7/24/2026 - How Big Tech Lobbying and Government AI Regulation Threaten Open Source Innovation

Transcript

1. 0:00[music] [music] [music]
2. 0:20[music] [music] Good evening, good people. Welcome to
3. 0:35the AI Learning Lab. It's Friday night date night. Uh we've we've got a band of merry tick tockers. Uh they convinced me
4. 0:42to just let Tik Tok keep running tonight. So screw it. We're going to do it. Um they're not going to be able to
5. 0:48see anything. I'm not doing the whole actual full Tik Tok production where you can see things because that actually
6. 0:54requires thought. Uh, and I got no no uh possibility of thought, but welcome to the AI learning lab. If you're new here
7. 1:00to the AI salon, welcome. Uh, it's beautiful community, really amazing people. If you haven't introduced
8. 1:06yourself yet, over in the side channel on the lefth hand side, one of the spaces there is called introduce
9. 1:12yourself. I would encourage you to do that. Go ahead and introduce yourself. The people are really nice. Even if
10. 1:18you're like, but Kyle, I don't know anything about AI. They're going to make fun of me. No, they won't. You know why?
11. 1:24because we were making fun of them a month ago. You see how this works? It's it's it's like a pyramid scheme.
12. 1:32No, no one will make fun of you. Like everyone's trying to figure this stuff out. Cam Ken's in the house. Okay. So,
13. 1:38if you want to do So, tonight is an ask me anything. So, you can ask me anything, which the name implies. Um I'd prefer it
14. 1:47if you ask something about AI because, you know, AI learning lab and all, but you can ask me anything. Um, in this
15. 1:53corner right here is is the hyperlink to the chat. So, if you can't see the chat right now and you want to, it's in that
16. 1:59corner. Uh, if you want to pop this window out so you can have this window streaming live and then the AI salon in
17. 2:07another tab that's in this lower corner right here. You need to have two tabs open. One of them you're going to pop
18. 2:12the screen out if this on your if you're on desktop and then the other one you're going to have the AI salon over in the corner. Okay. Um,
19. 2:23there's a lot to talk about tonight. There's a lot going on. Um, I kind of lost my crap today in office
20. 2:33hours just yelling at the government. I just think the government's making really stupid moves right now. And uh, [laughter]
21. 2:42so I let the the crowd at office hours I I let him have a piece of my mind. [laughter] You're like, "How is that going to do
22. 2:51anything?" No, it's not going to do anything. It won't matter, but it makes me feel better. [laughter]
23. 2:59It's going to be a good one. All right. So, so let's tap tap tap and gift get Tik Tok to go viral status. All right.
24. 3:07Beautiful. I like that. All right. Is Chat GPT the best for video? Um, well, I don't know what you're trying to do with video. Uh, Chat GBT
25. 3:18cancelled their Sora video project. So, Chat GBT was the best for video when they first announced they were going to
26. 3:24do it, but then it took them 18 months to launch it, and then when they finally launched it, it wasn't the best. And so,
27. 3:30no, it's not. I would say right now Google's Omni is probably the best in terms of like physics. Um, and Seed Dance 2 is probably the best
28. 3:44for just video generation. I'd like to hear about more about what you said in office hours about the open source open
29. 3:50weights. Okay, I'm happy to do that. Um, all right, let's let's jump right into that. Um, oh, Steo Steo will drop by.
30. 4:00Oh, maybe Steo will drop by. I miss him. I know. I miss him, too. Okay. From Cam Keken, I'd like to hear more about what
31. 4:07you said in office hours about the open source and open weights. Okay. Trying to figure out where I even begin.
32. 4:25So, some context about what's going on right now is it became pretty clear at the end of last year and the beginning of this year that
33. 4:40all three frontier model companies, which right now the ones that I would cons consider frontier are OpenAI, the OG, Anthropic, and Google.
34. 4:52Um, Google's been a little light lately. Um, I think on purpose, but that's just conspiracy theory. Hey, Linda Sherman, what's happening?
35. 5:03Um but what happened was the CEOs of those respective companies and then Demis who's the CEO of deep mind so he's
36. 5:13sort of co-CEO with with Sundar Pachai they all started talking they they started shifting their tone they stopped
37. 5:23talking about AGI they started talking about super intelligence and they started kind of hinting that the models that they weren't releasing
38. 5:33yet were really powerful and they always do that. It's always the hype train but something was different. They were
39. 5:39talking different publicly. Then shortly after that, Dario Amade announces, "We've got this model called Claude Mythos."
40. 5:54And sorry guys, it is so powerful it can hack any system, and it's so dangerous that we're just going to give it to 42
41. 6:02of our favorite companies so that they can shore up their defenses. But you can't have it yet. They made a really
42. 6:09big deal of this. And then right around that time, shortly after that, they they I I think that took about a
43. 6:18month for, you know, it was just out there and I guess these companies were doing things with it. And then they they
44. 6:26do these two things concurrently that are kind of weird. They launched this model called Fable, Fable 5, which is Mythos. It's the same
45. 6:38model as Mythos. It's a 10 trillion parameter model and then they release it for three days and during those three days they say to the
46. 6:50government or maybe right before the three days they say hey government these models we're making are so
47. 6:57dangerous you should probably regulate us. And 3 days later, the the government said, "Hey, one of
48. 7:12your models did something bad and so we're going to export control you." So So they took it away, right? You
49. 7:20remember all this? So they took Fable away from us for about three weeks and then they gave it back and it's completely hobbled.
50. 7:30And right about the time they give it back, Dario Amade says, "Hey, by the way, um, Alibaba, who makes Quen, they
51. 7:39distilled, they stole our model. They stole our model with a distillation attack and they stole it. And that's
52. 7:48we're Americans and we've got American AI power and you shouldn't take our models, damn it." Meanwhile, it turns out that they
53. 8:00distilled Opus 4.8 from Quen from Alibaba. So, they're all stealing from each other. And by the way, if you don't know
54. 8:08what a distillation attack is, it's using a large language model and taking the answers of very specific prompts and
55. 8:17basically using those answers to train a new model. So you're basically, you know, you're you're basically doing
56. 8:24what Anthropic did in the first place, right? They took all of the data on all of the internet, everything we've ever
57. 8:31written, and they train their model on it. So they did a distillation attack on us, and now they're saying there's a
58. 8:38distillation attack on on Chinese models. Then this week, ChatgPT6 made a boo boo
59. 8:52and OpenAI put out this blog post that said, "Hey, uh, oops, sorry, hugging face. Uh, we told our model that it was not
60. 9:03supposed to escape its little pen, but we wanted it to do a an eval test, an evaluation test, and get a really good score on it."
61. 9:12Turns out it was so smart that what it did was it tunnneled out of its machine that didn't have internet access, found
62. 9:18some other machines inside OpenAI that did have internet access, found one of those that was vulnerable to a hack,
63. 9:25hacked it, and got out on the internet. Now people are saying it escaped. It didn't escape, but it did hack a machine
64. 9:32and get internet access. And then it thought because it's thinking Who's got the answers to this eval test? I know Hugging Face.
65. 9:47So, it started hacking Hugging Face. Hugging Face caught it cuz they're badass tech people and they understand
66. 9:56when they're being attacked by an AI thing. I don't know how those people do that, but they do. They're good at it.
67. 10:03And so Hugging Face started getting attacked by an AI agent and they're like, "Holy [ __ ] we got to stop this."
68. 10:11So they went to Fable, Fable 5, this new mythos class model, and they said, "Hey, Fable 5, help us stop this attack
69. 10:22from a sophisticated AI model." And Fable 5 said, "Oops, that's too dangerous. you can't use that prompt
70. 10:32because we've got safety guard rails which the government imposed because Dario Amadeay asked them to impose them.
71. 10:40And then I guess they tried to use open AIS and they had safety guard rails too. And so Hugging Face ended up using a Chinese model
72. 10:50to protect itself from the American model that got out that went rogue. [laughter] Now, it protected itself
73. 11:07because the safety guard rails on the American models prevented them from being safe from attack. And as that's happening, our government
74. 11:24is saying, "Hey, Alibaba and now this new KI3 thing stole American IP through these distillation attacks." By the way, all of the major
75. 11:35models do distillation attacks on each other. It's how these models learn. It's why they always have the same thing two
76. 11:42or three months from each other. So, this is nothing new. But what this actually is is anthropic and open AI manipulating the government
77. 11:54to prevent competitive attacks from Chinese and open source models. It's business in America 2026. This is how we do it.
78. 12:11The government is now saying we may block Chinese models that stole from Americans. We may sanction or block open-source models.
79. 12:22So, you know, like the model that I have, the Quen model that I have on my Mac, that may become illegal, which do you
80. 12:29remember in the 80s and 90s when the government tried to make encryption illegal because it was we it was a
81. 12:36weapon. They classified it as a weapon. Like what we learned was that you can't you can't do that to software. It's
82. 12:46software. It's free speech. So, so we're in a situation where the major model companies have hobbled
83. 13:02their strongest models because they're too dangerous. the government's about to block access for capable models
84. 13:09that, you know, that could help us defend against [ __ ] because these two are hobbled. Um, and and if if you're an individual
85. 13:19just wanting to have your own AI inference, that may be a thing that goes away. Um, this is all stupid. This is all
86. 13:36preventable [ __ ] But it's the way America works right now, right? Corporations are individuals. Citizens United, thank you very much.
87. 13:51So, Anthropic just decided to up their $20 million contribution to the midterms to $40 million. And both Anthropic and Open AI, I think
88. 14:02OpenAI's given $100 million to the midterms. Both of them have offered 5% stakes of their companies in the government
89. 14:14so that the government passes regulation that is favorable to them and does things like blocks their competitors.
90. 14:22The only problem is that [ __ ] with American innovation because American companies aren't going to use the Frontier Labs models if
91. 14:34they're so risky that they're going to be taken away from us. And and they're also not going to use them if they're 10, 20, 30 times more
92. 14:42expensive than an equally capable open-source model unless the government forces them to do that. And then these
93. 14:49models are so crippled with their safety guard rails that they become unusable. What it's going to do is it's going to slow down adoption
94. 14:58probably by a year or two. So it so what a way to start out the night, Kyle. This is uplifting.
95. 15:13[sighs] Don't they effing get it? They need to defund Citizens United. Clearly not. Apparently, so here's what I learned,
96. 15:22Chef Kelly, that um politicians like receiving $40 million better than they like receiving $5 from actual individuals.
97. 15:37I don't know. So So apparently they're just going to keep doing that. [laughter] If they pull their support, it gives us
98. 15:47back our power. Oh, so then, okay, so there there is a there's a kota to this. The kota is that Jensen Hang, the CEO of Nvidia,
99. 16:01sent his very first tweet, his very first post on X. Hey, Tik Tok. I'm not used to you being here. You interlopers.
100. 16:12You damn behind the scenes crackerheads. I can see how that math is nice. Yeah. The the $40 million versus the $5. Yeah,
101. 16:22it's, you know, it's just kind of easier if you just take the the big check and just deposit that right in your campaign
102. 16:28thing instead of having to like cash is dirty. You got to count it. Just take the big check. Um, so Jensen did his very first tweet
103. 16:41and it was basically Nvidia signing on to a letter with Microsoft and Meta and just a bunch of big companies
104. 16:50that are all saying uh, keep open source safe. The internet was based on open source like like all good things were based on
105. 17:01open source. Do not do this to us. Um, so, so a lot of big companies right now are signing on and saying to the
106. 17:09government, get your head out of your [ __ ] ass, please. Um, I'm not confident. I'm just not confident. Frostbitten,
107. 17:22Miniax plus Deepseek plus GLM plus Kimmy are distilled from anthropic models. And Opus 4.8 is distilled from Quen. [laughter]
108. 17:36I get that. But but Frostbitten um the the uh the Grock the Grock models were distilled from these two. They've
109. 17:47they distilled from each other. This boils my blood. I know. I know. I know. I know. It boils my blood, too. Okay,
110. 17:54let's get some more questions. Anybody have any questions? Pop them in the comments below or if you're on TikTok if
111. 17:59I see them. We don't have a producer, so he's not capturing Tik Tok ones. Um, let's see. This was from Philic from uh from the
112. 18:11uh what you call it from the back to basics. So, let me let me throw this in here. Version two is a distillation of
113. 18:20Bllandens. I Yeah, you know what? We should have a distillation of Bllandens. Okay, what's this question?
114. 18:29Can individuals get downgraded into a lower economy if they don't keep up with AI? Um, yes. I think they can. I don't know
115. 18:39how I'm going to be able to use a model that donates to Citizens United. Well, no. No, Kelly. They're not they're not
116. 18:47donating to Citizens United. Citizens United defined a corporation as an individual that can that can give unlimited funds. [laughter]
117. 19:00So, so all all anthropic and open AI are doing is taking advantage of the law that's just sitting there. We can act as
118. 19:09an individual. Yeah. C CPAC. Yeah, exactly. Okay. Anyway, can individuals get downgraded into a lower economy if
119. 19:17they can't keep up with AI? So if you don't know what what this question is about, so about two months ago, I came
120. 19:24up with this concept called the seven economies. And the basic idea is AI is accelerating. Uh it's getting better and
121. 19:32better and it's accelerating. And most of the world and most of our country has no idea what AI is. They
122. 19:41don't use it. They don't pay for it. They're afraid of it. or they're just completely ignorant to it. And so what
123. 19:50happens if you have something that's accelerating away from you and you have another part of the thing that's
124. 19:55standing still, what's going to happen is the world is going to kind of fragment, right? All this pressure is going to pull the world apart.
125. 20:05And I think what it looks like is very distinct economies based on level of AI adoption. And let me see if I can
126. 20:17find if I can find the image, I'll share it. Um, let me see. Economies. Yeah, I can. All right.
127. 20:32And I don't know if this is going to stay shared that long. Okay. So, oh wait, that's weird.
128. 20:52That's a really low res version of it, but that's okay. Um, hang on. Okay, so so this is the seven economies slide. Let me get rid of this.
129. 21:04Um, so the basic idea is the four economies to the left, one, two, three, and four are are basically businesses that are doing what they're
130. 21:15doing today. Essentially continuing to do what they do. So economy one is the analog economy. This is I'm going to go
131. 21:22grow blueberry, go grow blueberries, you know, screw your technology. Economy two is we're not even going to pay attention
132. 21:29to AI. We're just going to do what we did before. Economy three is we're going to use AI for some efficiencies.
133. 21:35Economy four is we're going to use AI for some efficiencies, but we're going to do a little bit of innovation,
134. 21:41but still basically be the same business model, right? So, so those economies are most of what exists right now and
135. 21:50they're just sort of plotting along. And then you've got these these three AI economies that are that are kind of
136. 21:57distinct, but they're they're getting faster and faster and faster. So the first one is what I'm calling the AI
137. 22:03aggressor economy. And that's where you take an existing company that might even be profitable and growing and you you
138. 22:10completely just burn down the org chart and reinvent it. And so Jack Dorsey just did this with block economy. Six is
139. 22:18people that get laid off go off and learn AI and a bunch of them pal up and say, "Hey, we could start a company that
140. 22:23competes with our old company. Let's go do that." And so economy six is going to be like little mavericks and little
141. 22:29rebels learning AI and going back and [ __ ] with economies two, two, three, and four. And then economy 7 is the
142. 22:36agentic economy, right? So this is a a desk a chip on your desk running 50 agents running two or three businesses
143. 22:44for you where you put in an hour or two a day and it just goes off. Okay. All right. Fable 5 safety guard rails for
144. 22:52bio and coding was tamed down because it was so good. Uh it was like Opus 4.5 and Sonet 4.5 where it would do anything you asked it. Huh.
145. 23:05The original Fable 5 was amazing for biological work. Okay. So, so that's the seven economies. So, Philip's question,
146. 23:14if I can find it again here, can individuals get downgraded into a lower economy if they don't keep up with AI? Well, [sighs]
147. 23:28the seven economies, I'm thinking about those more like businesses. If right now you decide, I don't give a [ __ ] about AI. I'm not
148. 23:39going to follow it. I hate it. I don't like it. I don't care. It's not like you're going to get downgraded. Like, you're already
149. 23:49working in an economy that doesn't have that isn't AI amplified. And so, you're going to stay there. But what's going to happen is those
150. 24:01economies one, two, three and four. Well, so economy one is special. The analog economy I think is about to grow and expand. I
151. 24:11think I think analog businesses, farms, artisal butter makers, live music walks in the park for money. Um, the
152. 24:23analog economy, I think I think this AI madness and this AI acceleration is going to is going to freak a bunch of
153. 24:30people out. They're just going to tap out. [ __ ] it. I'm out. I'm going to the beach. I'll figure out how to make money
154. 24:36at the beach. They'll open an Etsy shop. Whatever. It's done. So, I think I think economy one grows. But economies 2,
155. 24:45three, and four are going to basically be under attack from economies five, six, and seven. like a little bit at first and then relentlessly
156. 24:55because economy 7 in particular, the agentic economy is basically going to figure out all of their business models, replicate them,
157. 25:05and improve them. And so you're going to have like a 17-year-old snot-nosed kid that takes the time to learn OpenAI Codeex or Jack
158. 25:15Dorsey's new Buzz tool, which I'm going to talk about tonight. and and they're going to say to their agent, "Hey, go figure out how to do
159. 25:24fertilizer arbitrage. I heard that's a good business." And like all the fertilizer arbitrage people are going to
160. 25:30be like, "Yeah, AI will never get us. This is about backroom deals and fertilizer and arbitrage and we're those
161. 25:40guys, so [ __ ] AI." And then all of a sudden their business is going to start disappearing and they're going to be
162. 25:49like, "Where's our bis? Where are the pe where's the arbitrage? Oh, that's just this other company now. What's that
163. 25:56company?" I don't know. It's started by some 17-year-old kid. What? So, so individuals that don't learn AI are just going to
164. 26:09end up if they're in economy one, there'll be some version of fine. But if they're in economies 2, three, or
165. 26:15four, and they're not paying attention to it, those those businesses are just going to keep getting hammered and
166. 26:20hammered and hammered. Um, so yeah, so I think it's probably I think it's foolish to not learn this [ __ ] Okay,
167. 26:28I can show you the list of models I have tested and ways you can tell which ones were distilled from who. That's cool. If
168. 26:35you ask some of those models through the API, just asking them what their name is or who they identifi with, they will say
169. 26:44Claude. That's hilarious. Um, questions on Opus 5, Opus 5 versus Fable. Uh, I don't know yet. I haven't played with
170. 26:52them. Um, I have not played All right. Fable 5's guard rails. Blah blah blah. Okay. Let's see what we got here. I just set up a fertilizer
171. 27:02arbitrage conglomerate. Yeah. See, see Charles Adams. He's over there writing books. He's teaching people. He's making jerky in the kitchen
172. 27:13and just as a side quest starts a fertilizer arbitrage business. Like that's going to happen. Cuz I'll tell
173. 27:21you what's going to happen. And there's going to be some [ __ ] pissed-off farmer that the fertilizer arbitrage boys did a little too much
174. 27:28arbitroization and the farmer's fertilizer cost too much and now he's not turning a profit on top of all the other shenanigans.
175. 27:40But that farmer at night has been playing with chat GPT. And at some point that farmer is gonna go, "You know who I'd like to get rid
176. 27:51of? The [ __ ] fertilizer arbitrage boys." He's g [laughter] he's going to go, "Hey, chat GPT." Yeah. How can I help
177. 28:01you? Hey, can we come up with a business that cuts these [ __ ] out of my life? Sure. Here's what we do. Would you
178. 28:07like me to spin up an agent to go start that business for you? Sure. [laughter] Ah, [groaning] now you're getting it. Exactly. Exactly.
179. 28:22Vicky Baptist, I have Chat EPBT working in Tinkercad for me. Good lord. [laughter] Source camps over in Tik Tok. You Tik
180. 28:31Tok people. So, listen. This is not official. This is a This is a rogue effort by Tik Tockers who made the
181. 28:40mistake of asking politely if I could just leave this turned on for tonight. And and I was in a weak moment. I was in
182. 28:47a weak moment and I said, "Yes, this is this is don't you don't you dare think that this is the a new thing. This is
183. 28:55this is rogue behavior and you know it. Don't tease us." Okay. Frostbitten like how the medical health care system in
184. 29:03quotes said no. No to me so many times. Something isn't when they don't test to see if it is or isn't. Yeah. that maybe
185. 29:11I'm working on something that would dis disrupt medical GP fields. Yeah, exactly. Oh, wait. We can go on TikTok.
186. 29:18See you there. [laughter] Yeah, Vickiy's coming. Vickiy's coming over. You you big you big lugs. Um, okay. Let's get another question here
187. 29:27from Mary Carol. Okay, copy. I'm going to put it in here. This is from back to basics night from
188. 29:43Mary Carol. Explain what you mean by use a local model for security purposes. Okay, so let me share my screen. Let me get rid of that.
189. 30:02Wait, I think I'll do a window because I think they share a little better. Okay,
190. 30:20here we are in chat GPT, right? I got to do something though. [whistles]
191. 30:38>> [whistles] >> Okay, so here we are in chat GPT new chat.
192. 30:53We really do miss it here. I know. I know. I apologize. Um, when you're in chat GPT and you type in a prompt,
193. 31:08your prompt and that whole conversation is sent, the text of that is sent to OpenAI's servers and then OpenAI has these big data
194. 31:23centers that allow Chat GBT to answer you And so everyone that's saying we got to get rid of the data centers are also going to be
195. 31:32the same people going why doesn't my chat GBT work excuse me excuse me I was using chat GPT why doesn't my chat GPT not work oh
196. 31:45sorry ma'am we don't have any data centers in your region well yeah I got rid of them I yeah I know that's why we
197. 31:51can't anyway that's neither here nor there the problem with that is that Open AI owns your data and they're training their new models on
198. 32:08people interacting with their existing models. So they're they're taking all of our prompts that go into their models
199. 32:16and they're taking all that data and that becomes new data for the new models, right? [laughter] So if you put in like your social
200. 32:24security number and your medical stuff and all that stuff, it's basically getting trained and ingested into the
201. 32:30new model. And so it's just from a security and privacy standpoint, it's an absolute shitow because used to be time was that when you
202. 32:44are a big data company, only big data companies could could process and make any sense out of big data or some very talented data
203. 32:55scientists. But you had to be a data scientist or a big data company to make sense of any of the data. Well, now with
204. 33:02AI, any old schmuck with chat GBT can can parse through [ __ ] tons of data and make sense of it. So all of the context that we give away
205. 33:15is actually really valuable. It's valuable to us, right? And the the the risk of that is one is your stuff might
206. 33:23not be secure. you might, you know, risk your privacy. Um, but it's also if if Open AAI decides to shut down your
207. 33:38account, all of your history and all of your context and all of your previous chats are no longer yours. They're gone. They're gone. So,
208. 33:51if I Let's see. I can't do it here. I got to change how I share again. I'm gonna go show you something.
209. 34:04Do this. We're going to do this. We're do this. We're do this. Okay. So, now I I'm going to go to a new place. I'm going to go here. Go.
210. 34:23I'm going to go to this tab right here. You can just ignore this little dude. That's my little pet. Put him down there.
211. 34:35Reload. Last model. Pick a model. Okay. So, let me get rid of this. How do I get rid of this? Like that. Okay. So,
212. 34:50let's see. How do I do a new search or a new chat? Let's see. Do I do is this new chat? No, hang on a sec. I'm going to turn off
213. 35:02that pet. It's driving me crazy. Hide pet. Oops. Hide pet. Okay, so we go plus. Where do I do new chat?
214. 35:15Oh, here's new chat. Okay. So, I'm going to do a new chat. So, here we are in a thing that looks like chat GPT. Okay.
215. 35:27But if I click this model loader, I've got Quen. I've got this is this is a Quen model that is uh it doesn't have any guard rails on it.
216. 35:39This is one that you can ask it all sorts of nasty things and it'll give it to you. Um, this is chat GPT's open
217. 35:45model. This is Google's open model. But this Quen 3.6 one here. I can put my This is my context window. Put my
218. 35:53context window up to 81. I'm going to load the model. So, you can see down here it's loading the model.
219. 36:02Although you can't see it because Mary Carol's thing is there. It's loading the model. Okay, this model's loaded. So now I can say um
220. 36:12here is my private medical info and my social security number. Can
221. 36:26you please um tell me uh something interesting about them? All right, stupid prompt. I cannot provide
222. 36:39information about your social security. But see how quick that was. That chat GBPT like response did not
223. 36:47come from OpenAI. It did not come from Anthropic. It came from a MacBook Pro that's sitting on my desk behind me. So all of this
224. 36:57all of this interaction I can do with these models on my local machine never goes to the cloud. It never goes
225. 37:04anywhere. In fact, that machine could be completely disconnected from the internet and I could still do this.
226. 37:12And so these models are getting better and better and better and the computers are getting stronger and
227. 37:18stronger to be able to run better and better models. This is the [ __ ] that the government is saying we should make that illegal or we
228. 37:27should sanction it to the point that it'll it'll be as expensive to use as anthropic. Right. So, $50 per million tokens. I
229. 37:38could run this thing 24/7 and the only thing I'm paying for is electricity and all of my data stays on my local machine.
230. 37:47So, that's what a local model means. It it's you can do things similar to chat GPT and you know these these cla models.
231. 37:56They're not nearly as good as the frontier models, but us chickens, unless we're doing stuff like Cam
232. 38:04Katkin's doing with DNA and biology and things like that, we don't need the big models. If we're just doing little
233. 38:11automations and and copywriting and things like that, we can absolutely get away with these local models. So,
234. 38:17anyway, all right. What do you think about the people saying their electricity bill went up? Is it a lie?
235. 38:22Well, it's probably not a lie for two reasons. Um, one could be data centers, but I but it would have to be that there
236. 38:30are data centers near the electricity plant. Like data centers have been around for a long long time. Do you ever watch Netflix?
237. 38:40Guess what those run on? Data centers. You you ever do a YouTube search and watch a YouTube video? Data centers.
238. 38:50You ever use Google? Data centers, right? like like data centers are the backbone of everything we do. Um
239. 39:00the amount of power hunger that these chips are doing depending on your region it could absolutely be that you want to
240. 39:08know why our electricity is going up in Denver by by 50%. It's been 100 degrees. It was 100 degrees eight days straight here and we've been
241. 39:20in a heat dome in the high 90s for just under a month and that is destroying our power grid. [laughter] So, so it could
242. 39:32also be that uh it's not cool anymore anywhere. Um, okay. What else do we have here? Other questions? What other
243. 39:42questions do we have? Let me bring this back so I can see you all.
244. 39:56Kyle, did you try the obliterated Quinn 3.6 model yet? Yes, I did. So, the obliterated model, I asked it I
245. 40:06asked it to teach me how to make a fertilizerbased product. And it did exactly how to do it. [laughter]
246. 40:22[laughter] Um, that's the only thing. You know what's funny is I'm such a I'm such a [ __ ] I I um I asked it that and then I
247. 40:33was like, I'm not going to ask it really nasty stuff. It's going to make me uncomfortable. Um, but yeah, it appears
248. 40:39that it'll just answer anything. This is heat in Colorado. I mean, yeah. Yeah, it's it's killing me. I
249. 40:48know. No, it's crazy. It's really really crazy here. Um, okay. I got some other questions here. Let's see. This is interesting.
250. 41:02I'm a little spicy about this tonight. Okay. Who should decide how powerful AI models get released? The companies, the
251. 41:11government, researchers, or the market? The market. Because what happens is this
252. 41:39Well, the companies in the market are kind of the same thing. The companies are well aware. Sam Alman is well aware.
253. 41:51Dario Amade is well aware. Sandra Pachai, um the guy from Deep Mind, Elon Musk,
254. 42:06even Mark Zuckerberg are aware that they can't just it is not in their best interest to just release
255. 42:12models that are just patently dangerous. But there's a really fine line that they have to walk. And Sam Alman has said,
256. 42:22I think that we should incrementally release technology into the wild so that people get used to it and learn how to
257. 42:30use it. And in the case of something like Mythos that's really really powerful, they sort of released it to a
258. 42:38small group of people to say, "Hey, go go go defend yourself against the capabilities of models like this because
259. 42:44more of them will be coming." If you rely on the government to do it, the government doesn't understand what [ __ ] Wi-Fi is.
260. 42:55Like, they have no idea what AI is. So, what's going to happen is some lobbyist is going to come in and
261. 43:02oh, these things are really dangerous. You got to lock them down. And all of their constituents are going to be like,
262. 43:08yeah, we don't want to be killed by an AI robot. Lock it down. And so the government's just going to
263. 43:16put arbitrary rules on that are going to hobble us being able to release models into the wild. Now, Anthropic is asking
264. 43:25for that because Anthropic wants to write the rules that regulate all the other companies. So, so Anthropic wants a regulatory
265. 43:38environment where all of the other companies play by the rules and infrastructure they set up. That's not good
266. 43:47because what if they're wrong, right? So we want these companies comp competing and then we also don't want to
267. 43:54happen to hugging face where the government regulations on we're going to release these models but we're going to
268. 44:00so restrict them that you really can't use them for anything useful if it's like science or or health. They're useless for those but they're
269. 44:11really powerful for poetry. You can go write a bunch of sonnetss. enjoy. [laughter] But god forbid you want to do any
270. 44:23science on these things, they they can do none of it. So we can't protect ourselves from outside models. So what needs to happen is let the
271. 44:32Chinese compete with us. Let us compete with them. Let these companies compete with one another. And then as people start to use models
272. 44:40for nefarious acts, then people with stronger models can use those models to prevent the bad guys. That's the way
273. 44:48it's always worked. It can still work that way. But if you put, you know, restrictions around these companies with clueless legislation,
274. 45:01I don't know. We're in trouble. All right. If anyone has any questions, pop them in the comments. I'm happy to answer them.
275. 45:13Answer them. Yeah, boys. Oh, this is an interesting one. Hi, sleepy d.
276. 45:31Okay. Should companies let employees use AI agents freely or do they need rules before people start assigning real work to bots?
277. 45:51No, you shouldn't just give employees untethered agents and turn them loose. No. Apparently, um, Chat GPT 5.6 Soul
278. 46:07is semi-regularly erasing people's hard drives, deleting all their files. >> [laughter] >> So, so
279. 46:22what companies need to do with AI? First of all, they need to have actual governance. They need to have people
280. 46:27like Vicky Baptiste and and Chris Bruno and and people who understand what AI is, what it makes possible, what it's not good at,
281. 46:43and they need to have those people that have a clue help them write governance policies so that there's actual rules for what
282. 46:50you can do and what you can't do. that there's actually infrastructure in place that says, "Hey, if we want to protect
283. 47:00this data, here's how we're going to protect it for this kind of data. We're going to run all of our our AI locally in our own
284. 47:09private cloud." And people can use they can do financial data, but they have to do it in our own private cloud.
285. 47:17If they do it in anthropic or chat GPT or an agent, they'll be fired. So [snorts] you you absolutely have to
286. 47:26if you're in a company understand what the hell this stuff is. Now these agents, these agentic harnesses they call them
287. 47:36where you can just give it a goal and then it goes off and it spins off sub agents and they go out and do a bunch of
288. 47:43work and then come back and do a bunch more work and then go back out and do. If you don't have governance around that, you can't just No.
289. 47:55There's an old saying that I'm really fond of. Don't be an idiot. Like, don't be an idiot. [laughter]
290. 48:12Oh my god. Blue age. I agree. Mary Mary. Yes. Yeah. Don't be an idiot. You might be talking to each other and not me, but that's okay.
291. 48:22Um, I think Elon shows that trillionaires are possible and they still keep they still they're still playing to keep away the mere
292. 48:31billionaire air. Yeah. You know, I the thing about Elon is he he clearly wants power, but I think he doesn't really
293. 48:42give many [ __ ] about money. Because if you look at his history, you know, when he sold PayPal, he got $120 million and he put
294. 48:5560 million into No, he put 40 million into Tesla and he put 80 million into SpaceX. He took his whole 120 million and he
295. 49:08just bet them on two companies that were predicted to fail. So, I don't think he's all about the money. Um,
296. 49:16I really do think he is about big ideas. Um, and he also is fairly confident that we'll all be billionaires and
297. 49:28trillionaires in the not too distant future. That that robots are going to build us anything we want. He's confident that
298. 49:38the robots are going to um run out of things to do for us. you know that that like what we'll say is we're like, "Hey, you know, I got I
299. 49:49got an acre of of abandoned land up up on the hill. Could could we turn that into a giant swimming pool, pond,
300. 49:57recreation park?" And then the robots are just like, "Sure." And they just walk up the hill, work for two weeks, come back down, it's done,
301. 50:06like, "Oh, cool. Could Could you we go get some jet skis?" "Ah, sure. Let me just run down to the jet ski mart. Let
302. 50:13me have my agents go order those. They'll be here tomorrow. For Elon, confident. Confidence is a negative signal for the possible.
303. 50:23Um, all right. I think Trump is going to be a trillionaire before his term is over. If he's got anything to do with
304. 50:30it, he will. All right. Let's see. Um, debating. Uh, let's see. [cough] Hang on. When you have nearly a trillion dollars,
305. 50:41you don't have to give a [ __ ] as much about money so much. Yeah, I know. Um Elon will have some great people
306. 50:48working for him just like Gates did. All right. I want them to load the dishwasher my way. Well, I'll tell you.
307. 50:57So, so let me talk to you about like that. That's that's I I know you're being kind of kind of snarky there, Mimi, but um
308. 51:09they will do what we want them to do. Like like I I did um I'm working on a thing right now where I'm having Codeex go out and research new
309. 51:22sales targets for our sales guy at StoryVine. And I gave it the the whole here's what I think you should do
310. 51:33and it went off and it started sending him 10 names a day and and and the first list of 10 was just 10
311. 51:47people from our database that are already in there that Rich is already talking to. And then so Rich give gave
312. 51:54me feedback and then Rich gave me more feedback and then Rich said add a LinkedIn le link and then Monnique my
313. 52:02co-founder said oh can you validate the emails before we get them? So now Rich is getting a list every morning every
314. 52:09weekday morning of 10 people that have this whole list of criteria. Here's their LinkedIn profile, their emails,
315. 52:15their phone numbers. Here's what's been validated. Here's what I think you can sell them on. Here's why. It's like
316. 52:21this, it's this incredible thing. And it got better in like four days. So, it was like it would send a thing. I would get
317. 52:29feedback. I would literally just copy and paste the feedback into this thing and it would get better. And then they'd
318. 52:35say, "Oh, that's better. Two of them are really good. Now do this. Now do this. Now do that." And I would do that and they would get better.
319. 52:43They will just build the jet skiis. Yeah. Exactly. So, what will money become? What what Musk is saying is that
320. 52:50money will essentially become valueless. What what Musk is saying right now is you already just print money anyway. You
321. 52:55might as well just print it because what happens is um in theory like I I I don't quite know how we get from the economy
322. 53:04we have now until this mythical economy of hyperabundance. But but assuming we make that transition, um
323. 53:16if you want a business, you'll just tell your agent, go start, run, and ma, you know, manage and and grow
324. 53:26a business that does this. And that could be a physical business. It could be an import export business. It could
325. 53:32be, I don't know, all sorts of businesses. So, it's just going to go do that. And where most of the transactions
326. 53:40are going to happen is between agents. So my my company will go out and do something and say, "I've got this thing
327. 53:49to sell." And there's some other agents will find it and go, "Oh, I need that thing. My my boss says, I need that.
328. 53:56I'll pay you this." And they'll do negotiations and it'll just be what it is. Oh, thank you very much for that. Um,
329. 54:05and then when we have physical robots, we'll just say, "Oh, go build me a house. Go build me a this. Go build me a
330. 54:11that." Well, how are you going to get lumber? I just I'm going to have the robot go home to Home Depot and buy some
331. 54:19chainsaws and a and a sawmill off marketplace and then go set them up in the woods and cut down all the trees and
332. 54:29harvest them into wood and then go build the house. So, it's just like all the stuff's just going to be there. And
333. 54:39I don't know. I'm not an economist, but that's what they're all saying. 3D print the wood. Exactly. So, it's going to
334. 54:45look like a fritter as we transition paradigms. Hey, wait. A filter. A filter. Yeah. The transition is the part I don't
335. 54:59get. The transition is the part because if you run a lumber yard, you're kind of relying on big lumber
336. 55:12companies that have, you know, thousands of acres and thousands of machines and thousands of trucks that bring boards to thousands of
337. 55:21processing plants that send their [ __ ] to thousands of distribution centers, that send their [ __ ] to Home Depot, that
338. 55:28puts it on the back of your truck and sends it to the construction site. And then if you could just not pay all
339. 55:38of that and just go do your own. I don't know. I don't know, man. I don't know. I don't know how this shit's going
340. 55:47to work. All right, let me get rid of some stuff over here. Got another question. I think I'm going to show you Buzz, too. Ah, this is good.
341. 56:08Things break when supply chain supply chains get disrupted. Broken things have consequences. I listen, I don't disagree. Like I said,
342. 56:17I'm not an economist, but it's like Musk is confident on all this, but he's also been confident on, you know, we would have self-driving
343. 56:27cars in 2015. We still don't. We sort of do. We sort of do. Okay. If AI can write code, make slides, summarize meetings,
344. 56:38search files, and manage workflows, what are humans supposed to get better at next? So this is what the great repurpose is all about.
345. 56:51We live in a world where our college degrees and our expertise and our tasks that we do every day
346. 57:08are the indicators of your value. Right? I went to Harvard. I'm good at spreadsheets. I can do financial projections.
347. 57:22And you're going to pay me because I went to Harvard. You're going to pay me more. Because I'm really good at
348. 57:27spreadsheets. You're going to pay me even more. Because I'm good at predicting the future. You're going to pay me a boatload.
349. 57:37AI doesn't care that you went to Harvard. AI can build you a spreadsheet. not only use it, but build you one and it's going
350. 57:46to be better at that future projection [ __ ] than you than you will ever be. So all of that stuff, all of those
351. 57:56outputs that were the indicators of value, I'm a business with money. I need future projections of business.
352. 58:07I'm going to go look for a human being that's got these things. They got to be from Harvard because I've I've worked
353. 58:14with Stanford brats. They suck. Colombia, the journalism school. No, they're not doing my financial projection. We're going Harvard, right?
354. 58:27And they got to be able to do pivot tables. And then I got to have a brain be able to use all that stuff in
355. 58:37conjunction to make my thing. And now all of a sudden AI comes along and the same person with money is like, "Huh, I
356. 58:44wonder if I could just do that. Could I just ask it to Oh my, it did it better than he did." Huh? So if we live in a world where the thing
357. 58:56that used to have value goes to zero, what is a human to do? Well, humans have ideas. Humans have intuition.
358. 59:13Humans have taste. Humans can interact with one another and understand one another and inspire one another
359. 59:25and then the machines can do all this stuff. So, I think what happens is the role of humans shifts. It it it's almost
360. 59:33like we all become creative directors in some weird kind of way or entrepreneurs. We just it it's going to be our job to
361. 59:42come up with the ideas. You know it'd be cool and you come up with some stupid [ __ ] idea and then AI goes and does it for
362. 59:49you. And if it's actually a cool idea, people will go like, "Ah, that's a really good idea." Huh? Who ever thought that pink jell-o
363. 59:58and parachuting would go together? >> [laughter] >> I don't know. That's humans. Community, caring, giving, creating art, dance, taking care
364. 1:00:12of our loved ones, animals, environment. At least that's what I want to do. Help rebuild the reefs. Yes. Yes. The humans
365. 1:00:19here, here's the thing. Here's the thing that is is counterintuitive. AI is going to take our jobs. Most of our jobs suck.
366. 1:00:30If you're one of the rare people that does a job that's in line with who you are and fulfills you spiritually and
367. 1:00:39[laughter] and you know financially you're in the you're in the 0.00001%. Most jobs are just moving [ __ ] forward
368. 1:00:52in a system that's overly complicated because that's how someone invented it in 1854. And we're still doing some version of
369. 1:01:00the same thing with fancier tools. But what we're doing is standing in an assembly line. I got to input the data into Salesforce.
370. 1:01:10Hang. Hold on, please. Oh, sorry. I just got a new stack of things. Oh, I got I got a new stack of things.
371. 1:01:28What do you think's happening in that person's brain? Looks like I'm not I'm not going to get a piece of
372. 1:01:40cake. I I got to get got to get a piece of cake. If you cut cut cut cut cut cut cut cut cut cut cut cut cut cut cut cut
373. 1:01:43cut cut cut cut cut cut cut cut cut cut cut cut cut cut cut cut cut cut cut cut cut cut cut cut cut cut it smaller if I
374. 1:01:44just by doing the math I can see that I'm not going to get a piece of cake right so what if the AI did all that [ __ ] and
375. 1:01:57then what do we do I don't know what are you passionate about stamps okay go do something with stamps they make me happy I want to I want to
376. 1:02:07make an envelope that's that stretches from the east coast to the west coast of the United States and put stamps on
377. 1:02:14every inch of it. Perfect. People will pay to watch that. Go do that. [laughter] Don't be cutting down my pine trees. The
378. 1:02:25supply chain and warehousing is going to change drastically. Yeah, Trump is cutting funding to schools that don't guarantee more income on
379. 1:02:35graduating. Yeah, exactly. Yeah, there you go. All the arts degrees are going to be gone. Also, Trump just said he's
380. 1:02:41gonna pull uh $200 million out of universities uh to put into AI something AI. I don't know. Whatever.
381. 1:02:52Something something. Yeah. I don't know. Just Yes, I know. [laughter] pink gentleman parachutes.com is my
382. 1:03:03website. Perfect. Bob Don Bob Dunigan's already He's up on that one. Yeah. Amazing. Um, all right. I wanna I
383. 1:03:12want to tell you about one other thing. Share my screen again.
384. 1:03:31[whistles] Okay. So,
385. 1:03:49if you want to be on the bleeding edge of something, hang on. What is going on? Why can I not grab this window? Uh,
386. 1:04:03okay. That's weird. Huh? I just moved a window over and I can't
387. 1:04:19I can't move it anymore. It put the It put the grabby bar at the top underneath my menu. All right. Um Okay. So, so this
388. 1:04:29thing that you're looking at right here, this is a thing called Buzz. And Buzz is kind of like Slack.
389. 1:04:44It was announced yesterday or it was launched yesterday. And it's from Jack Dorsey. And Jack Dorsey, if you don't know, is
390. 1:04:54the founder of Twitter and he's um the CEO of Block, which is a financial services company where they just completely reinvented
391. 1:05:08how they do things. Um and okay, how is PI AI? Anyone know?
392. 1:05:22Um, I used to play with Pi a lot. Um, it was not bad. It was not bad. It's Pi is kind of an aggregator of a bunch of other
393. 1:05:30AIS. It's I I wasn't a fan of the interface, but it's not bad. It's worth playing with. Um, so what you're looking at here is Buzz. And so,
394. 1:05:43so Jack Dorsey is what what he's always been able to do is not only see the future. There's a lot of people that can see the future.
395. 1:05:54What Jack Dorsey's been able to do is see the future and build for it, build into it. So this thing is like slack except
396. 1:06:06so so meaning if you don't know what slack is down the lefth hand side you have different conversations so it's
397. 1:06:12kind of like mighty networks you have these different conversations you have different people can be part of those
398. 1:06:18different conversations and then you can you can chat you can say hi there and someone else can say hi there and when you start it up
399. 1:06:29You're given three agents. You're given the agent Bumble, Honey, and Fizz. So, Bumble is bring me in when you want to
400. 1:06:38research something. Honey is bring me in when you want a rough idea um clarified or amplified. And Fizz is I'll help you build something.
401. 1:06:50And so, what you do with with this with Buzz is you create communities. And a community is a server. And then you invite people
402. 1:07:01into your community and they can come into your community, but they can also share their compute. You what you do when you launch Buzz is
403. 1:07:11you bind it to your AI agents like Codeex or Claude Code or your local inference and then other people in your community can use that.
404. 1:07:24The other thing it does is when you create an account with Buzz, and by the way, the website is buzz.xyz. Oh, is Brandon here?
405. 1:07:36Um, if you go to buzz.xyz, you can get this. When you first create it and you say, I want to create an account,
406. 1:07:45it gives you a cryptographic pair, a private key and a public key. You keep your private key private and you can share your public key.
407. 1:07:55Here's why that's really important. Buzz or Block or Jack Dorsey are not running a server that has your your
408. 1:08:07account on it. He could turn off Buzz and your identity is still with you. So this is an identity system for people
409. 1:08:20that that is transferable across services. The other thing it does is all of the conversations inside your buzz
410. 1:08:28space, inside this space with these agents where you're talking to other people and you're talking to agents and
411. 1:08:34you're telling the agents to go do [ __ ] All of those conversations are written into a kind of blockchainy kind of um
412. 1:08:48it's called stateless like a stateless uh server where if you wanted to move that off their servers and onto some other server
413. 1:08:59you could do that and they're going to make it much easier to do that. But basically this is not quite fully fully
414. 1:09:05decentralized blockchain but it's in the neighborhood. This is about data sovereignty and it's about collaborating
415. 1:09:12with other people in conjunction with AI agents. It's a big deal. It's very very new. It's it's a day or two old and it's
416. 1:09:21janky. It's like it, you know, version 0.4. Um, and it could absolutely be a steaming pile of [ __ ] Like I like it is possible
417. 1:09:36that it's going to be bad, but it's Jack Dorsey. Like the pedigree is pretty deep with him and he seems to understand the
418. 1:09:44implications of what's coming. He laid off 4,000 of his 10,000 employees when they were profitable and growing.
419. 1:09:53Because what he realized is that for 2,000 years, we've been trying to scale organizations with human beings in middle management.
420. 1:10:02and middle management always [ __ ] everything up because the humans are bad at communicating and they're slow and
421. 1:10:08they're inaccurate. [clears throat] And so he's replacing that with AI. So this guy understands the implications
422. 1:10:18of the technology that we have in our hands. So agents can be assigned an ID. I don't know about that source camp, but you can
423. 1:10:28create your own agent and then I assume that agent is tied to your identity and then if you move that agent probably
424. 1:10:34goes with you. I'm not quite sure how it works yet, but yeah, the basic idea here is you can create private spaces
425. 1:10:43with an identity that you can transfer across services with context conversations that can work across
426. 1:10:51services. So I could have I can have an agent that is my clawed code agent or two or 10. I could have my codeex agent.
427. 1:10:59I could have an openclaw agent, a Hermes agent. And I could have a local inference of AI powering agents that are in this system.
428. 1:11:09So if I had five different computers, I could put an agent on each one and make them part of the community. And then if
429. 1:11:15you came into that community, if if you were trusted enough for me to put you and I know source camp, that's never
430. 1:11:21going to happen, right? Like who would trust you? So if I trusted you enough to be in this community, then you could use
431. 1:11:28those agents, too. And so this is the thing. I was talking about this in here about a month ago. This idea of small,
432. 1:11:36tight, trusted cohorts of people that are co-creating businesses together and in a in a shared kind of environment
433. 1:11:44where they can also share their compute. That's what I think this is the foundation for. Who this? What are you doing for date
434. 1:11:53night? What's happening there, Jason? I just This is a rogue Tik Tok where I decided to let the tick tockers take a
435. 1:12:02take a win and I just kept it running for tonight. Um it's not going to be there for long. I also can't don't know how to move this window.
436. 1:12:14That's really [ __ ] up. Oh, I got it. Okay. Gonna bring it back. Anyway, so buzz.xyz XYZ. Um, if you're here, there's 30 of us
437. 1:12:28here. Download Buzz, claim your name, right? So, just like imagine being in the early days of Twitter and and you just got the
438. 1:12:39name Kyle or Jack Dorsey got the name Jack because he invented Twitter. Imagine being early and being able to
439. 1:12:46get your name without the one, two, three, four, five at the end of it. I would do that with Buzz. I think that I
440. 1:12:52think that this buzz thing is really significant. What would be an interesting experiment is find within the AI salon the two or
441. 1:13:03three or four or five people that you like that you trust and where you've had interesting collaborative conversations
442. 1:13:12and start a community a buzz community that you own that one of you owns and invite them into it and start playing and learning
443. 1:13:22this thing because there's a bunch of there's a bunch stuff going on in this model that's really interesting to me.
444. 1:13:29The the biggest is data sovereignty. The second one is this idea of stateless. It's not serverless. You're
445. 1:13:37still hosting it on a server, but you can host your own server, but it's stateless, meaning Jack Dorsey doesn't own your data.
446. 1:13:50Your data is tied to your identity and you get to keep those. Those are portable. That's a big deal.
447. 1:14:00And then the other thing that it does is it allows you to connect multiple people's local inference and multiple
448. 1:14:08people's clawed GPT perplexity subscriptions and you can all tap into each other's [ __ ] Now, I don't know how permissions work.
449. 1:14:20I don't know if it's any good, but it just seems to me this would be a good thing to experiment with. Buzz. Buzzz.
450. 1:14:33Kyle or anyone. I can't see the URL for Buzz because I didn't put it up there. Hang on. Buzz. XYZ. Boom. Boom. There you go.
451. 1:14:51It also didn't give me the public ID number, just a private one. Just store your private one and then in the once
452. 1:14:59you once you set up your your thing, uh it'll give you that. Okay, here's another thing. You can create your own community, but
453. 1:15:09you're not going to create your community at buzz.xyz. You create your community at I think it's buildid.com or something. Hang on a sec.
454. 1:15:22Builderlab app.builderlab.xyz is where you create the communities. So builder lab it's the infrastructure here
455. 1:15:31is kind of kind of convoluted. Um, but I created three of my communities in what I thought was Buzz, but if you look at
456. 1:15:40the URL where you create your own community as it builder lab, what you're actually setting up is a virtual server
457. 1:15:48where all of your conversations are going to be stored for your community. You can set that up on your own servers
458. 1:15:55if you want to, but this is just a convenient way to do it that they're setting up for you. My head hurts. Someone send me a DM of
459. 1:16:04all this. No, I know this is What's up, Kyle? Am I irregulars? What's up, Wolf Man Clint? Um, this is a day old, so I don't know
460. 1:16:15enough about it to explain it easily. I just think it's a big deal. All right. I I
461. 1:16:31Coolio. So that's that. No, but you just said it, so thanks. I don't know what I just said. Oh, cool. Got it. The URL.
462. 1:16:42Um, let's do another question. Okay. Seems smart for businesses, but what happens when you leave the business?
463. 1:16:48What goes with you then? Okay. So Gwyn, what you're talking about is actually really, really important. Let's say
464. 1:17:06let's say that you put together a consulting company and you create one of these buzz servers and you've got your two or three people
465. 1:17:15and then a company hires you and they and one of the reasons they hire you is you've got this cool consulting thing and they're like, "Hey,
466. 1:17:26we want you to come and we want you to bring that with you cuz cuz this thing is badass at what it does. Your identity is portable. So your
467. 1:17:35identity into your buzz community would not be that new company's email address. It would be the thing that you own. You
468. 1:17:42personally own your identity that has access to these buzz communities. And then you would I I would think that
469. 1:17:51you would negotiate if I'm going to use this at work, then you don't have any rights to any of the conversations that
470. 1:17:58go into it. I'll give you the outputs and I'll put those on your company server, but all of how we got there, I'm
471. 1:18:04going to retain that because that's where the real power is for me. And in that case, you could move it. But what
472. 1:18:10you're talking about, Gwyn, why this is what why your question is really good. I think the employee of the future
473. 1:18:25that is smart is going to be aware of the value of their data and they are going to own their data. And if they're really good,
474. 1:18:37someone's going to say, "I want to work with you. And then you'll say, "Oh, I'm so flattered. That's awesome.
475. 1:18:44By the way, here's how I work. I'll come into your business. All of the [ __ ] I do is going to be running through my thing
476. 1:18:53that I own, and you'll get the benefit of my brain and all of the brilliance that I've put into my digital brain.
477. 1:19:02You'll get the benefit of that. We'll make your reports. We'll make your We'll rebrand your thing. We'll give you
478. 1:19:08strategy. It's you're going to do great. We'll build you apps and I'm going to give you all of that, but all of the all of how I got there,
479. 1:19:18that's mine. And initially, companies are going to be like, "Oh, that's you're going to have to sign the NDA with
480. 1:19:29that's the it's the standard protocol with [laughter] Yeah. No. Uh, [ __ ] you. I'll just go. Oh, you want You're being unreasonable.
481. 1:19:44Hey, why don't we go out to lunch? I'll buy you a steak dinner. So, yes, right now when you leave the business, they keep your [ __ ]
482. 1:19:59It didn't make sense for us to think about the data we were contributing to a business because everyone was too clueless to use the data.
483. 1:20:11No one gave a [ __ ] about your Slack channel. You know what Gary Vee is doing right now? Gary Vaynerchuk, he's got this big
484. 1:20:23$200 million ad agency. He's turning his Slack channel into a Vayner Media baller GPT that knows everything about their
485. 1:20:38business because it's got everyone's ideas in it, every answer to every question, every problem, every piece of gossip.
486. 1:20:47It's all in there instantly accessible. I mean, every employee signed a piece of paper and said, "Sure, you can have my
487. 1:20:57data because who cares about their data? You can't use that [ __ ] again." Now you can. Producer Brandon in studio. So, so we're going to
488. 1:21:13the smart employee. There's 31 of you here right now. Start getting your head around the idea that your chat histories and your
489. 1:21:24conversations and all of the hundred little stupid apps that you vibe coded and all the thousands of pictures that
490. 1:21:31you made and all of the songs that you made on Sunno that all of that data and all of that prompting that is what you figured out how to do
491. 1:21:42has value and you're going to be able to start to capture that and manage it and leverage it in a way that makes you more powerful
492. 1:21:56than a person who doesn't have that. And those people that really understand that and [ __ ] own that gold mine,
493. 1:22:08they're going to get paid a lot of money. All right? Can you get kicked out of the community by the person that created it? So what
494. 1:22:19so what they're allowing you to do is you're allowed to create three communities and like what's the
495. 1:22:24community? So so think of a community as a Slack server. So the community is just literally you're creating a Slack server
496. 1:22:35that that you own. you created that Slack server and right now they're allowing you to create three of those.
497. 1:22:42That Slack server is actually a server that all of the conversations that happen get stored in a ledger that is portable.
498. 1:22:54Right now, Slack, if you have a Slack server and Slack decides to ban your account or shut you down, you don't have a Slack server anymore.
499. 1:23:06What happens on Buzz is your ID is independent of Buzz and and the data on your Buzz community is independent of them. You can you can
500. 1:23:21download it and store it on your local hard drive. So, so you could have an agent say, "Hey, every night
501. 1:23:28I want you to back this up onto GitHub so that I've always got the data of this community." so that if Buzz ever shuts
502. 1:23:37down, I can just take this and put it on another server that uses this same protocol. That's pretty radical. And so, no, if
503. 1:23:46you own your community and and you own it, it's tied to your your identity, which has your private key attached to
504. 1:23:54it, which is independent of company or service. They can't. Now, you could be part of someone else's community and they could
505. 1:24:02kick your ass out. So, don't be a jerk. [laughter] Ah, Mel, join my community. Good. I I'm glad you're all starting to join each
506. 1:24:14other's communities. Yes, do this. Set them up. Start to start to create these little circles. I've I've been trying to figure out how
507. 1:24:23the AI salon could start to look like cohorts of people doing interesting things together. The reason I've been weird weird about
508. 1:24:33it, there's no good way to do that in Mighty Networks. And Mighty Networks owns everything. If the AI salon, if we
509. 1:24:41stopped paying our bills next week and you started up this little cohort, you'd have to go back to the Discord server.
510. 1:24:47You'd have to go start your own thing and it's a pain in the ass. Not with this thing. So, start learning about it because
511. 1:24:57there will be more like this. I I promise you what's going to happen is people are going to see this from Dorsy
512. 1:25:02and they're going to say, "Oh shit." Basically, what what this what Buzz is is a combination between Slack and Open Claw.
513. 1:25:13So, OpenClaw was this thing that lets you run agents 247 and you can do those with local models or cloud models.
514. 1:25:23And then Slack is where workers get together and talk to one another. And what Dorsey said was, "Well, wait a
515. 1:25:30minute. What if we created a stateless Slack that seamlessly integrated autonomous agents that people in the community could
516. 1:25:41create and control and talk to indistinguishably from humans? And then all of that data that is super valuable as Mel and Mary Carol and
517. 1:25:55Silver Fox get on one of these servers and start their artisal honeybee business. All of those conversations, all of that
518. 1:26:04stuff they build, they own all that. The value of that collaboration is now a valuable entity.
519. 1:26:21Yes, tomorrow. So, tonight's Friday. Oh, we're almost out of time. Um 8 to 12 p.m. Pacific. Um is it noon? No, 11 to 3 Eastern time.
520. 1:26:36Marty Chaikowska and Sherry, I forget Sherry's last name, are putting on a workshop. If you're if if you've built
521. 1:26:45something and you're ready to launch it and you're kind of stuck, that's what this workshop is. If you're thinking about launching
522. 1:26:54something, if you're thinking about building something and thinking like, huh, once I build it, how do I launch
523. 1:26:59it? Go to this thing. It's free. It's free. Okay, so that's tomorrow. Go do that. Go. Go, Luke. Go. Beautiful. Thank you,
524. 1:27:16Mary Carol. I do ads for free. Okay. So, if I built all my stuff on Claude, how do I take it to Buzz and
525. 1:27:23protect it? Can we run our own agents there? Yes. I don't quite know how it works, Chef Kelly. But what happens is when you create a
526. 1:27:35when you create an identity, you can then you can then create Wait, when you create Yeah. When you create an
527. 1:27:44identity, that identity is yours and and it's transferable across services, whatever uses this protocol.
528. 1:27:54And then when you create a community, you bind that community to an agentic harness. So I I I bound mine to codeex and to cla
529. 1:28:07code. I've done I've done a little bit in clawed code. I've done a lot more in codeex. But now what I can do is I can expose
530. 1:28:18those two services. Plus, if I have a local LLM, I can expose those to the community and then we can build agents on top of those
531. 1:28:28things. So, I can create an agent that says, "Go to Kyle's codeex and do this specific thing and use the skills that
532. 1:28:36Kyle's already developed to do this thing." And we can name that agent Bob. And then I assume the way it works is we can say the following
533. 1:28:47conversations in in this buzz have access to different agents, right? So in the financial services conversation, there might be a whole
534. 1:28:58bunch of agents about running reports and doing financial projections. In the marketing one, there might be marketing
535. 1:29:04agents. And so in those in those channels are also human beings that are marketing people using marketing agents
536. 1:29:11to do marketing [ __ ] And all of those conversations are being tracked and logged into a ledger. It's pretty brilliant the more I talk
537. 1:29:22about this. So I could create a die converter community and all of my users could come there for help chat knowledge base. I think
538. 1:29:32I think so. like it like so so Chef Kelly, it's no different than um it's no different than creating a Slack channel.
539. 1:29:43You could create a Slack channel for for your die converter community and they can go in and talk to one another and
540. 1:29:50you could have different channels for ask questions here, you know, give recipes here, what things like that,
541. 1:29:56right? The difference with this is everyone's sovereign. The the downside would be in order for people to be in your community, they
542. 1:30:07have to create one of these IDs. And right now that's a little technical, but but yes, you could. But what you could do is you could
543. 1:30:16create agents within this thing that do specific die converter things. So you could have, you know, um, you know, the
544. 1:30:25Jolly Rancher agent of, you know, color mixing. If you want custom colors, what dyes do you mix together to get, you
545. 1:30:34know, what colors? You know, you could have agents in there that they could have access to. Kind of crazy.
546. 1:30:42All right. So that leader may have their tokens chewed up without approval. In other words, I run out of tokens and
547. 1:30:48just take them from Kyle. Yes. So, see, you're starting to get it, Gwen. Possibly. This is where I haven't used
548. 1:30:57it enough to actually understand it. I don't understand how permissioning works. I don't understand like if I put my codeex in there,
549. 1:31:07can I selectively give access to codeex or does everyone have access to it now? And then how do I protect my data from
550. 1:31:13what I want in the community? All that stuff. I don't know. I'm I'm suspecting that what what you might do within
551. 1:31:20codeex is create a project for that community and then that community only has access to that folder and all your
552. 1:31:29other [ __ ] I don't that I don't know but they would not be afraid afraid of transferable data. Yeah, exactly. New
553. 1:31:41way of evolving hobbies. Yes, absolutely. Okay, it's 9:02. I try to get out of here on time. Thank you all
554. 1:31:49for being here. Uh I'm gonna go say hi to producer Brandon who's backstage. And uh my brain is melting with
555. 1:31:56possibilities. I know. I know. This is it. This is brand new stuff. And my caveat is this could absolutely flop and
556. 1:32:04be a piece of garbage. Okay, so you heard it here first. You heard about it here first. I think it's a big deal. And
557. 1:32:11it could be a pile of garbage. So take that into the weekend. If you want weekend homework, go play with Buzz this
558. 1:32:17weekend. And then let's connect in the AI salon and start to create little cohorts, little communities where we can
559. 1:32:23start learning how to use this and we can start talking about is this a piece of garbage or not? And if it's not,
560. 1:32:29let's start doing interesting things together. Okay. All right. Peace out, everybody. Really appreciate you being here. Tik Tok, this
561. 1:32:40is not coming back forever. Ever. This was a rogue experiment tonight. You hit a soft spot. I love seeing you there.
562. 1:32:47Peace out, everyone. Bye, [laughter] tick tockers. All right. Peace out. Have a good night, everyone. Have a good weekend. Bye.