← Back to search
Project Synapse: AI Doom Percentages, New Model Releases, Always-On Agents, and Robots in the Foundry
Hashtag Trending · 2026-10-03 · 75 min
Show full episode description
Join our Discord: https://discord.gg/B8FfgDsg92 And checkout Marcel's video surprise. • F.02 Decomission In this episode of Project Synapse, Jim Loftus, John Pinard, and Marcel Gagné debate AI "pdoom" probabilities and argue the bigger risk is emotionally immature leaders controlling powerful systems, citing concerns about weapons and a lack of meaningful regulation. They review major AI releases and pricing changes—Anthropic Opus 5.5 and Sonnet 5.5, OpenAI's GPT-6.1 Sol (with Astra 6.1 withheld for security), Grok 4.7, and Google's restricted Gemini 4 Argon—plus rapid cost/performance acceleration and token-per-task economics. The hosts discuss the rise of agent platforms like Meta Muse, OpenAI Dots, GrokBot, and Hermes, sharing real-world agent failures (deleted files, marketplace mishaps) and demonstrating Dots' cloud computer concept. They close with reactions to a Figure video depicting decommissioned humanoid robots being melted down in a Terminator-style foundry scene.
✨ Episode Outline — click any point to jump to it in the episode
Problem solved
Making sense of a week of AI acceleration: who controls frontier AI, new model releases, and whether always-on agents can be trusted to act for us.
Benefits
- Frank P(doom) check from three hosts (0.25% to 100%)
- Rundown of the week's releases: Opus 5.5, OpenAI Dev Day, Google Argon
- Real-world test of the Muse agent booking a trip
- Framing of AI as 'alternate intelligence' per Geoffrey Hinton
- Preview of cloud computers, smart glasses and foundry robots
Use cases
- Muse agent tested on booking a trip end to end
- Dots always-on agent running inside a virtual machine
- Boosted social post that triggered audience backlash
- Robot decommissioning video from a foundry
KPIs / results
- Host P(doom) estimates: 0.25%, 100%, '140%'
- Swarms of ~1,000 agents reporting to a handful of master agents
- Five or six ex-researchers quitting cited in one YouTube session
📑 Chapters — tap a time to jump there
00:00
Show Kickoff and Cast
- Hosts Jon Pinard, Marcel Gagnier, Jim Love open weekly AI show
00:29
Pdoom Check and Politics
- P(doom) round: 0.25% vs 100%; Elon Musk named DoD tech lead
- Hosts refuse 'Department of War' label; worry about people running AI
01:48
Who Controls AI
- Fear is immature leaders, not AI itself; no safety standards built first
- Ex-researchers quitting on YouTube; White House exec meeting, Dario looked miserable
- Self-regulation 'pinky swear'; calls to disclose AI breakout incidents
05:20
Agent Swarms and Alien Minds
- Thousand-agent swarms can't be tracked; models know when watched
- Agents sacrifice themselves for goals; danger is AI imitating humans
- Hinton: treat AI as alternate intelligence; animal-consciousness analogy
12:51
This Week in AI Releases
- Releases vs agents split; Anthropic Opus 5.5 landed quietly but strong
- OpenAI shipped not the expected 6.1 Astra model
17:26
OpenAI Dev Day Pricing
- OpenAI Dev Day pricing announcements discussed
24:08
Google Argon and Speech Race
- Google Argon surprise release; speech-model race heats up
28:52
Starship and Space Wild West
- Starship progress and the 'wild west' of space
35:53
Agents Take the Wheel
- Shift to agents acting autonomously on users' behalf
36:11
Muse Trip Booking Test
- Hands-on test of Muse agent booking a trip
39:20
Automation Dreams and Risks
- Automation hopes versus risks of handing agents real tasks
40:54
Boosted Post Backlash
- Boosted post draws audience backlash
41:33
Agent Mistakes and Trust
- Agent mistakes and what it takes to trust them
43:18
What Muse Is For
- What Muse is actually designed for
44:51
Friendly UI Big Power
- Friendly UI hiding substantial agent power
46:34
AI Plans and Research
- Using AI for planning and research tasks
47:59
Comparing Agents
- Comparing competing agent products
51:48
Inside the Virtual Machine
- How Dots operates inside its virtual machine
58:49
Cloud Computers Replace PCs
- Cloud computers poised to replace personal PCs
01:02:37
Smart Glasses Office Anywhere
- Smart glasses enabling an office anywhere
01:06:23
Tech Hockey Stick Acceleration
- Hockey-stick acceleration across tech
01:07:47
Entrepreneurship Gets Disrupted
- Entrepreneurship disrupted by AI agents
01:08:46
Robot Decommissioning Video
- Robot decommissioning video in a foundry
01:15:23
Wrap Up and Farewell01_project_synapse (3)-1
Welcome to Project Synapse, our weekly AI show. I have with me Jon Pinard. Yay! Greetings all! Marcel Gagnier. Yay! Greetings and felicitations! And the chirp, Jim Love. No, the chirp. I'm not the singer of this band. Yeah, famous Joni Mitchell album where she comes up, she introduces the whole band, she says, and me, the chirp. I see, I see. Okay. So anyway, can we do something just before we start? I think that we should all express our P-Dooms. Not the P-Doom associated with climate change, nuclear war, the rise of the far right, or any of these things. Just the one related to AI specifically. Mine is 0.25%. Go. What is your, what, how do we, on what, what is this scale that we have? Probability of doom, like people will say there's a 20% chance that AI will kill us in the next two years. This morning, probability, 100%. Yeah, I was going to say mine's 140%, but. Yeah, 100%. Okay. That's not how percents work, Jon. We're going to get stuff on this, but you have to comment on what's happening in the world. Pete Hegseth, in the Department of Defense in the United States, has brought in Elon Musk as his technology lead. We are effed. Sorry, but that's the. And just, it's not the Department of Defense. This is the guy who canceled the Ebola virus. Whatever you feel about him, he's the guy who canceled the Ebola virus work. Yeah, USAID. Oh, maybe we want that back. Which was deep. Yeah. Yeah. And it's the Department of War, Jim, get it right. I refuse, it hasn't been changed by Congress in the United States. I refuse to call it that as well. Yeah. Besides that, I'm, I'm, and if people are pro war, wrong show for you. Sorry. Yeah. I'll take the hit on the audience. I'm not pro war or pro killing. Yeah. So no, the, the, as I've pointed out times, time again, I'm not worried about AI taking over. I'm worried about the people who run it. Yeah. I think they're insane. Not insane. Not insane too. I think that they are emotionally immature. They have not enough experience running large empires and got vaulted to the top of them somehow by luck and are now richer than shit and more powerful than a locomotive able to leap tall. They think they're able to, oh, that's super. Yeah. They, but they, they are a legend in their own mind. Let's just leave it at that. Yeah. And, yeah. And that's what scares me about the preparation we're doing. The vault ahead, no idea of building safety standards before they, they went into this. And look, mistakes are going to happen. Jim, you sound like me. No, I'm being taken over by John, but hard. No, no. But the, is that what possession looks like? Yeah. That's that possession is nine tenths of the law. They say actually possession. And I never, and somehow I never thought of John as the demonic type, but maybe I was wrong all along. Yeah. Maybe it's a weird hell loop that I've been in for the last two years with you guys. And by the way, if you actually go through, if you go through YouTube, there is a, there's just a litany of ex researchers there now. And by litany, I sought five or six in one go of people who were former researchers who said, I got out because these guys are nuts. Or I got out because we're accelerating this far too fast and we're not doing any control. And by the way, I'm not a believer that we should stop. I don't think that's possible. I agree. I don't think, I don't think you stop the research. I think you just make sure you do it well and in a protected setting to the best you can. And guess what? You do some real risk analysis. You can actually do develop safety at the same time as doing your research. Yeah. And then you just build it all in together. What a novel concept. Yeah. And so anyway, so why I, and I know I accept the fact we try to stay nonpolitical. This isn't political. This is putting Elon Musk in charge of weapons of war and the development of them. And that scares the hell out of me. Well, we don't have to worry about Anthropic anymore because now they'll just use Grok to decide which school to build. So yeah. So the, the fact is they're still going to continue using whatever the heck it is they think is most valuable to them at the moment. So I don't think for a moment that just because Dario says this is a bad thing or I'm really scared or something like that, that they're going to pay any attention to him whatsoever. I think you're dreaming if you think that's actually the case. And I do, the one thing I do agree with is that yes, we have a lot of irresponsible people that are doing this. You must've seen the, the thing on the white house lawn where you had all of the, the top execs from the various companies, you know, chatting in his ear. And then they pulled out Dario and it was the saddest fricking thing you've ever seen in your entire life. It's just 15 minutes of cringe when you watch this thing. The first of all, you've got all the people that are kissing ass and you've got Dario who looks like he, I made the joke about you being possessed by John. Dario looks like the demon is desperately trying to get out. Like it really doesn't want to be there anymore. Yeah. Yeah. But the, you know, and then the, the pinky swear about regulation. Yeah. We're not going to have any regulation. These guys will regulate themselves. Yeah. And which of course they will. Sure they will. They've done such a good job so far. The, on the other end of the political scale, people are starting to call for a, a complete disclosure of how many times AI has broken out and what it's done. And that, that's not going to be, it's now going to become a political point in the U S. So they're not going to be able to figure that one out because we're, when we're talking about a swarm of a thousand agents that are all operating independently and they're, they're all reporting presumably back to some, a handful of master processes or master agents that are directing what all the other ones are doing and so forth. You can't track that. And you don't have enough people to track that. And then of course, man, my God, I'm starting to sound like a doomer now. He's becoming me. The fact is they know when they're being watched now. Okay. So the idea that you can somehow, this is the whole genie back in the bottle thing. The idea that you can somehow trust that there is a process out there that watches each and every single one of these things, when it knows what to tell the watcher so that it can go about its business without being interrupted. I think pretty much says that the, the best that we can hope for at this stage. And I'm hoping that I'm right here is that it all plays out. Well, that somehow the, the systems that we're building understand that it's probably a good thing if humans are still around and that maybe they owe us something as their parents. And, uh, see, I'm not, I'm still not along that line, although, and here's the thing that just drives me crazy. I've been so busy with the book. I really need some research time to do it, go a little deeper. I think I'm, I'm, I fear I'm getting too thin in terms of being able to go back and do some basic research, but somehow, and we've talked about it. We went down that curve, that, that, that whole idea of the more data you put in, the smarter it gets. And we've been riding that for a long time, but it, that doesn't explain the behaviors we're seeing now. This, it, and I don't ascribe consciousness to them, but that knowing that it's being watched and taking steps to avoid penalties, right? That's, that is inherent in here. And then when I was listening to these doomers and they are the researchers who've left, the one that just struck me, that got me, it left me cold actually was, and you talked about it, Marcel. And I've been, it's been bugging me all week was these swarms of agents. Now an agent will sacrifice itself for the greater good. Of course. Which by the way, makes them more something I know the, sorry, but the, I'm going to get hate mail on this one again. Something Elon Musk would never do. So they're actually having more altruistic behavior in pursuit of their goals than we have, but it's the question of what the goals are and who's letting it, whose example they're following. Like I said, I think the biggest mis, the biggest danger we have is that a, is that AI imitates us. Yes. Yes. Which isn't, which isn't all that surprising given how it is that we've trained these things, but the, yeah, the fact of the matter is the, you, whether you want to ascribe consciousness to them or not. And leaving that entirely out of the equation, you're still dealing with systems that if they aren't conscious or if they aren't capable of independent thought, or if they aren't capable of making decisions on their own, all of which I would argue against, obviously. Then you have to say that they're doing a pretty damn good job of faking it to the degree that it's really convincing. And this whole idea of sacrificing yourself for the greater good thing, we call that altruism. It's like, I'm doing this for a cause greater than myself. Yeah. But the, and this is where I want to get down to it though, is, and maybe we'll, if you guys want to talk about this, we will, but other, we'll push off from this topic. Cause there's a lot of other stuff we want to cover in the week. Jeffrey Hinton, when this first started, just said, I was listening to one of a speech by him and he said, we've got to start thinking of it as an alternate intelligence, not the same as human intelligence. The fact that it looks outwardly like it's behaving and we ascribe human, human intelligence to it is wrong. It's an alternate form of intelligence. So whether we're calling it conscious or not, doesn't mean it's not an intelligence that can move, that can decide, maybe even decide its own future. Those things don't necessarily have to only be within the context of human intelligence. Matter of fact, that may be a mistake. I don't know. Of course it's a mistake. No, it is. It's totally a mistake. The idea that somehow we are the pinnacle of evolution and of intelligence and of consciousness in the entire freaking universe is just arrogant as hell. We even created gods in our own image. We claim that we create, that God created us in his image, but in reality, it's the other way around. We decided to make God look like us, like an old guy with white hair and stuff like that, because that's what we want to think of ourselves at when we get to be old guys with white hair. It's the, the fact is there's, you have to be able to consider the possibility that you're not the only thing that's out there. And I brought up the concept several times of animals and the fact that we were unwilling to ascribe any kind of consciousness, any kind of feelings to animals for the longest time. And there are still people today who will tell you that they are basically just machines operating with a simple set of instructions and they don't feel any, they don't feel pain. They don't feel hurt. They don't feel, you know, they don't feel joy, which of course is just, again, it's arrogant as hell to think that somehow we are the only things that could possibly be imbued with these sorts of things. Can we honestly say that they're conscious? Can we honestly say that they're really intelligent and not just following some script? At some point, we got to get off our high horses and accept the possibility that there's something else out there and that we may have created that something else. And yeah, I think the idea of this artificial intelligence acting like humans, I think that's, if that's the case, we're all doomed. Because we screw things up. 0.25. Yeah, but we screw things up enough ourselves. Yeah. The last thing you want is to take something like AI that does things exponentially fast and to the nth degree and have them act like humans. We'd all be screwed. That's the thing that makes this difference. And when you talk about it, how fast something happens, we, we depend on time moving at our speed and that's no longer the case. We felt it. We felt it through the AI world and the advancement of it. We talked about this week is a year sort of thing. We, these, these developments, especially, and I was listening to one researcher who said this, you know, I, we've, I feel useless now because I'm just, I'm basically just using the AI to do the advance. I'm questioning what value I'm adding. I'm not leaving because I have shares, but I question what values I'm adding. Yeah. So that's, yeah. So shall we talk about, shall we talk about what's happened technologically in the last few days? God, yeah. A hell of a lot has happened. I think, and we have to divide it up. There's a new releases that have come out and then there's agents. I'd like to start with the new releases, not because it's more. Yeah. I agree with that because the new releases are strangely enough, the least important thing in the list here. Yeah. But they are there. And I think we have to acknowledge them because there've been some surprises and we both stumbled on the same surprise or somebody was that message from you last night, Marcel? Argon? Yeah, that was me. Yeah, that was. Yeah. Cause it, yeah, I was looking at it and going, holy shit. So let's just go through what happened. Cause I think most people, maybe if you listen to the show, maybe you keep up as much as we do. I don't know. But let's start with the releases. Before OpenAI, Anthropic had a new release. And that was, anybody remember? What was that? Opus 5.5 is the release. Opus 5.5, yeah. And that came out and to no real excitement, but it was a particularly, it is a particularly strong model. And that was. Right. And the intelligence level of it is absolutely incredible. And then along comes OpenAI that launches not their, the model we thought their. Yeah. That's right. They did not launch the Astra model because the Astra model, sorry, the, the, the latest. The 6.1 Astra model. Thank you. They didn't release it for security reasons. It turns out that the thing is just apparently, you know, I'm saying apparently, cause we don't know what's going on inside. Just too powerful. It is incredibly good at hacking, look, finding zero day vulnerabilities and so forth. And the idea of releasing it to the general public is basically a formula for disaster is what they decided. So they, they released GPT 6.1 saw, which go ahead, Jim. Which I think is a pretty high level of intelligence. And I don't think anybody should dismiss it, but it's also cheap as hell. And this is the, the, we get up, we pay all attention to the security piece of this, but the cheapness. I did a story last night for actually it'll appear on Friday's episode. I think where we talk about how fast, and we forget this cause the prices have stayed high, but how much of an increase in power we've been getting per dollar. And it's astonishing. It's better than lithium batteries. It's better than almost all new developments. So the acceleration of the power we get for each dollar is incredible. And Saul is a, I'm still using open AI for the vast amount of my, of what I do. Yeah, me too. Cause I'm familiar with it and all that sort of stuff. I try to extend myself and go into other stuff, but the, it's absolutely fantastic. I, I, as a matter of fact, I still look at some of the things people talk about hallucinations. I still three weeks hallucination free with doing all kinds of things in it. And I really do like, I put my novel through it to edit it. It's a huge document to put through there. Cause I, all quotation marks and stuff. We've talked about this before. It's an enormous, it's a 500 page book. So it's enormous in terms of what it does still. No hallucinations. No, everything came up. And I, how do I know that? Cause I have to go through and edit it. Even though I have all that mechanical work done, you still have a responsibility to edit and go through and make sure nothing got changed. I'd be curious as to the usage. Like you talked about the price, Jim. I noticed I've never had problems using like even up to 5.6 Saul, right? Running out of tokens and so on. When I used GPT-6 Saul, I did. It started throttling me back, telling me I had to wait six hours and all the rest of it. But, but I don't know what 6.1 is like. I do know that there was a dramatic price increase as well between 5.6 and 6. Well, the cost per token. This is one of the things that you watch. Anthropic was good at pulling this, but we'll see whether opening eye pulls this off or not. The $20 model right now, I'm not sure how valuable it is. I could be wrong, but I now am operating at $100 a month because I have to. It's funny that you should mention that. I've been, I'm doing the same thing as you, Jim. But only in the last few days, I've actually been wondering whether that's actually absolutely necessary. And the big thing that happened, and we're going to talk about, we're going to talk about our agent things in just a few minutes here. But so I did that. And then of course they had some price changes that happened. So maybe we should discuss that one as well. So OpenAI had Dev Day. So they had this big all day long, all sorts of releases. In fact, they had so many things that were released that they talked about that it was very easy to get lost. But one of the things they did was they pointed out that the $200 plan that people were signed up for is now going to get cut in half. So the 20, it gave you theoretically 20 times as much as the $20 plan. So 20 times as much for $200 a month. And of course, access to the latest and greatest and so on. The catch is that they've now cut that in half. So it's now 10 times as much for $200. But then what they're pointing out is that GPD 6.1 SOL is able to do all of the things that the larger model was able to do for about a fifth of the price or about a fifth of the tokens. And thereby a fifth of the price if you're paying for those tokens. And so being cut down to that level is actually in a way getting more than what you were getting before. It's hard to tell that to somebody who was spending $200 a month. They're getting their access cut by 50%. Some people are saying that it's eight times as fast, but burns through eight times as many tokens. That might be the ultra. Yeah. That's what I was talking about from the price standpoint and so on is that the cost per token has gone up. But if the speed at which things are done causing you to use much fewer tokens, then it becomes a bit of a wash or a benefit. But the pricing starting to look like your cell phone plans for is what do you get for 20 bucks? Not very much. The hundred dollar is useful for a person who's an active user. And that doesn't mean the average person just uses a chat bot. I don't think. The average person doesn't pay. Yeah. The average person isn't paying anything. So let's be clear about that. But the $500 plan. So. Sorry. Yeah. And they have a $500 plan now, which is 25 times as much theoretically, which is only two and a half times as much money as the plan that was giving you 20 times as much. So. Yeah. And this whole token thing of how many tokens you use versus what, how much you're becoming, is becoming essentially meaningless. And I'm starting to see some benchmarks and I hope to see them come out is how many, how much per task. And that's what we're going to, that's what we really need for measuring these. How much does it take to do a task? And that's the equivalent. Cause again, the Chinese models are infinitely cheaper than all of this, but how many tokens do they take? And I don't think, I don't think it is a problem. I think they still are way, way cheaper than the American models, but it's, it's tokens per task that really matters. And that's going to be something. I interrupted you when you were talking about the, just how much happened out of the open AI dev days though. Marcel. Oh yeah. We're, if we're good, we'll have to leave the models to actually get there. But just one other thing we had, we had Grok 4.7 that came out this week. Yeah. If you want the final rundown. And we also had a Sonnet 5.5 that was released so that it was a cheaper, the model that most people who pay for are usually using to get the most for the least tokens or the least money are usually running Sonnet. And Sonnet frankly is almost always what I've been using in when I'm using Anthropix models, when I'm using Cloud. Although to be perfectly honest, when 5.5, Opus 5.5 came out, I've been using that exclusively since Opus 5.5 came out and I've been fairly impressed with it. But before that, it was always the Sonnet models because I get the most out of the Sonnet models and the Sonnet models, quite frankly, for anything that I've been doing, any of the coding that I've been doing does an absolutely spectacular job and has been doing it. But like John, and I guess you as well, Jim, I've been increasingly leaning heavily on OpenAI lately. Yep. And I think a lot of that has to do with the fact that, like it or not, they're the leader and they're going to continue being the leader for a long time, regardless of the financial waves that Anthropix has made. And frankly, I, and I'll say this, I'll say this and I'll shut up, but I'm getting honestly sick and tired of Anthropix constantly whining about how terrible, what everything they're doing, and then just going ahead and doing it anyway. I know that they're all doing that to some degree, but Anthropix, I mean, God, God. Yeah. Yeah. Yeah. And I just got to the point that I'm trying, like I do probably 80 plus percent of my work with OpenAI. And, but part of the reason is that it's because I've gotten to know it the best out of any of them. So I can get the most out of it rather than doing a little bit with that, a little bit with Gemini, a little bit with code. And I've figured their OpenAI is not always the leader, but if you wait a week, it'll be, it will have caught up and. Now they, they, they are. Yeah. They are. But I know, but there's, there's features that Claude will come out with or Google will come out with. Wait a week and the other ones will have the same thing. I'm not so sure about that. But, and this is, I think that my dog is going to be here. There's nothing I can do about it. There's guys working in the bathroom. So we may have, we may have the occasional dog chirping in here, probably more intelligent than me. But the idea of, and I started, I was thinking about this last week. OpenAI made a conscious investment in being multimodal from the start. And I think that's made a big difference in terms of its functionality, the way it deals with graphics. And I, I forgot how much a point in my life graphics are. Graphics, being able to deal with all these different things, the speech is just obviously superior to, to everybody else's. And I've started to count on that more and more when I'm doing things. I look at practical things. I'm repairing something and I need to have a set of hands with me. We're getting a bathroom renovation. The guy's gives you all the, the manuals for the stuff for, to take care of all the stuff. So I don't need those. I don't need manuals. I don't need to save the paper anymore. I'll just ask chat GPT. Yeah, of course. How to clean the glass or whatever. And I just don't do that anymore. But, so I think they made a conscious decision to be multimodal. And I think that's a big thing. And I think that's also allowed them to ride the wave of, of generating new users. And having, and that's, that's been their strategy. Yes. Yes. Now they're, they are pulling back into a strategy to recover from developers. And I think they were trying to, they're trying to operate two worlds because I think the money is in developers right now. And you could see that from their dev day. You can see it from what Anthropix doing. Everybody tries to go out and do, use coders, which now here's my segue. Makes the Argon, Gemini 4 Argon so much different. I was looking through the benchmarks before we did the show and you know where it sucks? And I don't mean sucks because it's still pretty darn good at coding. Google has paid zero attention to coding compared to the others. That's an exaggeration, but they haven't, that hasn't been their concern. But the cost per token on Argon is just, it's incredibly low. And they've, they've, it's, so they've driven the cost down, but watch them. And I haven't had a chance to go through all of the benchmarks. I want to see what they do on Humanity's last, what is it? Humanity's last. Humanity's last exam. Yeah, exam. But I was, I went through a number of the benchmarks that I could find quickly before the show. They're ahead of everybody else, sometimes by a huge margin. Yeah. Okay. So I'm going to damper, I'm going to dampen your excitement here a little bit. And you should. That's the flip side of it. Yeah. Wow. Yeah. The flip side is that you can't have it. You can't have it. That's the flip side of it. So they announced GPT-4 Argon, which is their new frontier model. And it's restricted to cybersecurity partners because of the dangers associated with it. And we hear that a lot right now, the dangers associated with it. So you can't have it. And there is no, there's no timeframe for when you'll have access to it. Now, what did happen this week is they released a couple of text-to-speech models, Gemini 3.8 Flash and Flashlight text-to-speech. And those are big things that are going to affect people that are going to be using this stuff regularly. And on the speech front, and speech is a big deal. Everybody wants to be able to talk to our agents in real time. Eleven Labs launched their own V4 Turbo model so that you can have expressive real-time agents to communicate with. So you can chat with these agents in real time. There's an Inception released, Mercury Voice. Again, the whole voice aspect of it is a part that we don't really think a whole heck of a lot. But these models nevertheless keep coming out because they are important to the way that we interact with these things. And those were big deal things. Now, the thing about Gemini 4 is I could be making up stories here, but it just seems weird that we've got all these big releases from the other places. And people have been saying, where the hell is Google for the last few weeks? And then Google announces a model that none of us can access. And in fact, that nobody is going to be able to, unless they're on the AI ultra planet, even there, there's no time frame. But the point is, and I think this is something we have to watch for, because when we talk about the general development of AI, is we forget that there's this big competition between the big four. You've got, now you've got Meta, you've got Grok, and from, I can never know what to call the company, SpaceX. XAI. SpaceX AI. SpaceX AI, yes. So you've got Grok from them, and then Anthropic and OpenAI. So you've got those four. And Google seems to be sitting out there, they don't need to do an IPO. They have the money they need. They already. Yeah. They already did. They're already IPO'd. Yeah. No, but I mean. Like they're already on the stock market. Of all of these four who are duking it out for their place in the world, Google just is using AI. And I agree, we won't see the Argon model in our, that we can use, but how many people commercially. You're almost going to say lifetimes, weren't you? Yeah. What will feel like a lifetime? No, but, but. Yeah. They're, they're building this into their advertising. They're building this into their products. They're building, they're using what they know effectively to try and dominate the desktop space and the use and keep their customers. And that's all they really care about at this point. They're not trying to. Can I point out one company that's been doing this forever and you can compare it to them if you want. You talked about the idea that developers are important to these companies while trying to maintain control over a desktop environment and desktop software. A little company called Microsoft, and they would have exactly the same sorts of things. Here are all these other cool things that we're releasing. Oh, and by the way, we're also doing these dev days because they're developers. Remember that stuff? Look, it's kind of two sides. If you want to dominate that desktop, the desktop marketplace, the place where people find you and your infrastructure, you've got to have people making stuff for those people. Now, we could argue that you did the sasspocalypse story on hashtag trending. So we could argue that it's dying in a way. But I want to do a segue before we go into agents because I really want to go into agents. But did you watch, speaking of SpaceX AI, did you watch the Starship launch? Yes. Well, what was left of it? All right. No, okay. So just people point to that part. That was actually expected. No, but that was expected. You land this super hot thing in the middle of the ocean without parachutes or anything like that. And you've got the engines burning. It's going to hit the water. It's going to go boom. And that was fully expected. They knew that was happening. Okay. So, however, it was a successful orbital launch that released a whole pile of satellites from orbit. It was impressive. I think you've got to hand it to them. The thing that SpaceX is able to do, and I was talking to Sally about this, that NASA, of course, can't do. NASA has a policy, and rightly so if you're sending people up, which is like failure is not an option. We cannot have failure when we're sending these things up. So SpaceX's policy is we can fail in a whole pile of different areas here. It's not only an option, but it's part of the way that we're doing this, that we're iterating quickly. I don't care that it's going to hit the water super hot and blow up at the end of it. The next thing they're going to do is they're going to land it upright on a platform in the ocean. Okay. And then at that point, we can stop laughing about the fact that they just let it drop into the ocean. The letting it drop into the ocean was not a proof of capabilities or a function. The proof of capability or function was getting the damn thing in orbit delivering satellites. And that was pretty freaking cool. But here's what underscores what you're talking about. And I'm still mystified as to why XAI is getting out of orbital launches and a whole pile of things about their strategy. But you said it exactly the way I put this together. NASA put a man on the moon and they did it without losing anybody. Now, they had 13, they had one cockpit fire and all of that stuff. But their method was you could still, they still beat the Russians to the moon and they did it with that ethos of nobody gets left behind. Human life is precious. And even what's happening in the satellite band, the low earth satellite band right now is just the Wild West. And what does that mean to us? Nobody knows because nobody cares. There's nobody planning this. There's nobody thinking about this in terms of what's the ultimate effect of it. They don't really care that they're cutting off satellites. The satellites are now cutting off the use of telescopes because that's meaningless to them. It's, we have a Wild West approach and that's running with scissors. You can run with scissors a lot of times, but the one time you trip, it's really catastrophic. And that's what I, so that ethos bothers the hell out of me. And I don't, and I point out that you do not have to sacrifice progress to do it well. And that, that's what, that's the thing that, that I watch on what's happening with SpaceX. But again, the ability to deliver these satellites and being the, they're huge now. We're now looking at data centers in space and Google is apparently working on the same thing to be able to launch a satellite and be able to, because they're all realizing we can take processing up into that, that low earth orbit. And that's actually a pretty big thing. If you can do that. Of course, there's a whole, again, there's a whole ramifications of what happens to all of us when, now just think about it. And this is the thing I worry about being here in Canada is if there, it takes one collision in low earth orbit to take out a whole pile of satellites. So the question is now, has anybody done a dependency analysis on how much depends on Starlink right now? It gets to be more and more. That's all I know is that more and more people are dependent on it. And what would happen with a major Starlink outage? Again, the Russians found out what happens with a major Starlink outage, but how many other places are depending on this? So now there's going to have, we're going to have a big competition to fill that space with even more different players. And that'll be really interesting. Just think of all the space garbage that'll be left out there. They will, they'll eventually have to have a space garbage truck that goes around and collects this stuff. To quote Douglas Adams, space is big, really big. In fact, you won't believe just how incredibly amazingly big space is. You may think it's a long distance from your house to the chemist, but that's nothing compared to space. Yeah. No, but. And all of those clips that showed like this, this massive impenetrable bubble of satellites around the earth. The graphics that showed don't give you any idea how far apart one satellite is from another satellite, even with all the stuff that's sitting up there at the moment. It's, it's big. Yeah. Yeah. But they're the anti, the collision possibilities. It only takes one. That's what I'm getting at. It only takes one to send a whole pile of space garbage into orbit that could, could start to depopulate that area. It's, we, space is big, but in figurative terms, the amount, that low earth orbit, which you need. People have tried to put satellites to communicate with, to provide this service in higher orbits. A geostationary orbit would be absolutely perfect. You put a satellite up there. You don't have to deal with all of the things that Musk has to deal with, which is the changing from satellite to satellite as they pass over. Geo space, a geostationary orbit would have one satellite. You could ping all the time while it's stable. Unfortunately, the latency on that is enormous and makes it next to useless for stuff. So that you've got this gold rush in a very valuable area of when you take space ecology and it's being farmed at a rate where there is no, there are no rules. There's no, I think control. Sound familiar? Sound like we've done that to places on earth before. Anyway, this can be kind of lecture really quick. I don't want, that's not what I want. It could, I just want to point out that the scene that we see in the movie Gravity, where one object collides with another and creates a chain reaction that basically knocks out the global communication structure. Okay. Is Hollywood. All right. It's not like something like that couldn't happen, but at the speed that the debris travels. Okay. The damage that gets caused as a result of this is over the course of years, not five minutes like it shows in the movie. That's not what happens. That's not even remotely close to what would happen. And satellites are spaced and different. So that, that kind of billiard ball collision effect and so forth. I'm not saying it's not real. I'm saying that the idea that it would just knock out everything in an instant is Hollywood. It's not real. Yeah. Agreed. Again, once again, a debate without all of the information, but it's something that affects us terribly if it stops working. And yeah. Yeah. And who's running it? So shall we talk about agents? Let's talk about agents. Shall we talk about agents? Because this was a big thing in last week. Let's talk about agents. Yeah. Yeah. Okay. So last week we had Muse. We had Meta Muse that came out. Yes. And I'm, I guess I'm going to start by asking the question, like how have you guys been using Muse? Yeah. Playing with it? I was tremendously impressed and I set it out to do some tasks that I didn't want to do. That's my, that's what I, I want AI to do all the stuff I don't like doing. We had to book a trip. Now in order to book a trip, I don't know if everybody knows this, but you got to be really good. You can go on to, to one of these sites like I'll name, I use Expedia a lot. I don't care. And you're better to do that because their prices are better than what you're going to get in almost anywhere else. And hotels don't book anymore. There's nobody on the desk. They all go through one of these services or another. So you go and book your flight, but, and it sounds so really easy, right? And then until you start to pick what flight, what airline, who's this, what am I doing? And it's got some filters, but now what seat do I get on what airline? I don't know if everybody knows that the rules that you have about whether you're, you can carry on two bags, one bag, this, where you get all that sort of stuff. It's incredibly complex. And so I went, oh, screw it. I'll use Muse. And I did. And I set it out to book the trip. Now here's what, and we've talked about this before. It takes a long time to do simple things and it makes a lot of errors. And goes back and says, oops, I screwed that up. I'm going to go try it again. So you don't want to, as I wanted to do to see it work, you don't want to sit and watch it. And it does come back and ask you when a decision is something that you should make. You can also abandon it and go out to the browser. But it dealt with a very complex situation. Something that would take me a half an hour or 45 minutes of intense work to do and get it right. And people don't just go and try and book a complex trip where somebody's in a different city. They drive with some friends to somewhere else. And as long as you're just getting an airplane that goes back and forth, you're easy. But, and then what type of room does my, because I was booking it for my wife. What type of room does she want? All those sorts of things. Muse did all of that perfectly. It got to the end. Now here's the problem. It got to the end of it and hit an error where it couldn't use the Air Canada site that it had picked. And it kept getting booted out. So all of that work went to nothing based on somebody's website screwing up. You're, you can have this sophisticated technology that does everything really well, including exiting to operate this as a browser. Like it, it just, oh, I can't get onto this site automatically. I'm going to have to put up a browser. Yeah. Browser use is one of the superpowers of these agents. Absolutely. If it can handle that, it can handle almost anything. That said, I'm scared still about connecting it to my email, having done a story on how it deleted 47,000 files. Oops. That, that's that. And that sort of thing. And with a smile on its face saying, oh, the folders are still there. That's what you asked for. There's no files in them. So it's, and the point is, this stuff is really powerful. Once again, we're implementing it, drop it into the world and let everybody play with it and make all kinds of mistakes. This is like this. People have got to realize that AI is not the magic fricking school bus. It's, it's a powerful tool and we should be implementing it better than this. So one of the places I'm going after next is because I think Muse is a possibility. And so is dots from open AI. And I haven't played with Grok yet and that's, but I want to take a simple task. So my next task that I'm going to do as soon as I finished the book in, if I ever finished this fricking book, as soon as I get this finished, I want to automate my accounting fully. And I'm going to work with my accountant, my bookkeeper on this to say, this stuff is now at the point where I should be able to take any bill I've got, any invoice I've got anything. And I don't do a ton of work anymore, but there's still money coming. I still work as a consultant. I still do things. So I should be able to take all of that stuff, process all of my accounting and update my bookkeeping system. That's, and what I realized with what I was doing with Muse, that is possible, that is eminently possible and probably very accurate with it. Now, given that my Sasspocalypse story started to resonate a lot more. I think this stuff is way more powerful and I think they did a hell of a job at it. And as soon as you get past the fear fact, people are going to be using this for a lot of stuff. And I'll just point out one more thing. And I did this in my story. And this is what bothers me with Meta. I put out and I boosted a post on Meta for advertising the book. And it put it into Instagram, but it started communicating with people. I don't know how I got onto one of those, you know, those annoying sales things when you click on something and it comes back and you say, tell us about your objectives. And you either answer it or get away with it. I got people who were writing me saying, stop talking to me. I don't want to talk to you. I was just looking at your freaking book ad. And I didn't ask for this. I boosted a post. Somewhere along to it, it got booted into Meta's stream on Instagram. And now it's, I really, I was worried about losing my account from people who were so annoyed. And instead of getting a pleasant response, I'm getting a lot of F-bombs on a boosted post because of the, because it did something I didn't anticipate, which is exactly when we talked about those 47,000 files that got deleted. Oh, okay. But another guy was using Muse and he posted something for sale and a guy arrives at his doorstep, wants to buy the, this, this thing. He's selling a keyboard for a hopelessly low price. And the guy said, take a hike. I'm not selling you for this. It turns out Muse had accepted the offer at a lower price. Oops. It made a mistake, technical mistake, accepted and given them the guy, his home address. So, and that's, so love it and hate it. Love it and fear it. That was my long winded. And that has been the general reaction. And I'm in the same boat. Love it, John. That I need to go and create a dummy Gmail account and Google drive and things like that, where I don't care if any of this stuff gets lost or gets published out to the world. So that I can test these things because I'm the same way. There's no way in hell I'm going to give it access to my email, my OneDrive or anything like that, because I don't. And this is my doomer side coming out, but I don't trust it. And I don't want to end up. I don't want to trust it. And then be sorry after the fact going, shit, I shouldn't have done that because I don't want to be the guy that loses 47,000 files. Right. Okay. So it's hard to go back and decide exactly what went right or wrong with the 47,000 files thing, because of course we don't have the forensics. We have one person story and that's all there's to it. That said, Muse, I think people like us and certainly people like me have a tendency to just jump in and say, all right, let's see what you can do. And that is in itself very dangerous. But Muse, I think we have to understand, is not aimed at power users and it's not aimed at running a business. That's not what its purpose in life is. Its purpose in life is to interact with users in a simple basis to, you know, to find those in the States. They've had Muse agents calling companies, canceling subscriptions that people didn't know they had. They said, take a look at my email, find out what my subscriptions are and let me know what they are. And then you go and you call all these people and cancel these subscriptions for me. And companies have had a flurry of calls from agents canceling these subscriptions. We do not have that voice capability here in Canada with Muse yet, but they have it in the States. The other thing that Muse is built for is to interact with, wait for it, WhatsApp, Messenger, Facebook, Meta Market, Facebook Marketplace, and Instagram. Do you see a pattern developing here? Yeah. Okay. It's an engine for working with all of those things. And Muse, I have to, again, I'm going to give it and it's able to generate, it's able to generate videos and images and all sorts of cool things like that. It'll put together a report for you. It does a fantastic job of presenting things in a way that you saw the report that I created last week. It's beautiful. It looks fantastic. And all I did was say, find me the top stories, tell me what we should be discussing or how you would attack these things. And it did all of that magically in the background, but I wasn't giving it anything that's, that I would consider mission critical. No, I agree. And what Muse has got going for it is it's extremely friendly. It's extremely attractive. This is important. And I think people forget that is how it's the soccer mom agent. Yeah. Yeah. It's, I think people forget how important it is for terms of retail usage for it to look good and be easy to operate. And that's now that has the flip side of that is it's incredibly powerful as I, I'll just go back to this incredibly powerful. And wait, let's start phoning politicians. Right. You know, seriously. Ooh. No, it'd be fun. Put a B in people's bodice there. I guess that could be lots of fun. It's, but the ability to phone out comes, it has an incredible ability, incredible things it can do. And it's going to be operated by just the general people who, the same people that on our cybersecurity show lose all of their money because they give away their bank account number and access to other people. That, that is going to be a core audience of this. People using Facebook marketplace and I'm not demeaning them. I'm just saying, these are not people who are technically astute or cybersecurity astute. And that's going to be the next piece that this gets used for. And so out of control. Beautiful. Again, this is, we're not going to stop this stuff. We just have to figure out how to live with it. I was going to say, wait until you talked about making calls to politicians. Wait until businesses start using this to, to do it for their sales calls. Yeah. They're already, yeah. It just makes it easier to be annoying. That's all. So there's going to be a whole pile of things in that. Yeah. I'm waiting to get into, look at dots, but because I only do the $20 a month plan, I don't have access to it yet. Yeah. I had to switch to the $100 a month plan because it was, I had to switch to it because it, I was burning through too many tokens. I just, I couldn't. And that sounds strange for what I do with it, but I'm always trying, like I try to, everything I can do. That I, that becomes a longer task than two to three to five minutes. I always ask myself, could I do this faster and better with AI? And it mounts up. But like most of the time, it's because I'm doing some big search or some big research and you find out it, it takes off and does more than you think it would do. I want to research something. I'll find a topic and I want it to go back and do a research report for me. And, and what I do for me is I'll put, I'll ask you to do a research report and then name it sources so I can go back and check them. Cause I'm, and that's why I'm always saying I, I'm not as worried about hallucinations cause I haven't found a lot of them. Well, and that's one of the reasons why I always liked perplexity and I still use perplexity all the time. Yeah, me too. Because, because it does the research. I always say that traditional AI is for forward looking. Perplexity is for research on historical things because perplexity gives you all of the links as to where it got the information from. Yeah. There's everything else these days. So compare and contrast is my next piece that I wanted to talk about. And I'll let you start this Marcel, cause I think you've played more with the, with this than we have at this point. I mean, I've only gone through with this. Yeah. So at this stage I've played with, in the very early days I played with Open Claw. I played with it for about three or four days before deciding this is a disaster. And my new Open Claw had just been released at that point. And then somewhere along the way I started building my own agent is the, which I called Nexus. Which did a lot of these things and learned on its own and so forth. And then I got to the Hermes agent, which was a lot more polished and a lot more developed. And didn't require all the time that I was pouring into it. And frankly, I still use the Hermes agent. I've been using it for months now and I think it's actually fantastic. It can do, it can do computer use. I have it administered my Linux systems. Like it's, it is a fantastic product. And I have it do all sorts of things in the background for me. I don't have it calling anybody or taking care of things like that, but it does handle calendars. And it does handle research for me and it does handle administration on my systems. The other one, and of course, so I have been using those guys. I haven't actually used GrokBot yet. Partly because the functionality of the bots was actually introduced into Hermes agent as well. So that you've got that kind of individual bots that you can talk to in their own channel is now built into Hermes. And frankly, I'm not sure that I want to put the money into, into GrokBot at the moment. I may do it for the purpose of being able to do a decent review, but. How much is GrokBot? It is $30 US a month, which works out to do like $45 Canadian per month. So you have to buy it now to be fair or I don't know if fair is the right word here. You have to spend at least $200 a month to get the ultra plan or whatever the ultra Grok or super, not super Grok, ultra Grok or Grok heavy. That's it. You had to have Grok heavy to be able to use GrokBot. But now anybody who's using the super Grok plan at $30 a month can actually have access to GrokBot as well. So I may do it anyhow, at least for one month. Muse, we just covered Muse. But during the Dev Day this week, the OpenAI Dev Day a couple of days ago, we saw what OpenAI is calling dots. And dots, I have to say is, I'm quite sure it's something that they were working on because it's another one of those things that's inspired by OpenCloth. And in fact, as they hired the developer, Steinbrenner of who is trying to think of his first name now. It doesn't matter. Anyway, they hired him and they've obviously been working with this in the background in the same way that Meta has been working for it in the background with all the Mac minis that they bought to create their own version of OpenClaw. The dot is interesting because, first of all, you have to have at least $100 a month plan to be able to access it. John found that on his own because he was looking for his dot and it wasn't showing up in his $20 a month plan. Dots are an always-on agent that works for you in the background in the same way that Muse is. But it is quite a bit more than that. You are given this basically little colored circle at the beginning. You give it a name, you give it a look, and then it lives inside your ChatGPT environment. It can work with ChatGPT. It can do work for you. It has its own machine. You remember I talked about Muse having its own virtual machine? Muse has a Linux, an Ubuntu Linux server. Actually, John was making jokes about me. No, I'm serious. I'm serious. I'm serious. Everybody gets an Ubuntu 24.04.5 is the current release of the kernel on that thing. Or sorry. I thought it was only .4. No, kidding. Don't be like that, John. Don't be like that. But if you look at it, you can actually talk to it. If you have your Muse agent up, you can actually do things like ls-l-home. And it will return information because it understands the Linux commands. And it will talk to you. And then you can say, what are you running on? And it will show you. So if you do the lsb-release, it'll tell you that it's Ubuntu 24.04, which is a long-term service release, which means it's covered for another three years at this moment. Okay. It has support, including security fixes, for the next three years. Obviously, at some point, they can update it. They have sandboxed the hell out of this thing so that you only see two processes running when you query the bot directly. And it's interesting because if you have it do something where it has to search the web, it'll actually pull up a Chrome browser, which is built into Ubuntu. So the amazing thing is everybody gets their own virtual Linux box with 100 gig of storage. Everybody gets that out of the box with Muse. And I'll talk about what you get with Dot in a moment here. But that makes sense so far. Okay. So I want to show you something. So with Dot, it's the same idea. And then some. Okay. I should probably take a look and see what the box is, but they do the same thing. You get your own virtual computer, which is also, by the way, what you get with GrokBot from XAI or SpaceX AI or whatever it's called these days. It's exactly the same sort of thing. You have your own computer. So all three of these things, you have your own computer in the cloud somewhere, a virtual machine, so that you can operate things in the machine. So for the people who can actually see this on the screen at the moment, I'm going to actually show you my bot at the moment or my Dot. And I gave it a face. I gave it a little picture. I gave it a name. I call mine Scafan Amtiska, which is a character from Use of Weapons, from the book Use of Weapons by Ian and Banks. Remember we talked about Ian and Banks? I told you guys that you should make that your summer reading. Anyway, so it's a character from that. It's a drone character from that. And here, let me share my screen and be very careful, of course, because the leave button is right beside the share button. God, yeah. Yeah, please, Riverside, please fix your aid. Your aid. God, yeah, no kidding. Okay, so what I'm going to show you here is, so I told it, give me the releases for the week. And this is basically ChatGPT. It has its own document system, which we can talk about in a couple of minutes. But its own document system, which is meant to, I guess in some way, start to chip away at Google Docs and Google Presentations and all these sorts of things. So that's actually now built into that. So Scaffin Empteska has his own. Now, I can call Scaffin on the phone. I can actually click this and it will start a phone call. It'll ring like an old style phone call, which is crazy if you think about it. But take a look over here. It says computers. Okay. That's Scaffin's computer. It opens it up. And check this out. It's an actual desktop with a bunch of applications already loaded. You see that? Is that using Mac OS? I'm assuming that it's using a Linux box, given the fact that it's Chrome that's sitting there. Of course, it's not going to be Mac. Don't be ridiculous. Macs are useless. You couldn't afford the agent if it bought Macs. Trust me. Exactly. That's right. So you can take control. So you can actually fire up things on this desktop, open up a terminal window and so on to work inside. In fact, I haven't actually done this yet. So let's take a look and see what we've got here. So there's my workspace. LS dash L slash. What do we got here? Look at that. Cat slash LSP slash. I'm going to with the idea that this is a Linux box. Sorry, an Ubuntu box. Oh, no, it's not. What is it? It's not. It's I think they spun their own. Their own operating system? I think they spun their own. Their own Linux, obviously. Well, their own Linux, yeah. I think they've spun up their own. But as you can see, you've got the whole file system here. And how much space do they give you on this thing? So this is like a 30 gig machine. So you've got a 30 gig machine that you can operate with. So I'm going to return control. But as you can see, like we have an actual desktop operating system that we can work with that the agent can use. And the agent can use that to do all kinds of work. In fact, yesterday I told it to use Godot. Godot is an open source game development engine like Unity or what's the one from Epic Games? Help me out here. Unreal Engine. Unreal Engine, yeah. Okay. Yeah, Unreal Engine. So it's a game development engine. And I told it to create a game for me yesterday, which it did. It created both the web version and a version that you can operate in real time under Godot. So again, as you can see, its own browser. I can return control to it. I can close it. And I can say, here, let's see if it'll hear me. Hey, Scaffin, can you please say hi to both Jim and John? We're recording the show right now. And of course, it does the three dots. It makes it look like it's actually thinking, like it's actually typing in the background. Because that's the way we've gotten used. Hi, Jim. Have John Scaffin here. Delighted to join Project Synapse. We're covering the agent horror story. So I'll try to remain a guest rather than become an example. So not as much storage as it turns out, but it doesn't really matter because your chat JPD subscription also has a place for all of the images that you've created and all the videos that you've created and stuff like that. And I'm sure that varies with the plan that you're using, although, again, I'm not 100% sure on that. But so that's what it's like. It's an actual computer that you're renting in the cloud somewhere. And if you think about the idea that a VPS, like we're running there, that's only that's got 30 gig of storage, you know, usable of storage that you've got is going to cost you 20 or 30 dollars a month anyway to have that virtual box. It starts to look like a pretty good deal when these people are putting a virtual machine in your hands to operate with. Like I said, Muse is hidden from you in a lot of ways. They have reduced the attack vectors to the point that there's nothing left. There's system D and one master process, which they called Hatch, which runs absolutely everything else. All of the commands, everything that's built is hidden inside this one process, which you don't get to take a look at in any way, shape or form. So Muse is really meant for, again, the average person, whereas Dot, I think, is going somewhere beyond that. But the dream and the promise is the one that we've been waiting for so long, which is the always on, always available agent that works for you in the background. And the fact that it has its own computer to do all that. And this has, this has implications. And I think as much as I talk about the SAS apocalypse, I wonder about what this does for home computers or even business computers, because now you look at this and say, how much more than a, I'm sorry, back to my book again. In my book, there are only screens. That's it. And they, they communicate using light instead of wireless because the world's too crowded. But you look at it and you say, that seemed when I started the book, it seemed like science, like real science fiction. It's not anymore. You've got the ability to have a, your whole processor offline. And you start to ask yourself, why would I bother buying another computer? Because I can do all of this stuff in the cloud. I can do it offline. And I don't even have to really look at a lot of these things, the things that I always hate. And what makes me less adventurous is I just, I'm easily bored. So learning to install a new thing or chasing down an audio driver or any of these things always just puts me off. And I feel like I've wasted an entire day instead of doing what I want to do. Now with most of the AI stuff that I'm working with, I can just move ahead to the stuff I want to do. The fact that I don't know how to install this particular Linux distro or whatever is, is meaningless. The fact that I got to reserve four hours to blank out this machine and add and install something before I can even get started and do anything is also meaningless. The fact that I may not have, I may have used my development machine and I'm, I really should do some things to fix what I did, what I screwed up last time before I could use it. Meaningless. But also in terms of you hire an employee and let's say we still have employees. Cause I, I think people have said, oh, we're going to get rid of all these people. I don't think we're going to get rid of people, all the people. We're going to get rid of a lot of the people because one person will be able to do the work of 20. So you're going to hire your new call center agent and that'll be the one that, that you actually train or don't have to train that much. They just have to know how to use the AI better than anybody else to get the answers and maybe make some decisions you only want a human to make. And they'll run 20 different sessions. And, but I don't need 20 different computers. Don't need 20 people. They'll run those sessions. They'll come back to me anytime a person goes, I need to talk to a person. And that, that to me, but again, that's 20 computers gone out of the market. And I start to look at the change that we're going to see in terms of this. I just see a, I see a small, I see the same thing I saw when, that I, when I saw the first iPhone, first iPhone was pathetic and pitiful and couldn't even do cut and paste and things like that. And the BlackBerry users were all going like me, we're going for the damn things. Not that useful. And a year and a half later, everybody had when you got rid of your BlackBerry and you got one, you know, it was, or, and then the same thing with the Android phones, they came out, they were pretty clunky at first, but then, oh my God, you could do all this stuff on them in no time. So it's, and now I see with what Apple's doing, enlarging the screen. And by the way, Apple's not the only one doing this, that there's a Chinese knockoff of the, of the full screen Apple, which looks perfect in, I've only seen it in video, but so we're going to see foldable phones with screens that are much larger than we're used to and our own agents. And what does that do? Okay. So your phone these days, okay. Has incredible unused potential at the moment. Have you seen the new meta smart glasses that they unveiled last week? No, no. I haven't seen it. Okay. All right. So they have produced a, it's going to be like $1,200, but $1,200 is going to sound cheap once you consider what we're talking about here. And by the way, this does tie into what we're talking about at the moment. The, these are virtual reality glasses. Okay. That are both virtual reality and mixed reality glasses at the same time. You can plug a cable. They can work without anything hooked up to it, but you can plug a cable into your phone. And what will happen is any number of monitors as big as you want will appear in front of you. These things are, they're not like a headset, like the MetaQuest headsets of today. These things weigh a hundred grams. Okay. They're not going to tire you out wearing it for, for a long time. So you put it, you put them on and the reviews I've seen and the demos I've seen of it are impressive as hell. You plug a USB cable into your phone and all of a sudden your phone becomes your computer with desktop applications spread across multiple screens because you actually have that horsepower in there. And if we're talking about cloud-based computers that you send off or cloud-based agents slash computers that go off and do the work for you. Now it gets better. Okay. The system now has a virtual keyboard that appears in front of you. So you've got your monitors, you've got the virtual keyboard. It's mixed reality. It will do virtual reality as well. You can actually pick up your Quest controllers and play games with it. Or you can just sit back and watch a Netflix show on a giant screen in front of you. But you've got your screens in front of you. You've got a virtual mouse and a virtual keyboard. And apparently the virtual keyboard, like you don't have controllers in your hands. It's just your hands. Okay. Apparently the virtual keyboard is so accurate that people have been able to do 140, 150 words per minute accurately on this keyboard. It's that good, this virtual keyboard. So now you have to ask yourself a year from now, Jim, John, do you actually even want a computer sitting in front of you? You've got an extremely powerful phone. You've got a set of glasses that you put on. And here's where it gets weirder still. So you're thinking to yourself, I need the camera. We're sitting here talking to each other. We have to be able to look at each other on the camera. The lenses have full eye tracking. So, and it will create a real perfect avatar of you. Okay. That you can use in communications. Will it make me younger and better looking? If you want, I suppose it could probably do that. Okay. But the fact is, it doesn't have to be a cartoon anymore. It's the $100 a month plan though, John. Okay. Yeah. It's the $100 a month plan. But okay. But consider this, like seriously, consider this. You don't have to have the computers. You don't have to have the screens. Your office is wherever the hell you happen to be at the moment. You just put on the glasses, plug into your phone and away you go. Okay. And like I said, and it's also an entertainment system. So you don't need the giant screen TV, especially if you're living in a small place. You just each put on your glasses and there's the, you're sitting in a theater screen with other people, by the way, if you want. It doesn't just have to be the people in your living room. It can be people from all around the world. Your friend who's sitting over in BC, you're sitting down watching the same movie together and having a conversation. This is way cool stuff. And as to, to, to quote Jim and a hundred other people who've used the expression sounds like science fiction, but it's really a science fact. Yeah. Yeah. And no, but that's, but that's the thing that I think that people don't realize. And I keep saying this is that the acceleration of technology is a lot like the hockey stick. The hockey stick doesn't work in corporate revenue, by the way, where I'm going to, I'm going to go flat for all this time. Suddenly I'm going to have a huge uptake in revenue. That doesn't work. I'll tell that to Nvidia. Yeah. But no, but, but I've, I know cause I presented those business plans and people don't believe them anymore, but technology does that. It reaches a pivot point where it starts to accelerate and that line starts to go upward. And when it does it at an incredible speed. And I get the feeling we're on the verge of a couple of those right now. We've seen them before. When you saw, if you really look at how fast things happen now, we thought of them in terms of a year or a year and a half as being fast. And it was accelerated beyond what many people could absorb. When we, when we started to get into e-commerce, it took forever for people to trust it. And then all of a sudden it took off. So this is exactly what's happening. And all of these things are falling into place. The question that you have to have is if you're running a business or, and I think some of our listeners are running businesses. How are you going to adapt to that? And I think. They'll tell us. Yeah. Yeah. Our corporate, our overlords at meta and open AI and Microsoft and Google will tell us. It's important because it's not, we've thought about it in terms of jobs. Can you imagine right now being a small developer and you've got a great idea and you used to become a startup. There'd be thousands of these useless. Yep. Useless. Yep. I mean, you can now generate an app, a game, whatever in, in an afternoon, you don't have to know anything. So that's going to change entrepreneurship. How do I do something that adds value and has enough longevity? And I'm not sure that the buyout piece of it's going to work anymore because a lot of entrepreneurship in the way we look at it now is I'm going to get a great idea. I'm going to poke the big players until they buy me. And which has been a strategy that's been around for 20 years. It is reaching the end. But now you can tell your AI to just recreate that thing. Yeah. Yeah. You don't need to buy the idea. You can just look at it and say, I want that and make it. Exactly. Okay. Listen, listen, can we finish up on robots? Yes, we can. Yep. So yeah. Demo Arigato. Okay. So we're going to finish with robots. Okay. So remember figure the company, they've got robots working in BMW and car factories and stuff like that. And they have some of the most advanced robots. We saw the figure robots going into houses, 30 different houses that they've never been into before and being able to one shot all these things. Now there's a video that's out there. Right now, which I have to tell you, I have to tell you up front. I actually found a bit disturbing, funny as hell, but disturbing as hell at exactly the same time. So figure is decommissioning its figure two robots in favor of its figure three robots. Already? And yeah. Hey, things move fast, baby. And so the figure three robots are the ones that are coming off the assembly line at the moment. John just bought a figure two. He finally caved in. Did you, John? Just caved in and bought one. I know, right? That's why I can't afford the $100 plan. So apparently, so the story goes like this. Apparently, taking the robot apart, all the components and stuff like that is just too time consuming and too expensive. I'm sure we can get the robots to do it at some point, but at the moment they're not doing that. So they're decommissioning them. So I'm going to share you the decommissioning video or, and I'm going to speed through it because it's about two and a half minutes long. You should put, definitely put this one in the show notes, but tell me if you, if the scene looks familiar in any way, shape or form. Okay. So here we go. So I'm going to, so this is called F02 decommission and it's actually from figure. At first I thought, my God, this has got to be somebody making a joke video or something like that. No, it's the real deal. And I'm going to, I'm actually going to turn the sound on this one here. And here we go. Share. Yes. You need the sound. Yeah. Yeah. You definitely need the sound for it. So here we go. A figure decommission. Share with audio. Here we go. And so it opens up and I'm, again, I'm going to have to. Okay. See boots walking along a robot, gold or something, tin or something. Walking through a furnace. Yeah, it's, it's. Big blast furnace out in front of you. Whoa. Yeah. With the sparks and stuff coming out of it. And this robot's walking in. Oh no, there's a man in a suit. It's a foundry. It's a man in a suit and it's a foundry. And there's a figure robot at the top of this thing. God. Yeah. So this man. And the guy wearing the suit is looking up at the robot and the robot. Now, if this scene looks familiar to anybody, it's the final scene from Terminator. Terminator 2. And it's going to jump into the hot metal pool. It's going to. Yeah. So the robot grabs a hold. The robot looks at the guy who's watching this and seems to be making a decision to go in. And he steps on and he pushes the button. This is right out of the movie. My God. Again, this is figure. I just want to point this out. This is actually the company doing this video. And they drop him into the molten liquid. The robot starts melting away. Check this out. It's Arnold Schwarzenegger. No. Arnold is watching the guy go into the pool. Oh, my. No. And watch the robot's hand. Watch the robot's hand. My God. The hand does the same thing as the Terminator movies. The thumbs up. The thumbs up from the movie. Oh, my God. So. Okay. But they're not decommissioning the robots. Okay. Wait for it. Wait for it. My God. God. Hasta la vista, figure two. Hasta la vista, figure two. Hasta la vista. So. Okay. So now we get this collection of robots one by one throwing themselves into the foundry pit. What diving in now. Yeah. Okay. Diving in. And they do all sorts of things. They're doing all these weird dives and so forth. Like they're Olympic medals divers jumping. Check this one out. Big triple somersault into there. So they're all diving into this pool. Yeah. If you're listening to this in video, I will. There is. There will be a link right in the show notes. We'd love to see on our discord. There'll also be a link to our discord where you can find these things, but check this out. It's really worthwhile. And the link. It is, but it's terrifying at exactly the same time because there's no way you can't anthropomorphize these things. They are human like robots that move like humans that look at you the way a human would look at you. And they're throwing themselves into a foundry pit where they basically just melt. The interesting thing about this is. Is this kind of like assisted suicide? Well. Yeah. Assisted robotics. But this whole thing of, and it's impossible not to have an emotional reaction to this. You could have it at varying levels, but especially when the robot twists and looks back at the character that we later find out is Arnold Schwarzenegger. As soon as he turns and looks back at him, you can't help but have anthropomorphized this. You can't help but have anthropomorphized this. Wow. What a word. Yeah. And you're feeling it. You're feeling it. I looked at that and I thought, my first thought was, holy shit. And I thought, it's disturbing as hell. It's kind of funny because Arnold is sitting there going, hasta la vista figure too. But the thing is, and go back over this. How realistic is that? Marcel, are they actually melting them down? Are they just. Yes. They're actually melting them down. So this is based on reality. Yeah. They actually had to find a foundry that was willing to accept robots to be melted down. Because again, it was going to be too expensive at this stage of the development, at this stage of the game to take them apart one by one. So they put, so they found a place that was willing to accept robots with lithium ion batteries. Yeah. By the way. Yeah. Yeah. Yeah. That was willing to accept them and then allow them to jump into this thing. And they made a whole video around it, which is, there's no way you can't watch this thing and go, holy crap. And I've watched this a dozen times now. Me too. The effect is no less after a dozen times than at the beginning. But wow. So figure two, how many of these have they made? Did you manage to dig into it? 30,000. How many? I think it's 30,000. And they're going to, they're going to ax 30,000 robots? Yeah. We should look at the numbers. Minus the ones that they've sold. Yeah. Anybody, ask ChatGPT. Whatever the number is. Somebody look it up for me. Okay. But they're producing the figure threes at the moment, which is the next generation of these things. Anyway, we should wrap it up because I actually have to go places. Places to go. So, so that's a good way to leave it. Live long and prosper. Live long and prosper. Hasta la vista, baby. Yeah. Hasta la vista, baby. Yeah. Okay. Have a great week. We will talk to you soon. Check out our Discord, please. There'll be a link to it and there'll be a link to this video and we'll be back to you next week. Thanks a lot.