knowledge to go into and say, "Right, this is the most relevant piece of content, and it's filling the gap." Yeah. But- Now you've gotta get... go and g- get, find a way to get a lot of people searching for that topic.

[00:40:44] Yeah. Maybe have it, I don't know, eh, like, have some kind of traditional PR event that sparks a lot of interest. But, but you say that- Or- ... and this is, this is one thing that I've always tried to stress to advertisers. I used to have loads of people come to me [00:41:00] wanting to run YouTube ads, and it was all based- Mm

[00:41:03] off of keywords and phrases or, you know, everything around that. Now everyone was like, "We wanna get the most views as possible," but you... but the relevancy was everything. So if you ran a campaign and only the people who you, who were perfect, right, landed on the video- You, let's say you got 200 perfect people, that would convert obviously much better, but it would also, um, mean that the engagement was 10 times higher.

[00:41:34] And the people that found the content, they found it far more valuable, right? Great signal. Mm-hmm. Whereas if you made those keywords and phrases a little bit more broad and you had a lot of people land, you had t- five, 10,000 people landing on there, you might have, follow me on this, but you might have 3 or 400 good people in that.

[00:41:54] You, therefore you've got- Mm-hmm ... more potential customers, but you've also got [00:42:00] 4,700 people who are n- not as relevant and are gonna land and they're gonna go, "Mm, no," right? Mm-hmm. Or they're gonna bounce off the video or the piece of content within the first 30 seconds, giving a bad signal, i.e. if it was a webpage, a bounce, or a video, a very short watch or low engagement.

[00:42:16] And then that- Mm-hmm ... content then dropped off. So the question still exists, if it's very, very relevant, and I believe personally if it's low volume people but extremely valuma- va- valuable and relevant, that is what I think gives you the citation rather than something that just gets shared and then the, the signal gets diluted.

[00:42:39] Are you with me? Yeah. So it's, it's better actually to have less reach, less sheer audience reach, but have a stronger signal with better engagement than it is to have a, a bigger reach. Yeah. Yeah. Sa- satisfy the need of the person that was searching- Mm-hmm ... and then prove the engagement through the f- the fact that [00:43:00] they watch it or use it or it then, that would say to me, an AI would go, "Right, this piece, Simon has found this piece really useful.

[00:43:09] He's watched all of it. He's in, he, it, it matches what he searched. He watched it all. He's used it. He's digested it. What did he do?" So the next piece then is what did that person then do next? And that's the really interesting thing in where we're going in answer engines and LLMs is, what do I ask next?

[00:43:26] What do I then want? What do I... 'Cause that's what it wants to learn, isn't it? Mm-hmm. It wants to learn how we then go to, well, what did I start, what intent did I start out with? Did I learn what I needed? And did it give me- Mm-hmm ... the result I wanted in as short a time as possible? Mm-hmm. Yeah. I guess this is probably why Y- YouTube videos are appearing so much in AI Overviews.

[00:43:50] Yeah. Because then Google can see if people are... It must mean that people are clicking into those YouTube videos to watch and to supplement [00:44:00] their answer with that video view. But then also YouTube can observe in its, within its own ecosystem. Maybe they, they, they go to YouTube, and what is the next video that they watch after that?

[00:44:11] Or what is... Do they make another search in YouTube, or do they start watching similar related videos in YouTube? And m- maybe AI Overviews is almost a sneaky way for Google to just funnel more YouTube engagement, just to get more YouTube watch hours. Uh, so the quest- the question then becomes, so typically, uh, Facebook, right, Meta or whatever, static.

[00:44:38] A lot of it, there's not tons of video in it. Agreed? Mm-hmm. Right? Yeah, yeah. If you say the next step where we're gonna eventually get to is voice, right? Everything that goes on YouTube has got voice in it, right? Unless it's a blank screen with a, you know, a Well, music video. Well- I mean, it's maybe music videos or [00:45:00] things, but- Yeah, not everything, but there is-

[00:45:01] large majority ... well, there are, there is some specialist content on there- There's- ... that is not- Well, there's always audio. Yeah, there's al- always audio, and probably 90% of it there's human voice. Right. In Google. So if we're moving to, to a voice... Google said back in 2018 it was gonna make the entirety of, um, Google video.

[00:45:21] Mm-hmm. I think they said that would happen by 2020. Somebody can probably pull me up on it, and the dates might be slightly wrong, but, um, they did make a statement they were gonna try and push everything into video. That may well be where we're going next, um, because would you rather watch something? Now, if you take the analogy of all the kids, let's not go...

[00:45:42] Okay, the, the, the younger generation are all sat there doom scrolling through their phones, watching videos after video- Yeah ... after video. That's all they look at. Mm-hmm. Um, that's... They- these social networks know that's the most engaging content, right? It's [00:46:00] far more- Mm ... engaging than anything that's static.

[00:46:03] We don't wanna read- They're definitely not clicking, they're not clicking into, uh, into websites and reading blog posts, for sure. No chance. 100% they're not clicking on websites- No ... are they? No. I'd, I'd love to know the stats on, you know, do, do people under 20 spend any time on websites? Nah, I, I bet you it's near zero considering that all their internet time is on their phone and that's spent in messaging apps and social media apps and- Yeah

[00:46:32] maybe gaming, uh, gaming as well. Yeah. But not... I, I would even, I would even ask the question, what is the percentage of time in the Chrome browser at all for a person under 20 or 25? No. And, and obviously Google knows the exact answer to the question we've just- Mm-hmm ... asked, right? Yeah. So they know. At least for Android phones, uh, they, they know the answer to that.

[00:46:56] Yeah. Yeah. So they, they know what [00:47:00] they're trying to target, and it's not me and you, right? No. It's that, that new generation that's coming up. Yeah. All right, Simon, I've got to, I've gotta wrap it up now 'cause I've gotta prep for a call coming up soon, but, um, it's been a, a pleasure as always. And, uh, any final thoughts about does the, does the open web have a, have a future at all?

[00:47:27] Are you optimistic? Are you writing, writing, uh, the obituary already?

[00:47:35] So that's a really difficult question. And I know people have called me out on LinkedIn saying that I must be a crackpot for saying the internet could disappear. I don't think it disappears. It becomes... I think it becomes a library. You know, like it- It just goes underground. Yeah, yeah. It becomes a big database, but it's not a database that human people would go and see.

[00:47:55] It would be like if I went into, um... Well, yeah, it would be [00:48:00] like if I go, go into a library to start browsing books or into a card catalog or something, and then that's a real old example. Um- Yeah, but if I- A, a repository, let's call it that. Yeah, yeah. Or it could, I guess it could be like a, a, it's just gonna be a big storage space, and people could still tour it like they tour museums, and there's gonna be some nice beautiful websites that people might look at as forms of art and say, "This is beautiful.

[00:48:30] What a, what a, what a great piece of art this is." So, so here's- But- ... here's, here's the worrying thought though, right? Mm-hmm. Where is the internet stored as far as you're concerned? Where do you think it is or it lives? I mean, it's, it's across millions of servers all over the world, so it's distributed- Right

[00:48:48] of course. Who control, who controls the majority of that? A business called Amazon. Oh, with AWS? Yeah. Mm-hmm. Yeah, the actual- If you switch- if Amazon switched off [00:49:00] AWS, right- Mm ... then what would hap- the internet would collapse, right? Yeah. So maybe Jeff's got a play left where he owns all the data. Just- Oh, I don't know

[00:49:10] that might be the thing. Yeah, he, they do have some serious leverage by having AWS, and so s- yeah, just being the hosting provider for I don't know what percentage of the whole internet maybe. It's a, gotta be a big chunk- I reckon it's a big chunk ... it's hosted on. And, and that's- Yeah ... the play, right? That's why when Amazon, whenever there's a service outage at, at Amazon, at AWS, what, most of the internet goes down.

[00:49:35] I mean, uh, like, uh- Yeah, yeah, yeah. So- And it's- ... maybe we've discovered the master plan. Yeah, yeah. Well, good luck, good luck to them 'cause they, they definitely fumbled the, fumbled the ball with, uh, Alexa. All right. Well, Simon, I've gotta run, but it was great, uh, great talking to you, and, uh, we'll talk again on Friday, huh?

[00:49:57] Yeah. See you Friday. See you [00:50:00] later.

Do We Need the Internet Anymore?

Key takeaways

In this episode

Grok 4.7 and agent-idle anxiety

LinkedIn lockout from agent scraping

TOP: Do we need websites anymore?

Meta Muse and the harness wave

Amazon blocks Muse; Alexa / Siri missed bets

Walled gardens vs one chief of staff

Stitch video #1 and relevance over reach

Close: open web as repository

Full transcript

See how AI answers describe your brand

platform

Resources

About us

use cases

compare us