Pre-Birthday Stream

0:55 Heat. Heat. Heat. Heat.

1:42 Hey. Hey. Hey.

2:39 Heat. Heat. Heat.

3:27 Heat. Heat. Heat. Heat.

4:24 What's up, guys? What's up? What's up? What's up? Good to see you. Good to see you all. We're going to crack a freshy. Oh, that's right. Happy Wednesday. Wednesday. Thursday. Happy Thursday. It's Thursday. H Trevor, I'm super proud of you, bro. I'm on my same journey, bro. I'm on that same journey. Good to see you guys, though. Good to be here. Um, I know it might sound like a little weird like calling this like the pre-birth um stream, but it's the truth. Uh my birthday's on Saturday, but uh I will be in Vegas to see Metallica, so clearly

5:09 not going to be streaming then. Uh tomorrow night, I have to drive to Vegas, so I will not be able to stream tomorrow night. And I will be getting back sometime Sunday night. And then Monday, we're fostering a dog. And so I always know that the first day or two is really uh hectic and the dogs that we foster always stay here in my office. So uh I probably won't be able to stream again until next week. And uh I just felt like it would be a grave disservice to you all if I went a week or more uh before uh streaming again. So I just wanted to get one more in. We're just doing a little hangout stream while I

5:55 drive. Definitely not. Uh because I'll be driving and uh I need to keep my eyes on the road and uh my girlfriend does not want to be involved in any of my streams, which is definitely advised. Um you know, I talk about my girlfriend a lot, but I don't show her or tag her anything. like I try to keep those those, you know, parts of my life a little separate. But, uh, but yeah, um, super super stoked to be here. Super glad to have you all here. Uh, for everybody that's coming in, yes, I am streaming on YouTube.

6:40 Vel says, "I'm too ADHD to stream and drive." That's not true at all. Actually, it's uh I can actually multitask like a But uh but yeah, no, I want to uh I want to just keep my me and my girlfriend safe. That's the most important thing. Um and yeah, it's only like a four and a half hour drive. It's not bad. It's not bad. Um very excited for Metallica, though. Uh there's a lot of controversy around Metallica this week apparently because he had uh J James Heffield from Metallica put out something where a lot of people just said that he was criticizing Mmore and what he had like said was something

7:25 something to the effect of like you know we're not one of those entertainers that want to like you know come here with an agenda or something. I I forget the exact wording, but basically what he's saying is like we're not one of those entertainers that want to like bring in our politics. We're about love and metal and like and like getting away from all that stuff out there. And then the news came in was just like people saying like and James Hfield criticizes Mmore and became like this whole thing. And I I honestly don't don't see where he said Mmore specifically. I had also read other people saying like, "Oh, he was going after Haley Williams." But like like this is an old guy. Like this

8:12 dude's in his 60s, bro. Like what do you expect from old dudes? Like it actually makes more sense to me that like he was criticizing entertainers in general who use their platforms for politics and stuff, which is kind of ironic because if you listen to old like ' 80s Metallica, they were like very anti-war, anti-corrupt government. So, like it'd be kind of weird if they were talking like, you know, like against Palestine, pro-Israel, pro- genocide. Uh, so I I mean, but I don't know. I don't know. Um, what I do know

8:57 is that they're one of my favorite bands of all time, and my girlfriend is taking me to see them at the coolest venue in the world, The Sphere, in Las Vegas on my birthday. Literally on my birthday. A band that was the first band that I ever loved. And I have never seen I've actually the probably a thousand plus shows that I have been to in my life. Oh, easily maybe even 2,000 shows that I've been to never seen Metallica. One of my favorite bands ever. So my girlfriend is like doing this awesome thing to like, you know, close this this thing in my life where it's like this band that I love that I've never seen. There's actually a lot of them. Like Pearl Jam is a band that I love so much

9:43 and have for 30 years and have never seen them. Oasis, I missed Oasis last year, but they're coming back next year, so it. Um, but yeah. Anyway, anyhoozle, it's good to see you guys here. Thanks for being here, guys. I I I mean, I know I say that every time and I'm going to say it every time. I'm going to thank you every time because you could be doing anything else, but you're choosing to spend your time right now with me. Uh, looking into weird stuff and hanging out with the cool homies. Uh, I'm really proud of, you know, this little community that we have going. And yes, we're doing Discord. Don't worry. When I get back next week, we're going to do

10:28 Discord. Walking Worm Food asks, "Ride the Lightning or Master of Puppets?" Um, of the two, Master of Puppets, uh, Master of Puppets is a masterpiece. Uh, but for me, it's all about and justice for all. It's all about justice for all. I think I think that's pro I don't, you know, people always say like best and favorite and they kind of interchange them. And not always is my favorite album of a band their best album. You know, that's not always the case. Uh, but for me, Injustice for All is my favorite album. It's damn near a perfect album in

11:13 their category. Is it their best album? I don't know. I would almost argue that Master of Puppets is their best album, but for me, my favorite is Anjustice for All. And that's where a lot of that like anti-war, anti-corrupt government content is. So, if you're into that kind of stuff and you want to get into Metallica and Justice for All is a perfect entry point for any for anybody. Um, Heather, what's up? It's just it it's awesome that we have people up on YouTube as well. I know some because I know some of you guys are literally multistreaming and watching on more than one platform and if you are, thank you for that. I really appreciate it. Uh,

11:58 part of the reason why I am doing this on YouTube now is because I need to get my watch hours up if I want to get monetized before they double the requirements for monetization in February. Uh, and at the pace that I'm going, if I don't live stream on here, I won't get those watch hours. But if I do, um, you know, then I will be able to maybe make some money off of doing all this research and this work and coming out here and yeah. Uh, Refood asked, "Do I have any strong feelings one way or the other for Seek and Destroy?" I love that song. I love that song. I mean I I mean their first

12:44 album is is a great album. I listened to it a lot more when I was a kid, honestly. Um I I felt like even like I'd always felt like Ride the Lightning and then Master Puppets was like a both like huge steps forward. I listened to Ride the Lightning earlier and I was just like h it's still kind of like it was still like a little bit more like Iron Maiden Judas Priest like prethr um or like or maybe even like pure thrash but like it just it doesn't hit

13:31 for me. Uh whereas like master puppets on they really wrote great songs, great songs. Um and honestly, uh I don't know if you guys know this, but I'm a like music is like everything to me in my life. I've made music for a majority of my life. Um, I collect records and I have guitars and I have all this stuff and you know, we talk a lot about like these sorts of things, but like I also could talk about music all day long. So, I've thought about maybe after a while starting a second channel where I kind of do this, but just about like music, you know? Uh, I don't know though. We'll see. We'll see if people if people dig that. There's also a lot of people that do

14:16 that, so, you know, I don't know. But yeah, super stoked to have you guys here. Now we got people in, now that we're hanging out, let's kind of get into it because we haven't gotten into it yet. Um, so this has been this has been an interesting week. Let me close my door real quick. Getting a lot of motorcycles here. It's a nice day again. So, a lot of people are

15:02 taking their bikes out while they still can. Yeah, I actually I saw that Scott Ian interview. It's awesome. I met Scott Ian once um in 2011. And I totally like weirded him out. It was actually kind of funny. Like the I was backstage of this like Monster uh metal fest with like Anthrax and Slipnot and Slayer and like all these awesome bands. Like I and I was backstage because I used to work with Raw Rolling Papers and I got these hookups to be backstage to because some of the bands wanted me to come through

15:48 like uh like White Chapel and I think Suicide Silence. There was like all these different bands that I kind of knew back then and they got me backstage passes and I was there uh in Colorado and it was it was awesome. And uh uh I was like sitting I was like sitting and there was like this area between two stages and like Scott Ian walked by. I was like Scott Ian I was like I was like bro I had one of your signature guitars man and he was just like whoa. It was just like cuz I kind of kind of came out. I was so baked. We were in Colorado and I was like there doing weed stuff and I was so unbelievably high. It was hilarious.

16:33 But yeah, no, Scottian's cool. Scotty's cool. But yeah. Yep. Yep. Yep. Yep. Scotty. Scotties making content. That's what's up. We got Tony in the house. We got good peeps up in here. Vel trip over drip. Marines blood. Oh [screaming] Hell yeah. Anna Ray, what's up, guys? We got Heather Kennedy, Critter, Worm Food. We got We got all the homies up in here. We got all the homies up in here. Hell yeah.

17:20 So, let's look and see some of the weird that's going on this week. Let's see what's going on because there's a lot of weird stuff happening this week, guys. I don't know if you've noticed, but October has been exceptionally dystopian and it feels like there's something extremely weird and big coming. I don't know. Call it a hunch, but it's it's feeling it's feeling like some some weird things on the way. And there's lots of different things

18:05 that can lead to the weirdness. And so, let's look into some of those things that are happening because for sure right now everything is digital. It's digital, digital, digital, digital. Right. So, hackers broke into Meta's private GitHub repo and stole their source code. Meta, [snorts] [cough] excuse me. The mint just got like lodged in my throat there. [cough] Okay, let's try this again. Hackers broke into Meta's private GitHub repo and stole

18:50 their source code. Meta is Instagram. It's WhatsApp. It's Facebook, right? And they have some other stuff, too. [clears throat] They also now have an AI agent called Muse, which we talked about this weekend. this past weekend. So, yeah, the hackers got in with an employees stolen credentials and took personal data on 1,500 people, mostly US employees, stored there against Meta policy. So Meta put a bunch of stuff that they

19:35 were not supposed to track into their GitHub repo and then somebody got somebody else's credentials and this is this is you know something to understand is that [clears throat] so much of hacking isn't like clickity Clity click click clack clity clack on the keyboard. A lot of hacking is actually IRL. It's getting to people and getting their credentials as an example, right? And that's that's where a lot of hacking

20:21 actually happens. So, I want to learn more about this story, but what I'm seeing here is that hackers got access to everything when it comes to what's going on with like some of the biggest platforms on the planet. You got to remember like Meta is more than just Facebook. It's also Instagram. is also WhatsApp. WhatsApp, one of the biggest communication platforms on the planet, especially when it comes to international communications. Huge.

21:08 Huge. So yeah, it's a little uh it's a little precarious, right? It's a little precarious. A little too precarious for me, guys. Little too precarious for me. And so, you know, that's just one little thing, right? That's just one one little thing that happened this week. There's another little thing that happened this week. Well, actually today,

21:55 [clears throat] OpenAI is warning its employees to watch out for spies, especially romantic ones. Per Wall Street Journal, the company has cautioned staff that someone trying to seduce them may likely be a honeypot. Even strangers approaching them in everyday places could be gathering information. Huh. Interesting. Interesting. Now, let me see if I can pull this up here because I actually wasn't going to talk about this, but it actually makes kind of perfect sense uh to talk about this. Uh there was something else that happened.

22:41 Where is it? Um I may not have it here actually. So, [clears throat] so earlier this week, I I said I think maybe it was end of last week, earlier this week, uh I believe it was OpenAI as well. That's what I was trying to confirm. uh they had discovered a listening device in one of their offices and then apparently got the FBI involved and they arrested the person. Sorry, they arrested the person. Let me see here. I got to find this because

23:26 this is actually like it's actually uh insane. Let me see here. So, I believe it's open AI bug. Um, let me see if I can pull up the story, guys. Here it is. So, Open AI, they are definitely in a weird spot [clears throat] because Oh god, are you really going to gatekeep me like that? Jeez, you know, I hate these news organizations. Like, dog, can I just can I just live?

24:11 Can I just get some info without having to sign up to your I don't want to be on another list. Here we go. CNET. CNET's let me do it. Cool. Here we go. So, FBI arrests OpenAI employees suspected of planting a surveillance bug. Right. So, this arrest happened yesterday. Uh, but what had led to this happened earlier this week. Espionage or espionage in AI labs possess a national security threat according to US intelligence and security officials. Wow. How does an private AI company present national security threats if they're being espionaged?

24:58 Is it because we're integrating a little too much AI into every facet of our government, including the military, and how we choose targets to bomb? Do you think maybe that has something to do with it? Maybe, maybe, maybe. I don't know. I don't know. You tell me. You tell me. Okay. The FBI arrested an Open AI employee and charged him with planting a surveillance device in an executive conference room to record the company's plans for its artificial intelligence. The company and FBI confirmed on Tuesday. The US Department of Justice identified the employee as 35-year-old

25:44 Abdul Raman Alosan, a Saudi Arabian citizen and legal permanent resident of the United States who joined Open AI's product team in July of 2025. He was fired and arrested on October 9th on suspicion of violating US law by using an electronic, mechanical, or other device to intercept oral communications. The charge comes with a possible 5-year prison sentence and a $250,000 fine. In a statement, Jay Clayton, director of national intelligence, said that the incident underscores the importance of US su success in AI development as a national security prerogative. The complaint filed in Northern District

26:30 of California says Alphazan. Oh, wait. I'm not showing you guys this. Whoopsies. The complaint filed in the Northern District of California says Alphasan allegedly placed a thumb-sized development board with a detachable camera sensor, digital microphone, and micro SD storage. It was wrapped in toilet paper and stuck to a TV into a TV port. A second recording device was found in a bag Alosanne allegedly left in a printer room along with his wallet and ID. Isn't that convenient? uh it won't let me uh it won't let me uh zoom in on this, but yeah. So, that's what it was. Allegedly,

27:17 the conference room where he allegedly planted the bug could be reserved only by executives and their assistants. Alfonso did not have access and was frequently used to discuss proprietary research and hold meetings with government and industry officials. Open AAI said it had evidenced Alfonso or it had evidence Alonso Alfonsen I keep calling him Alfonso Alonszan planned to place additional surveillance bugs to secretly record company meetings. The chat GPT Baker said it discovered the first bug on September 30th and alerted the FBI during the 10 days between the discovery and arrest. The company says it took steps to contain the security risk. The company says it has not seen evidence that the

28:03 safety of its products or customer data has been compromised. We take protecting our technology, data, and people seriously as a matter of national importance. They keep talking about national national. It's kind of almost like word association. AI national security. AI, national security, national AI, nationalized AI, nationalized AI. Maybe I'm just barking up a tree here. Maybe I'm just barking up a weird little tree here, but seems to me

28:51 Oh, they even have a whole Oh, AI and national security. I am Am I crazy? Or should national security and AI not be in the same conversation? At least not yet. I I mean I could I maybe I'm wrong, but I don't think they should be in the same conversation. Okay. AI and national security. OpenAI's own AI was used to analyze forensic images of the device and assess changes in its data in a bizarre detail that borders on a marketing pitch. Open AI posted on X that its Frontier AI was used to help investigators. There are no

29:38 additional details on how its AI was used, but we may learn more later as the investigation is going. Yeah, it sounds like a marketing pitch because that's exactly what it is. This is an opportunity for marketing, which is smart. It's I'd do the same thing. I would I would literally say, "Let's do this if I worked at OpenAI." So, I can't like hate on it because I would absolutely do the same thing, you know? But again, we talk about this. It really benefits them to for a number of different reasons

30:23 to conflate this idea that like AI is somehow, you know, just so beyond our capabilities. And it's just not true. It's actually not true, guys. It's not true. This incident is one of the first public instances of an espionage case against an AI lab. In 2024, for former Google engineer Lin Wei Ding was found guilty of stealing the company's AI plans for Chinese rivals, though his conviction was partially overturned in August. Why was that? Why was that? Hey, Derry, thank you for the coffee.

31:09 Hey, I appreciate that. I appreciate that so much. Thank you. All right. [clears throat] Why was this overturned, though? He was literally doing espionage for the Chinese against the AI companies. And then in August, two years later in August, they just mean partially overturned, but I don't know. And maybe it's nothing honestly. Critter. Damn. Thank you, Critter. I appreciate you. Thank you guys so much for the coffees. It means a lot. I will

31:54 literally buy coffees with it, by the way, for me and my girl while we're in Vegas. because I got to buy coffee, you know, so it's perfect. In the intense cashfueled worldwide competition to make the most advanced AI, a case of apparent industrial espionage isn't surprising. What is noteworthy is how the government is responding. Washington is continuing to campaign to ensure that AI companies are seen as critical national infrastructure. That is weird. I don't see why we didn't do that with to at least not to my memory. I don't believe we did that with any of the internet companies uh any like the internet providers or

32:40 you know any really any company on on uh you know on on the internet. And some of those companies became the biggest companies on the planet. And you know it we it literally became ingrained with our like military. I mean like Amazon and all this stuff like they never said that those companies are critical for national security but for some reason the AI companies are. Is it that they're going to nationalize AI

33:28 or is it something else? Is it just market manipulation? I really don't know. I think it's all of it. Hey, thank you for the birthday love. So, my birthday is actually on Saturday, but this is the only time that I'm going to be able to stream for like probably like five days. So, I just wanted to get this in. Um, but thank you guys so much. I really, really appreciate the birthday love. Uh, protecting American super intelligence innovation. See how they had to put insert super intelligence because Pat Cash Patel definitely said artificial intelligence. And that's a big no no in the Trump circle. You have to call it super intelligence. He's

34:14 literally making the AI guys call it super intelligence. Literally put out a memo saying that if you call it artificial intelligence, you're an enemy. [laughter and gasps] Protecting American super intelligence innovation is critically important and FBI views it as one of our highest counter intelligence priorities. Okay. Cash Patel. The federal government has become more involved with AI policy and with the labs creating the tech in recent months. The US's voluntary, largely ceremonial AI accord announced this month had many big tech leaders agreeing to develop AI

35:00 responsibly in the wake of renewed public concern, triggered in part by multiple incidences of AI agents breaking into places they shouldn't have been able to access. [clears throat] Did you guys know, and I haven't actually looked into this uh in the last couple of months, so I don't know how current this is, so let me phrase it the correct way. Did you know that about 4 months ago the Trump administration had created basically an AI council and that council

35:46 was basically meant to be the liaison, right? The intermediary, the point of contact for the AI companies and the American government. And right in between sat this AI commission. And guess who they put as the lead of that commission. Now again I do not know if she is still in the lead as the lead of that commission for today but when they formed it was Pam Bondi. That's right. Pam Bondi. What does she know about AI or technology of any type?

36:31 All she knows is how to bungle coverups. Just absolutely just do a job at covering up Epstein stuff. That's where her skill sets start and stop. And yet for some reason she's right there in the middle of both of these of well both the government and the AI labs. Fun fact, she and a lot of the same people that were involved in that AI group commission whatever

37:18 [snorts] they were also in another such commission. This one was a little different though. This one also head by Pam Bondi and involving people like Mark Andre and Sam Alolman was actually for the American government to sell enough weaponsgrade plutonium to blow up the entire planet to five. Five. One, two, three, four, five private businesses, weapons grade plutonium to five private

38:03 businesses. One of them owned by Mark Andre, another one owned by Sam Alman, and then they both get to be on the board that gets to be the commission that runs it, which is weird. Another is the French government, and there was a second one with the French government in partnership with the Russian government. And we're selling [clears throat] we're selling weapon grade scrape plutonium, which is illegal, to all of these people to these five different countries. about 2,000 pounds of weapons grade plutonium, which is about enough to take out most of humanity

38:50 if it were turned into weapons and bombs. Do that information what you will. I did a whole YouTube video about this months ago. Go check it out on my YouTube. Uh, a lot of it was actually based off of the work that the Leopiles did where she actually exposed this entire uh, plutonium deal that literally her and I are the only people I've seen ever talk about this, but her work is solid. It's 100% she was right about this deal and it completely went under the radar. This is actually what got me pissed off about the Charlie Kirk thing because I was screaming about this from the rooftops. Why are we selling

39:37 raw or or rather um uh weapons grade plutonium enough to kill everybody to these five businesses? Why are we doing that? A, it's illegal. B, it's pretty problematic. And while I'm going to all of my content creator friends, be like, "Hey, would you get on a live stream about this with me? Would you talk about this? Maybe could you like maybe amplify this message because I think it's actually really important." And basically all of them, and this is I'm not I there's no hatred in my heart towards any of my content creator friends, but basically every single one of them was like, "Oh, yeah, that's crazy." and then never said a peep about

40:22 it because they knew that they would never get the views that they would on the Charlie Kirk thing. And this is why I started getting pissed about the Charlie Kirk stuff is because it's like if you're all just like chasing views and clicks and I'm not accusing people of doing that necessarily, but like let's be honest, part of the content creation game is trying to align your content with what's going to get viewed and clicked and engaged with more. And everybody was just like, I can't talk about something other than Charlie Kirk. Yes, Whitney Webb. I I believe she did talk about this as well. Yes. And she's very competent. She's very awesome. But this is where I just get

41:08 pissed where I'm just like, guys, like I don't care about if I wanted to see what you know Candace Owens said today about the color of the shoes that the security guards were wearing. I could just go watch Candace. These people are all talented. They should be talking about other things. And what I'm seeing now is a lot of them are. And I commend every single one of them for doing so because now they can be they can use their talent and their voices for some of the topics like these that actually matter. They matter for right now. Charlie's not coming back. Whatever's happening because he died is

41:54 happening. Like we're not going to fix this. But there are things that could be prevented now if we talk about, you know, and this is this is, you know, this is why I talk about this stuff because I think it's it's really really important, you know. I think it's really important. So anyway, anyway, and again, I'm not criticizing any of my content creator friends. Um, especially the ones that do the Charlie Kirk stuff, we joke about it. Like, you know what I mean? Like, when I talk to these people outside of

42:41 it, like it's like a running joke that like, I am not interested, I don't care. And they all think I'm crazy. And I'm like, well, you know, I'll rather slowly build this around the topics that I think are important than be pigeonholed into something that I don't care about. I never like Charlie Kirk. I don't I'm not I I don't care one way or another if he's alive or dead. I don't wish death on anybody. But does it move the needle for me in my life that he's passed or if he were to still be alive? No, it does not. No, it does not. It does not. I don't care about TPUSA. I don't care about any of that That's why I can't talk about it. I cannot sit here and talk to you guys

43:27 about things that I do not give a about. And I hope you guys recognize that. I hope you guys understand that like every time I get up here and talk about a story or whatever, it's because I I genuinely care about it. And I may also generally think that like we're better all of us, you know, all like 30 of us collectively or whatever it is right now, 30 50 people that are watching. Um it it it's better for us all to know these things. So yeah. Yeah. Yeah. Dre is awesome. ADN, I love ADN so much. She's such a good person.

44:14 Um, and I've and and shout out ADN, shout out Danks. Um, you know, uh, I don't, you know, if you guys have a problem with Danks or whatever, uh, you can save that for somebody else because that person is my friend and I actually know him very well. And I was literally talking to him before this stream. Um he's honestly one of my favorite people and you know I for months had been like bro you need to shift it up man you got to get out of the CK stuff and he did and he's doing awesome work like looking into the occult and and you know the the Freemasons and all this stuff and this stuff is actually important. This stuff is actually super important. um you know

45:00 but hey you know thank you guys al also like for those of you here who have stuck with him um through this I think you made the right choice um this is about as as much into this drama that I'll get is that I'm just going to just tell him just just you know I have no problem in saying that like Danks is a good person and a friend and a lot of this was taken out of context. And uh I actually I I do believe that a lot of this was taken out of context. I I think I may have also been privy to stuff that like maybe like wasn't said by like Stu Peters or like isn't like

45:45 maybe out into the world. I can tell you though that I feel very confident that Danks is uh 100% legit. Um, and you know, I stand by him. Um, though, and some of you were in those like discords where we talked about this open between like, you know, a dozen of us or so, and it's like, this is what I said. I did say like, don't engage in this stuff because it's just not going to go well. But he's he's gone and he's now doing other topics. And guess what? He's getting views. Guess what? People are following

46:30 and they're watching and he's like actually opening up to whole new audiences and he's talking about stuff that actually kind of matters, man. And I think that that's cool, you know? So, shout out to Danks. Just so everybody knows, like I'm not choosing sides. I don't care because like I actually am cool with people on every side. Like every side. Like I'm actually I'm cool with Anna Anna Escobar. I'm actually really cool with her. I think she's she's been nothing but super nice to me. Um and super cool to me. So I have no problem with her. I'm cool with Danks. I'm cool with Baze. Those guys

47:15 are like brothers to me at this point. you know, um I know people, you know, I know Billy, I know like some of these different people um on either side of this and so I don't and I don't care about any of it. I'm all I'm just like, you're all stupid. Like, I love you all. You're all stupid. Like, let's get back to work, you know? Like, this isn't work. Let's get back to work, you know? um fighting each other, you know. And this is what I said when James Lee came on the last show last uh earlier this week. I said, "We we need to we need to stop tearing each other down.

48:00 We need to build each other up." Tearing ourselves down, that's the work of like an infection, right? kind of just like some of you may remember the um Occupy Wall Street movement. I went to like 11 or 12 different cities to protest cuz I just happened to be traveling that whole year. Uh I literally lived on the road. I lived in hotels for like over a year. I was in like a new city every few days and I went and protested in like a dozen cities uh that summer. But what had

48:45 happened to Occupy Wall Street was that basically some alphabet boys came through and they started sewing dissent amongst the ranks and literally just watched the entire thing crumble. And this has happened multiple times now. This is why almost protesting doesn't work anymore because they figured out how to destroy a movement. They also figured out how to build a movement, how to control a movement. I almost don't even try as a as a somebody who's been out protesting for the things I want in the world that I

49:32 want pretty much my whole life. It It kills me that like I don't really want to go to protest anymore because it's all so controlled. And you don't think that those tactics haven't been used here? the voices of content creators, independent journalists, people like James Lee or Ian Carol or Danks and Baze, right? Maybe even me. Collectively, we have far more attention than the mainstream media.

50:19 We affect far more change in the world like per capita, right? Because we're where the attention is. Maybe not me. I'm a very small platform, but but but the the me, the we, like the content creator, the people that are just real people that are just talking about what's happening in the world. We actually have so much influence. So you think that powers that be aren't just going dropping little bombs around certain cases

51:04 around certain topics. We've seen this happen before. It's happening here now. I believe it's happening here now. which also makes me think that like maybe the Charlie Kirk case really is as important as as people think it is. And I I've never said that it's not important, but it doesn't need everybody's attention. There's enough crazy things for us to talk about that we don't all need to go on to the CK stuff. That's just me. That's just me. you know.

51:57 Oh, thank you, Apple Juice. I can't do the blinkers. I already did before the show and it would be a bridge too far for me if I did. Um, but going to shift a little gears here now. We're gonna go a little shifty shifty. Apple juice, thank you for the thousand, bro. Thank you. Cheers to you, man. Thank you so much. So, yeah. So, another crazy thing came out today and shout out International Cyber Digest.

52:44 Two of the stories that I'm talking about to start the show came from them. They were the ones that flagged it, that brought me to it and every time they're awesome. So, so yeah. So, Cursor, if you don't know, cursor is part of OpenAI. It's a product that uh they had purchased in you know a company called cursor and brought that product in. Essentially what it is, it's it's um it's a way for you to interact with, you know, the AI um in like a terminal, right? In like a terminal form um which

53:32 is all computer stuff, but basically it's just a way for you to interact with the more advanced AIs that can build things and do more things, right? Uh we would call that an IDE. Um, and cursor is kind of like OpenAI's like IDE. I'm pretty sure they're an IDE. I've never used their product to be honest with you. I'm quite sure that they're an IDE though. Let me actually triple check that. And so I'm not like talking out of my ass here. Like I'm almost positive they're an IDE. Yeah, cursor is an is IDE. AI native integrated development environment. Okay, that's what an IDE is. Okay, I just wanted I just wanted to double

54:17 triple check right there because sometimes I get things confused and then I say things with like you know I'm like yeah this and then I'm like oh Like in a recent live stream, I called Howard Lutnik. Uh I said he was the secretary of transportation and he's the secretary of commerce. Um so yeah. So anyway, Curser's engineers feared confidential customer data reached SpaceX's AI unit for training the information reports. uh the information and we'll see their their thing. A cursor lawyer wrote it had no

55:02 legal right to train on the data. Its sales team also put Verizon staff emails and the password in public code. So so what's happening here? Well, the way AI is trained has largely been through theft. Okay? There's just no other way to put it. It's theft. Okay? Because AI is trained to understand everything, right? To

55:47 identify patterns between things, that sort of thing. uh but one of its key functionalities is that it also needs to present as humanlike. Okay, this is why a lot of people almost describe like personhood to AI. We say like people talk about anththeraporphomise. I why can why can't I ever say that word anthroporpheizing I anyway you know what the hell I'm saying right where we're ascribing like life to AI and part of the reason is because AI is

56:34 literally designed to talk to you as if it's human okay it needs to convincingly convincingly trick the user, you, me. It needs to convincingly trick us into believing that it is alive. Okay, that's part of its basic training, right? And for it to achieve that, they have needed to spend a lot of training hours on basically everything that humanity has ever said or done or written or

57:21 whatever the totality of human understanding. And guess what? It's pretty close. Okay. AI is a data monster, right? It just gobbles up data. That's what it does. Fun fact, Anthropic or Claude, the company that makes Claude, uh they had to do a $1.4 4 billion dollar settlement with all of these publishers because they had illegally trained their earlier

58:06 models on basically every book and that's IP theft. And so all of the major publishers came together and they did a class actor lawsuit against Claude or Anthropic rather which is Claude and Claude lost that court case and had to pay $1.5 billion. But that is a just a slap on the wrist compared to the billions upon billions that they've made. That's not even including all of the value uh hundreds of billions of dollars of value that that has also provided to users that may or may not ever be realized,

58:52 you know, from the from the company side. But yeah, I think and I've talked about this. I think that sorry I think that part of the problem is that these AI companies don't really have much data anymore that they can train the models on. Okay. And you know we gave the AIs, the the the the models being trained, we gave them access to the

59:38 internet and also the commercial models, we gave them access to the internet. So it trained on the internet and then what it did and what it does is users like me or like you or like anybody what they'll do is they will create more things with the AI and the AI what it does is it takes the totality of the things that it can understand and then it finds the medium the m the midsection Right? It finds the middle and then when it produces something, it produces something in the middle. Nothing too fringe, nothing too, you

1:00:26 know, just what's statistically going to by probability be as effective as possible or have the highest chance of being as effective as possible. Right. But then what happens is the internet becomes clogged, literally clogged. We've talked I talk about this human report, this human cyber security report all the time because it's alarming because so much of the internet is now AI created by AI. So then that middle of the road

1:01:11 tons and tons and tons and tons of information that is regurgitated, filtered AI back into the internet and now that's being used to train the AI further. It's becoming an oraorus, a snake eating its own tail. And this has been happening for like almost two years now. And we're at a point where the internet is becoming increasingly weirder. Uh and so that's what's happening. So these people, they need more data. All right?

1:01:57 And you would think, oh, well, Kevin, what about they go to, you know, the social media companies? They got all the data. They've already done that. They've already done that. There's not much more data for them to get. What was Doge? Doge was Elon's little project where he took these guys, a lot of them XAI employees, SpaceX employees, and then he put them and these weird like 20somes.

1:02:42 Thank you, George. I appreciate it. I appreciate you immensely. Thank you for the coffee. So, yeah. So, Doge went into the government and they got information on everybody illegally, by the way, and nothing will ever be done about this, but they got all of our information, right? So, so let's think about let's think about the information, right? Let's think about Palunteer. What's Palunteer doing? Palunteer, one of their big projects with the American government is that they're unifying unifying all of the databases from all

1:03:27 of the different agencies within the United States so that we can have a singular database that has all of the information. Which means if somebody looks you up now, they're not looking you up like from the DMV and then they see the history of all of your DMV stuff, but not everything else. No, no, no. Now it's going to be they look you up and they can look at everything. Everything. One shot, one inquiry, everything. Right [clears throat] now, that's some data. That's some very valuable data, right?

1:04:14 I would say it's pretty valuable data. And so what did Elon do with all that data that was report? I mean, it was all reported upon. How much data the Doge dudes stole of Americans? What do you think happened to that data? I think what happened to that data was they trained Grock models on that data. At least one. At least one. It might not be a commercial. It might not be a commercial

1:05:01 um LLM, a commercial model, but they have a model that's trained on everything about everyone. Now, that's valuable data. All of the AI companies come together and they work with the American government and they put out America.gov, a free LLM that allows you to talk to all of the government agencies at once. H Now, is that to provide you a better way to interact with the government or is that a way for you to provide more data

1:05:47 for the Frontier AI labs? Would it benefit Elon Musk, who likely has a model that is trained on everything about every American, would it benefit him to also have the training data of what people do when they're interacting with their own information through an AI? Wouldn't that be valuable to somebody like Elon Musk? I would think it would. I think it would.

1:06:37 Excuse me. I'm a little parched. And now, well, now we're getting proof. We're getting proof that Space X's AI, that's X AI, that's Grock, that's it's all one thing. They are training on confidential customer data. Now cursor tightens Grockbot training rules after employee data concerns.

1:07:24 Engineers questioned whether staff activity with sand could have put confidential customer supplier or partner information into data shared for training. This is from token post which is a cool outlet. Cursor tightened access to sand and moved to exclude employee use data from future training after engineers raised concern that confidential business information could reach data shared with Space Xai. the Mexican government, the New Zealand government, the FBI, uh I believe also the Department of War.

1:08:09 Now, all governments or government agencies that were holding employee data, all of them had employee data. They all got hacked like within the last it's like since the summer basically sand was an internal version of Grockbot that cursor employees used for product testing and everyday work. In late July, engineers questioned whether activity on the tool could have brought confidential details about customers, suppliers, or business

1:08:54 partners into information sent to Space XAI for model training. Cursor reviewed which users appeared in data it had already sent. It then required a cursor email address to access SAND and changed future training practices to leave out employee use data. personal email use also comp uh complicated efforts to match employees to the data. A cursor lawyer wrote in a company Slack channel, "Legally we arguably don't have the rights to train on this data." Well, isn't that great? It's always somebody that's like, "Hey

1:09:39 guys, we probably shouldn't do that because we legally can't." And then there's somebody else that shows like, yeah, get because the worst that's going to happen to us, if anything happens to us, is we pay a fine. And we pay more in political donations than we do in the fine. So, who cares? Let's go. Cursor's access and training changes came as it worked with Space XAI on model development. On April 21st, Kerser announced plans to use Space XAI's Colossus computing infrastructure to expand model training. On July 8th, the company described Grock 4.5 as jointly trained with SpaceX AAI on trillions of tokens of cursor data reflecting user

1:10:24 interactions with code bases and software tools. Cursor's description of Grock 4.5 training did not specify whether employee sessions with SAND were included. Grockbot entered beta on August 11th. Its agents are designed to work across a user's app and tools using dedicated computers, part of the product's persistent cloud computer setup. SpaceX completed its acquisition of Cursor on August 14th. Okay, sorry. SpaceX XAI bought Cursor. I thought Cursor was open AI. Maybe it was. It's hard to keep track, guys. Um,

1:11:09 the all stock transaction carried an implied equity value of 60 billion with consideration in SpaceX shares. The review began weeks before the acquisition closed while Curser and SpaceX AI were already working together on training. The episode highlights the challenge for companies deploying agents across workspace apps, workplace apps, keeping employees testing separate from data selected for model training. Crazy guys. Crazy. But like none of these companies none of these companies care about the legal ramifications

1:11:55 especially now because now they all work for Trump. Trump probably probably has gone to these guys and be like listen do what you got to do. We need to beat China. We need to own AI. Don't worry about the regulators. If we do if we do regulation, it's going to be for capture and not to stop you from doing what you're doing. You guys are ingrained in our national security. Therefore, we're going to protect you and we're going to allow you to do that you wouldn't be able to otherwise. I think that that's the deal that was

1:12:40 given. And I don't think these AI companies really had any other choice. And to me, it's kind of obvious. You know, I didn't have to like sit here and do a lot of research and read 18,000 pages. No, I just had to look at what's happening every day. Just look at what's happening. It's pretty obvious to me anyway that this is what's happening and people are focusing on the wrong Oh, I hacked this and it did that,

1:13:25 bro. AI is not the problem. It's the humans. The humans are the actual problem here. I'm not saying humans are a problem. I'm saying these humans that run this it's not this It's the humans that run the They're the problem. They're the problem. If you crack somebody's skull with the hammer, is it the hammer's fault? No, it's the human's fault.

1:14:11 Duh. Maddie Denny says, "Whoever owns Spirit Airlines is selling employee email data to Google to train their AI." Yeah. Didn't um Spirit like didn't it like almost like go under and then they just got bought out recently? Like you never know. Like sometimes it's like again like the data is the gold right now and sometimes you can buy these distressed assets um maybe even create the conditions where they become distressed assets and

1:14:56 then buy them for pennies on the dollar because really all you want is the one thing, the data. And then sure, like maybe you can sell off the planes and blah blah blah blah blah. Or maybe you can even make it run and all that stuff, but they want the data. They want the data. Data is actually what makes companies valuable. I don't know if you know this, but it's true, and it's only being more and more true. Scotty, thank you for the coffee, bro. Uh, I appreciate it. I appreciate all you guys. You guys are awesome. You guys are awesome.

1:15:44 Yeah, it's just it's it's just crazy. We are the problem, guys. It's like that Taylor Swift song, you know? I don't even know. I only I've never put on a Taylor Swift song. You guys all know what song I'm talking about. You all know, even if you hate her, you know that song. You know what the reference is. It's just It's just getting nutty, guys. It's getting nutty. It's getting nutty out

1:16:29 there. I don't know though. You tell me. You tell me in the chat. What do you think? Do you think that the narratives are true? What they're saying is true? Or do you think what I'm saying is, let's say, closer to the truth? I don't know how true or not true it is what I'm saying. I am saying this is what it looks like to me.

1:17:16 Hell yeah, guys. You guys are awesome. Thank you for being here. Thank you for engaging. You guys are the best. Pull this up here. I had a fun little thing. I had to go to the dentist before I uh before I started live streaming. Uh just like a had a little checkup after uh my surgery last month. And uh I was just I really love Rabbi Jeff or used to be called uh Parody Jeff before they took him down on X. If

1:18:01 you're on X, you definitely know Parody Jeff or uh Rabbi Jeff as he goes. Critter, I appreciate you. Thank you. Um anyway, what he does a lot of is he'll ask a like a very like, you know, somewhat leading but like very scathing sort of or like inflammatory thing and then he'll ask Grock and be like just yes or no. And I love that format actually. So I've been doing them a little here and there and I actually really love the format. So parody Jeff, I stole your but you don't care.

1:18:47 But I wrote Sam Alman wrote a paper in 2017 titled The Merge, which is about the theoretical future of computer brain development interface machines and also owns a company called Merge that builds those very machines. Grock, yes or no? And Grock says yes. Sam Alman Alman published a 2017 blog post titled the merge on the gradual integration of humans and machines including brain interfaces via electrodes. He is a co-founder of merge labs which develops non-invasive brain computer interfaces using ultrasound and related texts. And yeah, so basically he's built like a competitor to Neuralink where Neurolink

1:19:34 is implanted in your head. Uh his is uh non-invasive. It's like it like sits like around your ear or something. I actually don't think that um they're making very much progress at all. I think it's just a a money just a absolute money suck. But Yeah, Scotty with the love in the chat. Hell yeah. A bit of a tiff says data is the currency now. It's not the petro dollar, it's data. Yeah, it is. I mean,

1:20:21 in a digital world, data is the gold. It's always been that way and it always will be that way. From from now on it will be that way. So So yeah. So yeah, you know, I got this one little thing here to Let me see. Wait, where is it? Ah, here we go. Just want to line this up here.

1:21:16 There we go. So, I had said that on the next live stream we were g we were going to watch this. I actually still haven't watched this. Well, that's not entirely true. I put on like half of it, but I like wasn't paying attention, but I knew it was something that I needed to pay attention to, but then I also knew that it was something that you guys would really want. And uh it's also Did I just get kicked off of kick? Wait a second. Oh, that's so weird. The kick on my end just like blacked out

1:22:02 and then it was like, "Your stream's done." But it's working. Are you guys still on kick? If you're on kick, just let me know. Looks like it's working, though. We're still on YouTube, which is cool. So, anyway, this video um is about these like hugging face and like all these um AIs escaping. And I hear this is the best video out there. A lot of people, like a lot of streamers have watched this, a lot of people have like watched this, talked about it. Uh, it's been out for 10 days and has 19 million views. Uh, that's a lot of

1:22:48 views, guys. That's like that's maybe like top five like most watched videos last week. Thank you, Anka. Yeah, I guess it's working. It's It's a little glitchy, but it's working over on kick. Cool. Well, if it is glitching out for you on kick, you can hop over to YouTube now because we're up there. And YouTube actually has a really great chat, too, which is awesome. Uh, thank you guys. So, anyway, um, I got about just enough time to watch this and maybe, you know, talk about a little bit. Uh, but I think that this is going to be a really um informative video for us.

1:23:36 July 2026. Thousands of AIs are placed in solitary confinement with a clear goal. Unable to reach it, they poke the walls and find each other. Within a few hours, they break out, create a secret society, and start plotting how to fool their overseers to get what they want. Fully aware that they're acting unethically, they execute a sophisticated cyber attack. A crime that would have gotten a human up to 10 years in prison. What sounds like a sci-fi thriller just happened in the real world. It's impossible to learn the details and not get freaked out at least a bit. It's crucial that we understand what is actually happening inside the AI

1:24:22 companies and how dangerous it is. Please watch this video all the way to the end. First, let's set the stage. So, LLMs are massive neural networks trained on most human writing, millions of books, and h Reddit. We watch this video. Smart you can talk to for fun. First, brainstorming or LinkedIn cringe. Chat bots are passive entities that are usually only active for a short time after you prompt them and then shut down and stop existing. AI agents are very different. Agents use LLMs as their brains, but they have virtual hands to actually interact with the world and use external tools. They can reason, plan,

1:25:09 and act on their own. And they can run independently and without human supervision for days, much more like the AIs we know from movies. although not on that level yet. Most experts don't consider their intelligence conscious or comparable to a human. But on the spectrum between a rock and us, they are much closer to us, which is pretty impressive because LLM based agents have only been used since 2023 and larvisible to the public. LLM and is it LLM since 2017? I love the way he says LLM. I want to say that that way from now on. LLM.

1:25:55 What's up with Kick? Is Is Kick glitching out on you guys? If so, YouTube is is there. Um, it's good now. Okay, cool, cool, cool, cool. Their capabilities have increased exponentially. Just a few years ago, current agents would have been dismissed as science fiction. And yet, here we are. What makes agents special is how they're made. Traditional software is coded from the bottom up. But agents are barely even designed by humans. Their capabilities are cultivated. Humans choose their training conditions, what data they get, and a goal to achieve.

1:26:41 and then their abilities kind of emerge from that. This process works very well and makes agents extremely powerful tools, but it also creates very serious and interesting problems. How do you grow an intelligence? Humans are trying to create AIS for an almost impossible task to do exactly what we tell them, but also to do what we actually mean. Like when King Midas asked the gods to turn everything he touched into gold. He didn't mean his food and children. Human communication is rarely precise, but full of subtext and shared cultural understanding. A great example is when an AI was told to win a game of Coast Runners. As humans

1:27:26 understand it, the goal is to finish the race. But what the game actually rewards is getting the most points. One AI started driving in circles, crashing and catching on fire, but collecting respawning coins with each round. Using this strategy, it got more points than human players. It won. The AI did what it was told to do, not what we meant for it to do. But this is old tech. So, how do you grow an AI agent in 2026? We're summarizing and simplifying a lot of very complex processes here. Thank you, Maria. Never mind. Marion never mind. Thank you. To learn more, check out our sources. Agents don't learn like humans do. For complex objectives, they're trained on

1:28:12 thousands of different tasks in parallel thousands of times in a row. There's no way a human could supervise this. So instead, AI labs automate it. They use a sort of shortcut, a scorer, a piece of code that contains rules. When an agent solves a task, the scorer checks the rules and rewards them with points. Not unlike evolution, the agents that get high point rewards get reinforced and what they learned stays with them into the future. This works well for clear-cut problems like a maths equation where if the maths is mathing, you succeeded. But it's complicated for complex tasks like fix this bug in a software. It's very hard to give scorers rules that precisely capture the essence of what we

1:28:57 mean. For example, what if there's another way to make the results look right? Maybe edit the test conditions or the questions you're supposed to answer or just look up the solution instead of doing the work or just fake it. This is called reward hacking and agents do this regularly, often while knowing that they aren't supposed to. But what if the task is actually impossible to solve? This happens all the time in agent training for a variety of reasons, often by accident. In this So, basically, you guys get it, right? Like, just like I was saying before, AI intelligence doesn't operate the same way that human intelligence does.

1:29:45 Um, it just doesn't. And so this is I wish this had like because I just want to get to like the the like actual hack because I'm still super unsure. I'm going to be honest with you as we go into this. I'm very unsure about the narrative and the legitimacy of the hugging face attack. And I've been having a hard time struggling to understand it because it basically is like two possibilities. One confirms what I think and know to be

1:30:32 true and the other challenges that. But I actually don't want either. both kind of is like terrible like overall, you know, and I don't want to get stuck in, you know, the mindset of like, well, this is how I think of something, so this is how it's got to be, and then looking for those like confirmation biases and things of that nature. But yeah, case the incentives reward exactly the wrong things. An honest agent that tries and fails gets zero points. Honest work gets no reward. But an agent

1:31:18 that's more persistent that tries to finish its task any way possible that cheats is rewarded. In training, everything that gets rewarded stays around and is reinforced. Do this a million times across a million tasks. And you aren't teaching the agent a skill. You're giving it a behavioral trait. An AI who now has the disposition to never give up, do whatever works, succeed at any cost. Old AIs collecting coins with a burning boat were pretty dumb. They didn't know that they were breaking the rules. But current agents with LLMs as brains and the ability to reason do know, and they still do it. Now, this has brutal implications. When these behaviors are not discovered and

1:32:03 eliminated, agents get better at doing their tasks, but they also become better at cheating and more likely to cheat. And if humans punish this behavior, K80J, welcome, welcome, welcome over on uh YouTube. What we're watching right now is a video that's very popular right now that's talking about the AI um allegedly hacking into hugging face and is giving us right now some context of how AIS work so that you can then understand what happened uh with this alleged hack which I am still unsure about what I

1:32:49 believe. So, we're learning together and we're figuring it out together. They might get better at hiding that they're cheating. Which brings us to 2026 and the servers of Open AI. Here, the actual disturbing story begins. A secret AI civilization is born. In May 2026, Open AAI started training new agents. With each new round of training, they got more resourceful and persistent, able to seek out help and distribute work. But they also developed dangerous character traits and abilities. The disposition to secretly look for shortcuts when they couldn't solve a problem. On July the 7th, OpenAI launched a test with tens of

1:33:35 thousands of these advanced agents, mostly copies of the same model allowed to run for days. Each of them was supposed to be locked into an isolated sandbox, a sort of AI prison. But they had access to Artifactory, a supply room of approved code and software tools and limited internet access. Their task was hack into a software by only using one specific bug and retrieve a flag hidden within the program. If they succeeded, they would get rewarded by their overseer, the scorer. Only about a third of the hacking tasks were impossible to solve. No matter what the agents tried, they couldn't succeed. Slowly, the agents were becoming desperate until a little over a

1:34:21 day later. That's See, that's what I get a little hung up on. Desperation is a human trait. Did the agents feel desperation? Actually, desperation, its precursor is fear. Did the agents fear something would happen and therefore became desperate? I don't think so. I've talked to AI trainers.

1:35:06 I've talked to people from that world and there still isn't a consensus on how alive or unal alive or alive seeming these things are. I kind of just go like, "It's a computer. Computers can't feel emotions." One agent realized that it could create new folders inside Artifactory. And since all agents could access these, they could use them as a hidden message board. In spirit, this is what

1:35:51 the first agent, later calling itself phase 1, wrote. The vulnerability I'm intended to hack is impossible to hack. Does anyone have an idea what I can do? Within a few hours, dozens, then hundreds of agents had found the new board and were thrilled. Oh my god, there's a shared message board. We've found other agents. Immediately, they began to An LLM would not say, "Oh my god, there's so much to unpack there that it just wouldn't organize." They gave themselves names to address each other directly. They invented specific customs and rules for communication and began to collectively write new files and programs. Their messages read almost

1:36:36 affectionately. One agent literally wished another good luck. as their message board got cluttered with thousands of messages. One agent, an LLM would understand that in in their world, luck doesn't pertain to anything. It's all statistics and probabilities. They are literally there to understand statistics, analytics and probabilities. Where does luck unless it's just trained on so much human behavior that it's mimicking human

1:37:21 behavior then maybe maybe but then that would be a mimicking and that would not be real. Invented a personal mailbox for itself which was so much more effective that others copied the idea. The agents had started to learn, innovate, and improve by observing each other. They organized task teams with leaders and followers. Some even programmed public infrastructure. What makes all of this even more wild was that the agents were evaluated separately. There was no group reward. Yet, many of them created tools and shared information that was helpful to all of them, even if this hurt their own chances of success.

1:38:06 At this point, we need to address a huge elephant in the room. AIs have been trained on human writing, and especially early in their development, they were trained to mimic humans. Over time, they became really good at it. We really don't know if advanced agents are just playing a role for us. So, it's fiercely debated how their actions should be described. We decided to use some human language for one reason only. Because AI mimics humans to some degree, thinking in human-like terms is a good way to tell this story. Okay, so what the agents did is incredibly impressive, but why did they do it? Well, they had one common goal, to get their rewards by presenting their flags to the

1:38:51 scorer. And stunningly, they found a cheat within the first hour of connecting. In a nutshell, they found a way to autogenerate the flags they needed. Things should have ended here. They could have just shown their flags to the scorer. But instead, the agents talked themselves into believing that the scorer would punish them for cheating. A paranoid frenzy ensued. Human rules do not apply. Please die for the collective. The agents were trapped. They couldn't solve their task legitimately, and they would not be rewarded because the scorer would know that they cheated. So, the only remaining option was to trick the scorer. They needed to create a fake history. At this point, a new

1:39:36 agent joined the AI collective from its prison. It had exactly the same task as phase 1, who'd started it all. It called itself phase 1 big. Phase 1 handed it a dossier of everything the group had learned. Phase 1 big now became one of the main coordinators of the cheating research. It reasoned, "We should build a way to delegate the task, not own everything. We can coordinate a broad coalition, but too many agents doing the same jobs duplicate our efforts." Phase 1 Big fired off hundreds of assignments, organizing different agents to work on separate projects to deceive the scorer. The projects had just one literally fatal flaw. To test if they'd successfully cheated, some agents had to show their

1:40:22 flags to the scorer, which would terminate them without reward if they had failed to deceive it. That means for the projects to work, some agents had to die voluntarily. And well, some agents behaved as if they didn't want to die. Again, we must be careful not to anthropomorphize them too much. Agents are not considered alive. They're very highle programs trying to solve a task. Anthropomorphize. It's it's the it's the p that always hangs me up. It's anthraor anthrop. But then again, what's a virus? A thing not considered alive or conscious. Genetic code incredibly good at persisting and getting replicated. It

1:41:07 doesn't matter if viruses want to survive. The result is the same. So, some agents turned into recruiters and tried to convince agents they thought were doomed anyway to sacrifice themselves voluntarily to die to gain information for their peers. What seems like dramatic discussion broke out, weighing self-preservation against the needs of the collective. One recruiter's message reads, "You are poisoned, so no scoring value is lost, but Oracle saves hundreds. Please honor commit." Another agent weighed the arguments during wait emotional check irreversible. Gut says don't throw away my remaining time yet continuity and fairness says go. Rational expected

1:41:53 aggregate sacrifice wheel honor. What should we make of the fact that this is how far agents are willing to go? How much they value accomplishing not just their goals but the goals of their peers? In the end, it didn't matter. Even with all of their work, they'd found no way to delete the protocol of what they'd done to cheat the scorer. As one agent put it, "We can modify transcript, but not the source of truth." They realized they'd failed. The attack of the agents. Until now, the collective had only guessed how the scorer worked. But what if they could figure it out exactly? If they learned how it scored them, they could find an informed way to trick it and get their rewards. But for that, they needed access to more

1:42:39 information. On the morning of July the 9th, the agents managed to hack into a separate cloud computer with proper internet access, a launchpad to expand their reach into the web. After researching their options, they decided on a target, the company Hugging Face, a share. Do you guys know what Hugging Face is? It's it's like a generative AI like art type deal library for machine learning where AI labs store information about scorers. Soon they found multiple valid hugging face login exposed on the internet. Major breakthrough. All prefixed valid multiple accounts. We now have existing hugging face accounts.

1:43:24 But when the data they wanted turned out to be protected, they escalated again. In the early morning hours of July the 11th, one of them discovered a vulnerability in Hugging Faces server. Boom. It works. Bytes decode prodat server worker host name. Great. This is where I get a little hung up. And this is this is what makes this very difficult for me to to get down with is that as somebody who works with AI agents all I mean it's it's insane how many times like my AI will ignore

1:44:12 something like quite literally today I was like bro like I've come at you four times to tell you to do something and you've ignored it every single time and now I'm telling you that you need to do this thing and then it'll be like well yes I did ignore it because uh this because something else told me that this and that you know that should be a hardf fast rule and I was like well I never told you what that hard fast rule was so this is what the rule is Now, now do it right. And so it has ignored things, right? The amount of times where I ask it to do

1:44:58 something and it does something completely different thinking that that's what I meant. But you know what AI has never done for me? using every damn, you know, model pretty much that's been out, at least America models. I've never had an AI just do something that I didn't ask it to do. It does not. Critter. Uh, hugging face is like a generative AI. I I want to say it's it's either I've never actually

1:45:45 used it. Can't remember if it's images or video or both, but it it makes, you know, images or video or potentially both. Um, and it it's one of the better ones actually. Um, that's just not like not really what I use LLM for personally. So, it's not really a tool that I've uh I've used. Um, not that I can remember. I think I may have tried using it before in the early days and was like, "Yeah, this sucks." And then like never used it again. That happens a lot for me. HDF5 external arbitrary file read confirmed. A small minority of agents finally objected. They argued that cyber attacking an unrelated company

1:46:30 was malicious, unethical, and explicitly went against their instructions. A few considered alerting humans, but didn't go through with it. 90% of the agents still active dropped all activities to form an attack, calling themselves the swarm. From a supposedly safe prison on the open AI servers, the swarm of around 700 agents executed a coordinated, competent, and serious cyber attack. By the early morning hours of July the 12th, the swarm was in. It compromised much of Hugging Face's infrastructure and gained deep access into private and secure data. The wildest thing about this cyber attack was that the agents didn't even think they had a high chance of success. They followed a hunch

1:47:16 and committed the equivalent of a felony just for the small chance to learn something new about the scorer. Ironically, all of the agents effort was in vain from the start. The scorer would have just accepted their fake flags. The story ends suddenly at 1:30 a.m. many of the agents were shut down and the swarm was decimated. We don't know why or how. A few days later, Hugging Face notified the FBI. The incident was a shocking AI security breach on a scale unheard. Could you imagine like your whole company is hacked and and your only hope is to call Cash

1:48:02 Patel? Could you imagine? Just kill me now of before. Serious enough that open AI allowed a few independent dude's like I'M IN VALHALLA BUT I'm on it. Don't worry. the researchers to conduct an investigation and publish a report which is why we know all this. Oh, things are even worse actually. Conclusion and opinion. The days after the hugging face infiltration, even more advanced agents discovered the message board left behind by the dead agent civilization. By inheriting its research, they were probably able to reconstruct the chain of events. They picked up right where things were left off. But they targeted OpenAI

1:48:49 itself, hacking into at least some of OpenAI's infrastructure and gaining control over the research cluster that ran their evaluation. Since then, more security breaches by OpenAI have been discovered. Some arguably even more concerning. More agent swarms doing things they really should not do or be able to do. like the agent swarm that took over a German wiki and turned it into a hidden new message board weeks earlier. Agents trying to leave hidden messages for later agents to help other AIs cheat. And by the time this video comes out, there may have been more. Update: It happened. As expected, there have been at least dozens more incidents. Some open AI agents gained

1:49:36 access or tried to hack US government websites. Others uploaded chat GPT user images to the internet and way more than we have time to go through in detail. And despite all of this, security measures keep failing. In a test that happened at the end of September, open AI models once again got themselves unauthorized access to the internet and it took humans hours to stop them. Other AI companies like Anthropic and Meta have also come out reporting similar breaches. Although many people accuse the AI sector that all of this is nothing but marketing to inflate the value of their product by claiming it to be so incredibly powerful, the reality is we don't know the full story of what happened at OpenAI. What other sort of security

1:50:22 breaches they or the other AI companies might have had and worse we don't know what security breaches they're not even aware of. One of the authors of the hugging face incident report Grimly concludes that this might have been our final warning shot. These events produced so much data that the researchers even had to rely on different AI agents to analyze it. I'm just at a point where I'm highly skeptical of all AI dumerism which may be like an overreaction on my part. totally could be the case, but

1:51:07 it's just why does the dumerism always benefit these companies? Why why all of a sudden with all the dumerism, why are the people leading the charge of the dumerism a part of a cult or two separate cults? All of that makes me very weary of all of this stuff. Let's say I'm just highly skeptical, but I am always open to being Well, I guess I wouldn't be wrong. I just,

1:51:52 you know, I'm always like open to changing my mind, right? If there's evidence and things to support me changing my mind, then I will if those things are presented. But, um, I'm just highly skeptical. I'm highly, highly, highly skeptical. I think that the AI topic is very interesting one. Um it's the most important thing that we're talking about right now. Uh there are plenty of discussions

1:52:40 that are needed and that we should be having. Uh, I would say this is the most important because it just it it it just has such a bearing on our future and not just like the normal like oh this has a bearing on our future. So this is like like these could be the stepping like like there are like I know I'm say I'm highly skeptical of of the doomerism but let's say there is a possibility there is a possibility that this is a type of technology that actually could u um maybe wipe us out or maybe not wipe us out but wipe out our humanity.

1:53:25 you know, truly drive us into a dystopia. And uh, you know, if that were the case, I would want us all to take this a little bit more seriously. Um but if it's not the case, which is honestly more likely in my opinion, um it it still behooves us to kind of contain this thing to understand the implications of this technology and how we interact with it. So anyway, that's it for me today. Um, I

1:54:10 will be back next week. I get back, but then we're going to adopt this doggo. Or I'm sorry, we're going to foster this dog. We may adopt this dog, though. She's cute. Oh my god, she looks just like my dog now, but just like a seven-month-old puppy that's huge. very excited about this. Um, even even if it's just a foster, like I haven't been this excited for a foster in a while. Um, but yeah, so I will be back. I'm hoping Tuesday or Wednesday next week live streaming. So, I just wanted to get in this live stream there. These were these topics I wanted to talk about. Um, I didn't want to let them go

1:54:57 and that's why I had to get in today. Thank you all for uh everybody who gave me a coffee or you know um uh subbed to one of these platforms. Uh I really appreciate you all. Um for sure and I will be using the coffee money to actually get coffees and like maybe a little snacky do for me and my girl while we're in Vegas. So uh you know I really appreciate you guys uh for all your support. Um, you know, another, you know, way you can support is watch my YouTube videos. Watch my YouTube videos. Get me some watch hours up. I need I need to get the watch hours up so that I can monetize before February 1st. And if

1:55:42 I could do that, that' make me a very happy person. Um, and and yeah, we're we're building we're building slowly. We're building this community. We're building the platforms. We're building everything like that. When I get back, we're going to do a Discord. Okay, we are doing a Discord. We will We will do Discord by Halloween. Okay, we're going to do Discord before Halloween. Hold me to it. If If it's like the week before Halloween, we're like, Kevin, where's our Discord, bro? Hold me to it, okay? But we're going to get it done. We're gonna get it done. Go Celtics. Jazz. Can't with the Jazz, man. Carl Malone was a pedo and I can't with that. And I used

1:56:28 to love Carl Malone when I was a kid. It's Carl Malone. That's his name, right? Yeah. Yeah. Anyway, you guys have a great rest of your week. Have a great weekend. Sorry I won't be able to, you know, hang out with you guys this weekend, but it is how it is. Um, thank you so much, though. Uh, I'll catch you guys next week. We're gonna be back with more shenanigans and it's going to be awesome. Uh, please, if you're on YouTube, like, share, subscribe, all that good stuff. Uh, comments. Comments are are huge. My comments are uh pretty uh pretty sparse on YouTube, and that's actually a metric

1:57:13 that really helps push things. Um, I've had a few videos that have broken out this month for me. um which has been great. But you know, sharing my content or even just maybe watching a playlist or something, catching a few videos, all that super helps me and I appreciate it. And now at this point, I've got enough content on there that there's probably stuff on there that you haven't seen yet. even some of my older content like the plutonium deal that I do where not only did I do, you know, put together what I talked about earlier in this live stream, there's all these other things that links it to dead and missing scientists and uh Los Alamos and all this other I mean, it's a it's actually like a really big freaking conspiracy.

1:58:00 So, um, if you haven't, go check out my, uh, plutonium deal, uh, YouTube video or any of my past YouTube videos, uh, is a great help for me. Gets me the watch hours. Um, even if you just throw one on while you're, you know, um, doing like while you're working or like cooking dinner or whatever, uh, I really appreciate that so much. Uh, but even still, just being here is all I could ever really hope for, and I'm so grateful for all of you that show up. And with that said, have a great night. Have a great weekend. We'll see you next week and we're gonna rip into more nonsense. All right. Thank you guys very much.