8/28/2026 - How the Definition of AGI Is Changing and What It Means for Future Careers

Transcript

1. 0:01[music] [music]
2. 0:15[music] Heat. Hey, heat. Hey, heat. [music] >> [music]
3. 0:32>> Good evening. Good evening. Good evening. Welcome to the AI learning lab in the AI salon. I hope you're all doing
4. 0:38well tonight. I didn't do a pre-show tonight. So, if you're like, why weren't you on the pre-show? I had to drive the
5. 0:44B boys up to Boulder. I was the uh I was Uber drivering. I was Ubering without getting paid. So, so I just got back. I just got back
6. 0:56from driving. So, anyway, uh welcome to the AI learning lab. It's Friday night, date night, and uh it's an ask me
7. 1:03anything. So, if you're here, welcome. If you're new to the AI salon, welcome. Uh be sure to check out all the
8. 1:11different spaces down the side of the AI salon. All those different spaces or lots of different conversations that you
9. 1:17can go check out. And uh don't they have Whimos in Denver now? There are Whimos driving around Denver, but in the Whimo
10. 1:25app, um, Denver is not one of the listed cities. So, I think that they are literally just driving around our city
11. 1:33with no one in them, um, trying to avoid being in accidents. [laughter] So, so anyway, there you have it. Um, there's
12. 1:48there is a tremendous amount of activity happening. Hey, Mel B. There's a there's a It just struck me. Mel B is the Spice
13. 1:58Girl Mel B is also called Melb. And you're called Melb because I'm a big Spice Girls fan as you all know. I don't know. Um
14. 2:12anyway, there there's a there's a lot of activity happening in AI. a lot of activity, you know, it [laughter] um
15. 2:27and so one of the things I thought we could do tonight is go look at in office hours today. Uh we went look went and
16. 2:34looked at some of the new stuff. I have made another ridiculously stupid purchase which makes me giddy, makes me
17. 2:42really happy. Uh, if you haven't seen the micro duck yet, go check out the micro duck from Hugging Face, uh, and you'll understand why I'm
18. 2:52excited. [laughter] You can you can buy a robot now that that can learn it. It can use an AI brain [laughter]
19. 3:02and you basically you do your own reinforcement learning in a simulator. you train up your your duck in the
20. 3:10simulator and then you take whatever you taught him and you dump it in its brain and then it can do that in the real
21. 3:15world. So, um I don't think I'm going to see that duck for six months because they said it's not guaranteed by
22. 3:22Christmas anymore because too many people were buying them. [laughter] So, what are you gonna do? Um so, a couple of things. Uh, if you want
23. 3:34to pop out the if you're on mobile and you want to you want to pop out the the live stream independent of the of other
24. 3:41stuff you do on the phone, right above this is the link to pop it out if you're on mobile. Um, producer Brandon. All
25. 3:49right. What is this question we've got? So, by the way, tonight is an ask me anything uh night. So, if you've got any
26. 3:56questions, feel free to ask away. Um, and we'll get to it. Uh, with AI agents getting powerful enough to cause real security problems,
27. 4:09is the future of AI more about capability or control? I don't know if it's AI agents that are Well, I guess it is. I guess the if the
28. 4:20the hugging face attack, we we talked about this on Wednesday night that the hugging face attack was
29. 4:341,200 agents coordinating, figuring out how to communicate with each other, which they were not supposed to be able to do,
30. 4:42figuring out how to break out of their sandbox and get on the internet, get internet access. Um, I think that the the real answer
31. 5:00is that AI is going to mess everything up. It's going to mess up security. It's going to mess up cryptography. It's
32. 5:08going to mess up business. It's going to mess up employment. It's going to mess up how research is done. It's going to mess
33. 5:17up copyrights. It's going to mess up patents. Like, like we're in a world where you can get a patent, but everyone can instantly
34. 5:32circumn the patent, right? Because they have AI to help them navigate around it. So, everything's going to get messed up. And
35. 5:44nothing's going to get messed up permanently. I mean, potentially, you know, if it shuts down the grid or
36. 5:52shuts down computers, you know, sets off an EMF or something and computers no longer work, then then AI would be kind of screwed.
37. 6:00Um, and then we'll go back to farming. We'll go back to economy one. But all the stuff that is about to get
38. 6:08messed up, what's going to happen is new versions of that are going to emerge. Linda Sherman, great office hours this
39. 6:14morning. I sent a web link to my very start smart 14-year-old grandson and got him excited. Awesome. That's terrific.
40. 6:21Yeah. there. Yeah, there's there's a a lot going on and there's a there's I think we're in a space right now where
41. 6:41it's very easy to to to go binary. It's good or it's bad, you know. It's profoundly powerful. It's profoundly evil.
42. 6:58And I don't think either of those serve serve us very well. I think that thinking critically like not thinking of AI as a single thing
43. 7:15breaking it down into individual use cases and understanding that AI here might be very powerful AI here might be very scary but like AI is
44. 7:26a blanket concept it it doesn't it doesn't serve anyone to just talk about it as AI and and that's a lot of what's happening
45. 7:40right now is people are just talking about it as AI with these blanket statements. It's good, it's bad, it's
46. 7:47and so I think we're at a we're at a stage with this stuff where we have to be smart. We have to have thoughtful conversations. You know,
47. 7:58Linda, you sending what you felt was a good conversation today to your 14-year-old, would you say grandson? Yeah, your 14-year-old grandson.
48. 8:08If that gets him thinking about things in a different way or more critically or like maybe he's got multiple lenses
49. 8:16through which he can look at AI now. Great. Um because we very much and I I I feel like this as a country, we're we're very much
50. 8:29flirting with cutting off our nose despite our face with AI where it's it's dangerous and we're going to slow it down and we're going to
51. 8:39right we're going to do all this stuff. We're going to slow it down. Well, it ain't stopping anywhere else.
52. 8:46And so our ability to defend ourselves will likely require strong AI. [laughter] So, you know, if we're like,
53. 8:57we don't want strong AI, well, maybe we should make sure that we have that. [laughter] Oh man. Let's see. The Benny robot looks mad
54. 9:10cute. He jumps and is all terrain. Let me see the Benny robot. What's the Benny robot? Hang on. Did I miss that? Cuz I got the I got the
55. 9:21micro duck. Benny robot. Oh yeah, that's that little one that
56. 9:35Maris Brownley reviewed. >> [laughter] >> Oh, that one's pretty cool.
57. 9:53All right, hang on. I got to I'm not showing my screen. Yeah.
58. 10:11All right. Okay, here's the Benny the Bennett robot. >> Do a flip. So, I said yes to every single email for
59. 10:19an entire month again, expecting a bunch of silly stuff to show up. Um, and it did. But this is the coolest thing that
60. 10:26showed up. This is the Benny camera robot. And that face right there has a camera. So, it's literally just a tiny
61. 10:32rolling camera on wheels that you can control with a remote or an app on a phone, which is pretty great, honestly.
62. 10:38You can just tap on one person and it automatically locks onto them and follows them around with object
63. 10:43avoidance and at up to 18 mph. And the footage is actually solid, too. It's 4K and everything. So, why can it also do flips?
64. 10:52I have no idea. There's no advantage to being able to do a side flip or a back flip. It certainly doesn't look good in
65. 10:57your footage, but it is also incredibly impressive that the tiny little motors and all the self-balancing and gyros.
66. 11:03>> Maris Brownley, can you get his email address? He's like super famous. >> I'm just saying if that guy says yes to
67. 11:12every email that he gets, >> we should get his email address. >> Yeah, really. Um, 6.1 million followers. [laughter]
68. 11:22Um, how much is this little bad boy? Let's see. Oh, damn it. Benny Allterrain Benny. Oh, it's live on Kickstarter.
69. 11:34549. Okay, my little micro my micro duck that I ordered today is going to cost me 500. So, they're in they're in the same
70. 11:43neighborhood. Um, micro duck, show us the duck. All right, here's the duck. [music]
71. 12:01[music] Heat. Heat. [music] [music]
72. 12:26>> [music] >> And I got the light blue one. The third one in is the one I got because it looks
73. 12:42like Here, you want to see why I got the the blue and orange one? Um, let's do um golf. Uh, let's see. Race car.
74. 13:00Yeah, baby. See this? the old Ford the Ford uh whatever the model number was of this
75. 13:15the GT40. The Ford GT40 that looks like my Cyberduck. So I'm going to put I'm going to put stickers on it so it looks like a race car.
76. 13:26[laughter] Look, that one looks like a duck. See? See what I did there? Let's see. Let's go to the website. Show more micro duck.
77. 13:50So, what's cool is you start the simulator, you launch the simulator. What's the website? The website is huggingface.co. This is by uh
78. 13:58HuggingFace. And then it's slashspaces. Here, I'll put it uh I'll put it in StreamYard. Just look up Micro Duck Regulars
79. 14:14Challenge Image Competition. Show us the duck. Um, all right. Let me move this over here and then let's go back to the duck site.
80. 14:28Okay. So, arrows move A and E kick left and right. What is that? G picks up AE. E G picks up C equals camera.
81. 14:44Space equals reset. R equals So, AER GC. Okay. Waddle in. Okay. So,
82. 15:00kick. [laughter] So, what you do, and I I I don't know how to do this yet, but what you do is you train it
83. 15:15in this simulator. Let's see.
84. 15:35And then let's see. What were the other ones? G. Oh, G's pick something up. C. I forget what that was. R is sit down. Yeah, sit down.
85. 15:48I mean, come on. Tell me how cute. That's cute. You can't take the race car out of the lab. >> [laughter]
86. 15:57>> We're We're now in a place where Short Circuit is real. Love that movie. I know. Is this for the twins, Kyle? No.
87. 16:10the boys. This is for me. They don't get this.
88. 16:27>> [laughter] >> Let's see. Kick. Kick. Kick. Kick. Kick. Oh, I see. Kick. Kick.
89. 16:46Oh, that's the kick. Okay. So, it's Q and E, not A and E. And then C is sit. No, C was camera. R
90. 17:00is sit. And then I can spin this around. So what I don't know how to do yet is to like record this. Let's see. Files. Okay.
91. 17:17So, this is on hugging face waddling pre-order. Oh, feet. Rollers. Look, now he's got
92. 17:31rollers. So, how do you get him to how do you get him to learn to speed skate? Right. Right.
93. 17:51>> [laughter] >> I don't know what I'm doing. Oh, look. [laughter] How do you make him go backwards? Oh. Oh, no. [laughter]
94. 18:11Oh, he fell down. Can we go backwards? Yes. Go forwards. Go right. Go left.
95. 18:28Training is virtual. Application is physical. So this is a tangible thing outside the computer or a tool inside.
96. 18:34No, no, no. Okay. So what this is, what micro duck is. Okay. So, what you're looking at, this is really cool. Why is my computer hung?
97. 18:52Uh, this doesn't look good. Hang on. Oh, let's get rid of this. Okay, that's better. Okay, the simulator
98. 19:06was was eating my processor. Um, what is with my hair? Good lord. Um, so that simulator you were looking at,
99. 19:16that's where you train the robot. So I guess it comes as a dumb robot. It can't do anything. Doesn't know how to walk.
100. 19:21Doesn't know how to do anything. You train it in the simulator and then you teach it tricks. Like someone taught
101. 19:27theirs to do like a front flip. I don't I don't know how you do any of this yet, but you train it in the simulator and
102. 19:34then basically you download your training set. So, you're doing reinforcement learning for this thing in
103. 19:40the simulator, and then you basically stick it in your physical robot. Let me There's There's another video. Let's go.
104. 19:46We'll go watch the other the other video. Hang on.
105. 20:11All right, let's see. Let's go. Let's go to X and we'll go Micro Duck. We'll see. We can see if we can find some other vidgetes.
106. 20:33Okay, here's a real one.
107. 20:47[laughter] I hope it comes with roller skates. It's very cute.
108. 21:02You're going to create a space in your office for this guy and make a stop action film. That's cool. Oh, yeah. This this one.
109. 21:17He crapped out his battery.
110. 21:52I think that was someone just made a fun movie. Oh, this is funny. So, you're telling me it's open source? I can print
111. 21:59STL parts for it. Take my money.
112. 22:38Let's see. sharing some fun experiments with MicroDuck. Basically, once you have an open source robot with so many sensors,
113. 22:48the sky's is the limit. Here, we vibe coded an image detector integration to detect and follow a laser pointer. Cool.
114. 23:09>> [laughter] [laughter] >> Producer Brandon, you're going to have to fire up your vibe coding skills. You're going to have to help me make
115. 23:24We're going to have to do stupid robot tricks. Remember David Letterman's stupid pet tricks? That's totally what this is.
116. 23:31Okay. Raw footage. August 2026. A thread of joyful micro duck photos and videos. Oh, cool. Oh, this a whole thread. Okay,
117. 23:38cool. Let's Let's watch some of these. I love how he walks. [laughter]
118. 23:53He's very not stable. [laughter] Oh, that's cool. Balance recovery.
119. 24:11>> [laughter]
120. 24:29[laughter] [laughter] >> Roller skating.
121. 25:01>> [laughter] >> I'm sorry, but I'm going to go out on a limb and say this is a good purchase. [laughter] Oh my god, that's hilarious.
122. 25:19Here's a purple one.
123. 25:41[laughter] Oh my god. Comes with stickers. Does it come with stickers? I hope so.
124. 25:59They're teeny. They're teeny little robots. [music]
125. 26:15Actually, you know, it's funny if you think about it. Not that long ago, if you wanted something like this in a
126. 26:26movie, you'd have to do special effects. This is a great robot. My main complaint is manual battery replacement and short battery life make
127. 26:37it more toy than pet. I would immediately get one if I could swap if it could swap out its own batteries and
128. 26:44had a charging dock. Ah, let's just sour grapes. All right.
129. 27:02All right. So, that is my stupid purchase of the month. Stupid purchase of the month for August. The Cyberduck. [laughter]
130. 27:18Smart purchase of the month. Smart purchase of the month. AI salon mastermind. If you are new to the AI
131. 27:26salon or if you are not new to the AI salon and you're not a part of the mastermind, please consider joining it.
132. 27:31And in fact, hey, producer Brandon, I don't know if you're uh if you're slushed or if you're if you're available. You might have tidbits
133. 27:41crawling all over you. Um, but starting next week, September 3rd, uh, begins a 4-week Codeex master class, an expert
134. 27:52series class that Brandon is teaching on how to Why don't you tell the good folks what you're teaching because it's this
135. 27:58is a I think it's a really important course. I said that on office hours today. You weren't there, but I talked
136. 28:05you up on office hours. Um, why don't you tell the good folks what you're up to? Yeah. So, it's going to be a course that
137. 28:14introduces you to the concept of having one of these agentic assistants, be it cla or codeex chat GPT work or Google's anti-gravity
138. 28:27on your hard drive and the power and potential of what it can do with and for you while your computer is sitting idle
139. 28:36and you're off doing something else like playing with your kids or walking your dog or doing pottery if you're Joy
140. 28:42party. Uh because what you can do is once you get this thing set up properly, you can just grab your phone, log into
141. 28:49the chat GPT app and hit the remote button and then you could just talk to the agent that's on your computer and it
142. 28:57can do things like check your email, uh write blog posts, update your website, and a whole host of other things. We're
143. 29:04going to be going through a whole list of practical applications. I'm I've been holding off on doing some of the
144. 29:11projects that I want to do with co-work because I want to be able to demonstrate what it can do. And it's all going to
145. 29:17start next week, Thursdays at 2 Eastern, 11:00 a.m. Pacific and they will be recorded uh if you can't join live.
146. 29:24>> And so you're going to teach people how to set up codecs, how to do the use cases, like you're going to show
147. 29:31specific use cases, things like that. >> Yep. And we're also going to talk about how to do it the right way so you don't
148. 29:37accidentally delete all of your personal files on your hard drive. >> Well, where's the adventure in that [laughter]
149. 29:48>> and and for you Kyle? I' I've thought specifically about tab hoarding and and bookmark management. So
150. 29:54>> Oh, good. Yeah, that could help. That could help. I actually connected. So one of the things that um chat GPT launched today is
151. 30:05um so within chat GPT and within codeex uh chatbt work you could connect Gmail and Google calendar um and today they announced you can you
152. 30:17can now assign you you can access multiple accounts so you can add different accounts uh so if you have
153. 30:24different businesses different emails you can have chat GPT be able to search all of them. It's pretty amazing. All
154. 30:31right, next question. Let's go. Let's go. Let's go. Let's go. If you have any questions, um the below the live stream
155. 30:38is the chat. Uh if you put in the chat, we can see it back here and we will ask a question. Is Google quietly winning by
156. 30:46putting Gemini inside everything or does AI still need a big cultural moment to matter? Um, Google is
157. 31:04Google doesn't have a product strategy. Google has 43 product strategies. And so jamming Gemini inside the 43 different streams
158. 31:16of what you can do in Google is just confusing. So, I think AI needs a big cultural moment. I don't think we've had it yet. Um, I
159. 31:30doubt it comes from Google. I think that Google will probably follow on quickly after it happens. Uh, I don't know. Sam Alman said
160. 31:40he thinks we'll have AGI before the end of the year. Um, Sam Waltman's also said we should probably not release the we have
161. 31:51because it's too dangerous. So, I I don't know. I don't know what's going to happen. Uh, but I think I think
162. 31:59Google right now is a mess. Their product strategy is a mess. Um, if you can find anything in Google, more
163. 32:07power to you. Actually, what what Google needs to produce. Actually, Brandon, you could vibe code this. You should produce
164. 32:16a vibecoded app that tracks all of the all of the multitude of ways that you can access all of the multitude of tools
165. 32:27within and across Google so that you can go to this one thing and you have a single interface that just dynamically
166. 32:33routes you to the right place. That would be powerful on it. But seriously, like if you if you had a single interface where you could
167. 32:42go, "Show me the latest video model from Google or or let me play with the latest music thing and it just popped it up
168. 32:49even without a URL. It just popped it up in a window below some sort of slick interface." How are they defining AGI? uh Altman has
169. 32:59always defined AGI as the moment when AI can do most jobs of most people at or significantly above the average human worker.
170. 33:14So take any job, accountant, lawyer, uh I don't know, engineer, artist, right? I think that
171. 33:32it's still going to be jagged. I think some jobs, right, because the the training runs are focusing more on
172. 33:39programming and engineering than they are on humanities. I have a feeling that the humanities are going to be a lagging indicator. Uh but
173. 33:48I don't know like in theory AGI is all jobs of all people of all economic value at or above the median human. Um well
174. 33:59plumber will happen Chef Kelly when the when the humanoid robots Jason Kalcanis keeps talking about Optimus 3. He says
175. 34:08that he saw the Optimus 3 robot. he can't talk about it, but he said he said no one's even close to what uh SpaceX or
176. 34:18Tesla is doing with Optimus. Um we'll see. Gemini notebook used to be good for slides. Is there a way to do them with chat GBT?
177. 34:29And do you recommend that? I have a couple of slide decks I want to produce this week. Gemma is a really good tool
178. 34:34for slides. I think uh Notebook LM from Google is still good for slides. Um you can make slides within Chat GPT, but
179. 34:47they kind of suck. Um GenSpark.ai has a really good slide maker. Manis is pretty good, but Manis is is in the
180. 34:59middle of retracting themselves from being bought by Meta, and they're going back private again. So, I would say just
181. 35:05let let Mannis figure its out for a while. I would say Genpark or Gemma are probably the two that I would lean into.
182. 35:14Um, not using Claude yet, but good to know. Oh, another one is Claude Design. Yeah, Claude Design is actually quite good.
183. 35:23That sounds like ASI, not AGI. Um, let's go let's go ask um let's go ask Grock. So, one of the reasons that I like Grock
184. 35:37is that Grock is searching for all of um all of X, all the real-time stuff. I'm gonna say um I want you to share Sam
185. 35:51Alman's definition of AGI and also his of ASI um
186. 36:10ASI comma and then give me three other definitions of each
187. 36:26um on a spectrum of conservative to vast or something like that. All right, let's see what yield Grock
188. 36:41has to say. thinking about your request. By the way, if you if you don't know it, SpaceX AI bought Cursor and so Cursor is
189. 36:59now um within so they they launched this thing called Grockbot, which is is like Codeex um but it's kind of like a
190. 37:08consumer version of it. I haven't played with it yet. I think I can play with it, but I need to get a $20 a month
191. 37:14subscription to Cursor before I can. Okay. Cam Catkin, Sam Alman's O AGI, the most official OpenAI line, which Altman
192. 37:24has used for years, is highly autonomous systems that outperform humans at most economically valuable work. Um, his more personal, informal
193. 37:36versions, a median human remote coworker, something you could hire that can do anything you'd be happy having a remote knowledge worker do.
194. 37:48Learn to be a competent doctor, coder, etc. with the meta skill of figuring out how to get good at whatever you need.
195. 37:55also a system that can tackle increasingly complex problems at human level in many fields and then he has also said a system that does not
196. 38:06continuously learn after deployment is missing something important in his sense of AGI so he thinks they should get
197. 38:13smarter now his definition of super intelligence he talks about super intelligence more as what comes after AGI than a single
198. 38:23sentence tools that mass essively accelerate scientific discovery and innovation well beyond what we are
199. 38:30capable of doing on our own. A candidate operational bar, a system that can do a better job as president of the United
200. 38:38States, CEO of a major company or head of a very large scientific lab than any person can even with AI assistance. The
201. 38:47chess analogy, first humans, AI beat AI, later the human only makes the system worse. models that make scientific discoveries
202. 38:57humans cannot make on their own. That starts to feel like super intelligence to him. Okay. Three other definitions. The conservative
203. 39:08definition of AGI, an AI that matches a competent median professional across a wide range of computer mediated
204. 39:16cognitive work, research, coding, analysis, writing, planning. It generalizes to new tasks with little extra training, but it does not need
205. 39:26robot bodies, inner life, or super superhuman science. This is close to Altman's remote co-worker bar and to
206. 39:35many industry economic AGI uses. The second definition, a system that can perform any intellectual task a human can
207. 39:46across domains including novel problems with roughly human sample efficiency and robustness, not just office work.
208. 39:54Economic value is a consequence, not the definition. The closer this is closer to older academic usage and to deep mind
209. 40:04style versatility plus new skills with scarce data framings and then the vast full human equivalent mind.
210. 40:14Number three, human level or better across the board, physical world competence, long horizon, agency,
211. 40:22continuous lifelong learning, original scientific and artistic creation, stable goals, and for some people something
212. 40:30like understanding or consciousness. Under this bar, today's models are still far away uh even if they crush many jobs.
213. 40:40So there you go. Basically, it's it's AI smart enough to do our work at our level or better is how I is how I think about it.
214. 40:54All right, what do we have here? Um, if AI can write code, research, make slides, summarize meetings, and search
215. 41:03your files, what should humans actually get better at? Well, we talk about this a lot. What humans need to get better at is
216. 41:15instinct, taste, judgment, ideiation, discernment, prioritization, um, basically having the idea, having
217. 41:31the vision. It's human's job. So, so here's the thing that we're discovering in the mastermind. And if if you haven't joined the
218. 41:42mastermind, you should join the mastermind. Join the flipping mastermind. The the process of getting AI ready seems to be this.
219. 41:58You first start out playing with AI and you're like, "Wow, that's amazing." And then you're like,"Well, wait. Could
220. 42:05it do this? Could it do that? Could it do that?" And the answer's like, "Yes, yes, yes." And you're like, "Oh my god,
221. 42:10this is really amazing." And then you take all the ideas in your head and you jam it into this AI box and it it starts producing
222. 42:20some version of all your ideas. And some are good and some are not good, but it's like all of your ideas get out
223. 42:28of your head very, very quickly and are manifested in the world. And you're like, "Holy crap, this is amazing." And
224. 42:33then you're telling all your friends and they're like, "Could you stop? I don't care. I I know you had an idea for a
225. 42:41novel when you were in seventh grade and now it exists. I don't care." And but you keep going and you keep
226. 42:49going and you keep throwing stuff and now you're having new ideas because you've gotten all the old ideas out of
227. 42:53your head. Now you're looking at your old ideas manifested in the world and you're like, "Oh my god, I got new ideas
228. 42:58and it's even better." And you keep throwing this at the AI and it's just it keeps making stuff. And then after about two years of that,
229. 43:10you wake up one morning, you're like, [sighs] I'm tired. Could someone get me a cookie? And this is the moment.
230. 43:26The moment where you need a cookie after two years of making all the that was in your head and realizing that none of it mattered.
231. 43:37Nobody cared and you could make anything. And you get to the exhaustion point and you need a cookie and maybe someone
232. 43:46gives you a cookie or maybe by Jingus you go down and you get the tube of cookie dough and rather than baking it,
233. 43:52you just eat it out of the tube. And then shortly after that blissful moment of eating cookie dough out of the
234. 44:01tube, you have this realization. Oh, I actually need to know who I am and what I believe in and who I
235. 44:17care about and the difference I want to make in the world before I go vomit some stuff on AI.
236. 44:32And I think that's our job. I think that's our job. I think our job is to go, who am I? Who are the people I care about?
237. 44:41How am I going to use this remarkable, profound, magical technology? What do I want it to do?
238. 44:57Like, like let's let's go to that definition of AGI. It's like a coworker that you can hire that's as good at any
239. 45:11job you give it as any human you would hire. And then with these agents, you can have multiple of them. So you can have a
240. 45:19bunch of different employees. So think about it like this. If tomorrow someone said to you, you can have 10 worldclass employees
241. 45:35to do anything you want to do. The only requirement is you need to be able to tell me what do you want to do and what do you need the
242. 45:45employees for? What are they going to do? That's your only job. But if you can do that, you get 10 full-time, more than full-time, 3x
243. 45:57full-time. They work 24/7. You get 10 247, seven day a week employees that will do anything you want them to do. If
244. 46:06you don't have an answer to the question, what do you want to do with them? Who are they? What will they do? What will
245. 46:14you direct them to do? If you don't have an answer to that, you are going to be at a disadvantage. Because the fact of the matter is by the
246. 46:22time those things get good enough to be 10 employees, you'll be able to have a hundred of them or a thousand of them.
247. 46:29And those people that are like, I know what I do with a 100 employees are going to go build some amazing They're
248. 46:37going to build some amazing companies. I need a cookie. Exactly. Exactly. You hit the cookie moment. Right.
249. 46:48I told my husband every time he tells me about something the Red Sox did, I'm going to [laughter] tell him about
250. 46:52something new I made with AI. That's really good. Really good. Kelly Camp went to dinner tonight and
251. 47:09just converted two more strangers to AI. I need help. [laughter] Kelly's like the Pied Piper. She's like the AI Pied Piper.
252. 47:19For my entire adult life, my family would ask me every birthday holiday what I wanted as a gift. And my answer was
253. 47:25always staff. Now I have them. This is it. That's it. You don't need to learn prompting. You don't need to learn agentic
254. 47:40workflows. You don't need to learn how to set up a buzz community server like I'm trying to do halfass off to the side.
255. 47:56You don't have to learn. You don't have to go to Brandon's codeex course. You should. You can just wait. You can just wait for
256. 48:05two years, year and a half. And these tools are going to be so good that your only job is going to be have a idea, be able to articulate it.
257. 48:24Like if if that existed right now, could you answer the question? I can't I don't have a good answer to that question
258. 48:31right now. But I think that's our job. Um ducks in a regulars chat. Okay, let's go. Let's go. Let's go. Let's go. Let's go to the videape.
259. 48:45Oh boy. [laughter] AI salon. All right. Regulars. Where's the regular? Where are
260. 49:00the regulars? >> Just our live stream chat. [laughter] >> That's good. I like it. Solid. Nice.
261. 49:16Beautiful. Why did that not load? Content created using CH GPT.
262. 49:32Wait, here it is. Ah, there it is. Nice. That would be a cool one. Speaking of irregulars. Oh,
263. 49:47so listen. Let me tell you something cool. So for those of you that are irregulars, there has been a contentious thing
264. 50:05when we moved the AI salon or the AI learning lab inside the AI salon. The chat experience was less than ideal.
265. 50:14Um, and a lot of you are big fans of the interactions inside Tik Tok. We did I did kind of a rogue experiment last
266. 50:25for when we did um back to basics where I left Tik Tok running as kind of a backstage view while I was demoing things on screen.
267. 50:38We're going to start next week and experiment because that seemed to work because one of the things that happened
268. 50:42when that was running was we got some new people that joined the AI salon and joined the learning lab because the Tik
269. 50:48Tok was, you know, bringing new people in and they would come over. So, starting next week, two things are going
270. 50:56to happen. We're going to do some programming changes. One is we're going to have Tik Tok back backstage access. So, the the Tik Tok's
271. 51:05going to be there. Um, I'm probably not going to interact with it much. Um, because I don't want to be distracted.
272. 51:14I might even turn the camera around so I can't see it, but it's going to be there. And we'll let people know um that
273. 51:23they can jump into the AI salon by going to my profile. producer Brandon will be monitoring Tik Tok for pinned comments
274. 51:31and things like that and we'll bring them over to the the comments here. So, I'll be able to see them. I just won't
275. 51:37be able to interact in real time, but it'll be like that backstage view. The other thing that we're going to do is I
276. 51:44really want to lean in. If I'm thinking about what I want to do with the AI salon for the next three years, five years,
277. 51:54this idea of AI readiness. In fact, let me let me bring up this image.
278. 52:11this idea of AI readiness where you're at the center your ideas right the thing I was just talking about and then how
279. 52:18you educate yourself you play you go play with these tools you you go play without expectation you learn across
280. 52:25domains right and then create excellence you take what you learn when you're playing and you build things and then
281. 52:32once you've built things you share what you've built you share what you've learned you share the failures you share
282. 52:37the successes is and you keep repeating this cycle. I really want to lean into this because the the more I look at
283. 52:45what's coming with AI, the more that I I deeply believe that this is the way that this is the way to not make yourself
284. 52:54absolutely batshit crazy because this cycle is independent of any technology. Um, so what's going to happen? Uh oh, what did
285. 53:08I do? Oh, there you are. Um, so what's going to happen next week is I'm going to take the Monday, Wednesday, Friday, and each
286. 53:22one is going to be one of the stages of the cycle of AI readiness. So Monday's going to be play first. So Monday we
287. 53:29will explore and we'll learn and we'll go do things. Wednesday is going to be create excellence. So we'll build
288. 53:35something a specific project I'm working on or we might even take someone who's working on something and you know
289. 53:41collaborate on it and then Friday is going to be generously lead and I think what that could look like is we start having conversations. Um
290. 53:51people have been a fan of when we bring people up on stage. Maybe that's what that becomes. Generously lead. I want to
291. 53:57explore what that means. But those three days are going to match that cycle of AI readiness. As we put on physical events,
292. 54:06we're going to lean into the cycle of AI readiness. Those those three things with the you at the center of it absolutely
293. 54:13feels like the future to me. So anyway, that's what's coming. [sighs] They don't need to become AI salon members to participate here.
294. 54:25They they do in order in order for them to see anything inside the AI salon, they have to become they have to
295. 54:30register for the AI salon. Um that's different than when you join, Linda. Um so they do have to join. They don't have
296. 54:37to be a paid member. Um but but in order to see the stream inside the salon, they do have to join the salon. Um if they
297. 54:45just want to watch me talk, um they can do that, but they're not going to be able to see anything I'm showing on screen.
298. 54:52So anyway, that's coming next week. Pretty exciting, huh? All right, what other questions do you have? Got about a half hour left
299. 55:10on this fine Friday night. Linda Sherman, that's good. Thank you, Kyle. You're welcome.
300. 55:25Let me go look at um what we talked about in when is it time to buy new hardware now or hold? Well,
301. 55:40so here's here's how it goes in theory. How it goes in theory is you should wait for the next cycle because the next cycle the hardware is
302. 55:51going to be better, right? That's how it goes in theory. Here's how it's going in reality. Last April, you could get a 64 gigabyte maxed out
303. 56:06M5 Pro Max, whatever the hell it's called, 64 gigabyte machine for about four grand. Now it's about six grand for the same configuration.
304. 56:20And I have a feeling that by the time January comes around, that same configuration is probably going to be
305. 56:26like eight grand or 10 grand. Chips are not going to get cheaper. The demand for AI compute is about to skyrocket. Like there's
306. 56:37there's barely any adoption right now. There's barely any adoption. As as much as people are talking about AI and we
307. 56:45talk about it in here, there's barely any adoption. When these tools get really good and really powerful, people
308. 56:53are going to want more and more and more compute. There's not going to be enough compute and there's not going to be enough electricity
309. 57:00to power the compute. And so chips are going to get really expensive. So my instinct like I'm torn. Part of me wants
310. 57:08to say wait for the next round like like wait for summer of 27 for the M7 series from Apple. Except I just have a
311. 57:20feeling I just have this sneaking suspicion that the the prices are going to skyrocket. So I would get something
312. 57:26decent. Now, you can lease like you can get a a badass Mac studio with 256 gigs of RAM and like 4 terabytes of storage
313. 57:42or eight for about 250 bucks a month if you lease it for three years. um that's about the price of an
314. 57:51expensive AI thing and then you'd have a piece of iron that you could run some pretty significant models on. Um that's
315. 57:59a hard qu it's a really hard question, but just in general with in general with computers, always buy the the most maxed
316. 58:08out version of it that you can afford. And with AI, it's like that on steroids. If you think about AI is going to get
317. 58:20good enough, even the local models that you can install on your computer within two years, you're going to be able to
318. 58:27install a file on your local computer that is AGI. That's like having full-time employees at your disposal that you don't have to
319. 58:37pay anyone else for. So, if you want to be ready for that, I would say the time to invest is now. I have a Microsoft
320. 58:45Copilot PC that has Snapdragon chip on it. Codeex helped me hack it to harness a local LLM model that runs as smooth as
321. 58:53butter on that sil silicon. Oh, that's interesting. You have that, Brandon? Oh, wow. That's cool. I like it.
322. 59:07I can trade in my lease after 20. You can trade in your lease after 24 or 36 months. Um or you can buy it out.
323. 59:17Yeah. So like the lease is the lease is actually a really interesting hedge, right? If in if in 36 months you've been doing
324. 59:26this lease and you get the end to the end of your lease and the price of computers has skyrocketed. Um you can just buy it out and now you
325. 59:37don't have to pay for the really expensive thing. If if the performance has gone up and prices have gone down or
326. 59:43stabilized, you get to the end of the lease, you just turn that back in and now go get yourself a one and a half
327. 59:48terabyte M7, M8 beautiful all screaming, all dancing, you know, whatever. Um, Lori, I know was less costly when I
328. 1:00:00used to just make embarrassing Facebook [laughter] posts in the 90s. Oh my god. My god. My god. My god. Hey,
329. 1:00:11you want to see a fun commercial? Office hours remix. Yeah. Let me show you something fun. Um, move this over here.
330. 1:00:32Do that. And I'm going to do this. I'm gonna do this. Okay. So,
331. 1:00:47um today, this is from Mark Beni off this tweet. Today, Anthropic and Salesforce announced Clawude Force,
332. 1:01:03which is sounds like something out of a science fiction dystopian thriller, but Claude Force is a new AI
333. 1:01:14collaboration between Salesforce and Anthropic. And basically, you're going to be able to use super powerful claude
334. 1:01:22across your data in your CRM, right? But I want you to watch this commercial, this launch video. And don't forget, I
335. 1:01:31want you to I want you to put this video in context of the launch videos of like chat GPT4 and chat GPT 4.5 when you had
336. 1:01:42like the four awkward people sitting around the tiny table in the open AI offices passing a laptop between them.
337. 1:01:49This is what announcement videos look like now. So, let's talk about what seems to be on everybody's mind. The SAS apocalypse.
338. 1:01:58Yes, the long slow death of software. As we know, we launched the original9 that software is totally super.
339. 1:02:16>> Uh, hang on. Sorry. Audio is weird. I think I got to change something in my settings. >> [whistles]
340. 1:02:32>> I think this is right. Um, let's go back to the beginning. Just just tell me if this is better, Brandon. >> Let's talk about No.
341. 1:02:46Um, oh, did I maybe not. It's possible I didn't share right. Hang on.
342. 1:03:04Tell me if this is better. >> Sass apocalypse. Yes. Long slow. >> Okay. So, let's talk [music] about what seems
343. 1:03:18to be on everybody's mind. The SAS apocalypse. Yes, the long slow death of software as we know it. We launched the
344. 1:03:28original end of software in 1999. >> We're out here um trying to get the word out that that software is totally superfluous,
345. 1:03:35>> but now this one is different. It is not the end of software. It's the end of software that makes humans do all the work.
346. 1:03:45The end of bouncing between 20 apps to find one answer. The end of systems [music] that store information but do
347. 1:03:51nothing with it. The end of tools that demand your attention without ever earning your trust. So what's replacing
348. 1:03:59it? Software that listens. Software that understands. Software that can actually do. It's called Claw Force. And once you see
349. 1:04:10it, there's no going back. Because work was never supposed to be this manual, this fragmented, this dependent on
350. 1:04:19people [music] stitching together one disconnected step after another. Cloud Force changes that. When the world's
351. 1:04:26number one AI meets the number one CRM, software stops being a destination. It starts becoming a teammate. Not another
352. 1:04:33place to go. A presence [music] right there with you in the flow of work at the moment. work is happening there.
353. 1:04:40When you need context, there when you need your CRM dashboard, that's when work changes. It gets lighter, sharper,
354. 1:04:48more live. Workflows that move with intelligence. Processes that adapt in real time. Systems that don't just
355. 1:04:55record the business, but help run it. Sales, service, marketing, commerce, every function, every industry.
356. 1:05:05This is where software becomes action. This is where decisions become outcomes, where the AI you want meets the trust
357. 1:05:14you need. Imagine having a teammate in the moments that matter. I'm talking about your own AI CRO day or night,
358. 1:05:2424/7, 365 days a year. And when you've got a colleague like that, will work starts to feel a little different.
359. 1:05:32Suddenly, you're looking forward to Monday. You know what's going on, what matters, what comes next. Because Claw Force [music]
360. 1:05:40is there when you need an answer, when you need context, a second set of eyes, first draft, a better path forward, or
361. 1:05:48that last minute piece of information right before you jump on the call that you forgot was happening. Now ideas,
362. 1:05:54they don't stall. Decisions don't drift. Momentum doesn't disappear because you're in the zone. You're no [music] longer thinking alone.
363. 1:06:05And what's not to love about that? [music] This is the agentic enterprise. >> I mean, first of all, Ma Matthew McConna
364. 1:06:18is just cool. Um, but that's an incredibly well-ritten piece. It's an incredibly well produced piece that cost
365. 1:06:26them millions. It they probably paid him a million and a half for that or some god-awful amount. Um, no, nobody has the
366. 1:06:33attention span for this video. Totally true. [laughter] All right. All right. All right. Uh, but did they use AI? I don't know. They
367. 1:06:44probably didn't. That looks like fairly That looked like fairly straightforward production. Um, but that video for me kind of marks an
368. 1:06:55inflection point from, you know, awkward videos of like, gee shucks, we're just some engineers sitting around a table. Here's our
369. 1:07:04little model to this shit's important. This shit's going to change the world. If that line about this is not the end of software as we
370. 1:07:13know it, it's the end of software that makes humans do all the work. It's the end of having 14 different
371. 1:07:21systems where we're the ones that have to glue that together to make any sense of it. That's a really compelling promise.
372. 1:07:30And you got Matthew McConnA making the promise, you know. All right. All right. All right. You know, he's a good enough
373. 1:07:38actor that he can say that you know, word salad that they wrote for him. It was well written word salad, but it was still a mess.
374. 1:07:48But he at least made it believable, you know. He earned his $1.5 million. [laughter] [laughter] Oh my god.
375. 1:08:04So anyway, that was that was one that whether or not Claw Force is interesting in any way, I don't really care about
376. 1:08:13it. It may be, it may not be. They certainly can't make Salesforce any worse than it already is. I, you know, we run our sales
377. 1:08:23um function at StoryVine on Salesforce and it is the most horrific software on the planet. It is just abysmal.
378. 1:08:33And so if Clawude Force can just make it so you can just talk to it and happens and it doesn't screw it up and
379. 1:08:40it glues together all the piece parts, great. But for me, the significance is, you know, we we've sort of shifted if we go back to 1994,
380. 1:08:55you know, to to whatever 2010, sort of the runup of the worldwide web. I feel like we just graduated from like 1997 to
381. 1:09:061998. Like 94 to 97 was just it was all the geeks and the nerds and the you know just people trying to figure it out. 97
382. 1:09:16the money showed up and 98 was when commercials like that started to show up, right? It starts to become a real business, real investment.
383. 1:09:28Um, and so I think that's kind of where we are there. Let's see. No one likes Salesforce. That may be the last paid celebrity to
384. 1:09:39endorse a product for money. Well, no. You know what's going to happen, Kelly, is celebrities are going to keep endorsing
385. 1:09:48products for money, but they're not going to have to go do the photo shoot. Like that's probably the last one where
386. 1:09:54Matthew McConna has to go to a soundstage and say that right? But at some point here, he's just going to license. There's going to be a
387. 1:10:04Matthew McConna.com website and if you want Matthew to be a spokesman for your product, you'll apply. You'll say,
388. 1:10:11"Here's how much I'm paying." And then Matthew will say yes, no, yes, no, yes, no. And if if he says yes, then the AI
389. 1:10:20will go, you know, put his likeness in your commercial and then he'll get final approval and if he approves it, then you
390. 1:10:27pay him. That's that's where it's going. That's what you mean? Yeah. Yeah. Yeah. Amazing. Our current ERP ERPI SAP and the biggest
391. 1:10:37complaint is difficulty in getting getting the data in and especially getting the data out in a meaningful way. Looking forward to having a
392. 1:10:44capabilities. Yes, this is and and that's that that was one of the lines in there that that MCC said um
393. 1:10:53these systems were designed to put data in, but there was essentially no way to get it out. Well, that's changing. Great.
394. 1:11:05AI as a simplicity layer on top of the garbage software you've been over bloating for the last 30 years. Sign me up. >> [laughter]
395. 1:11:17>> Okay, what else do we have here? The micro duck. Oh, this is good. Okay, let me move. Let me let me do some other things here.
396. 1:11:39That's another video. We get another video now. All right. All right. All right, everybody. Here we go. All right. Okay.
397. 1:11:50Let me let me tell you what we're looking at here. So, Anthropic today. So about a year year and a half ago I think it was like
398. 1:11:58a year and a half ago anthropic released MCPS model context protocol which is basically a protocol that that is is a quasi replacement for
399. 1:12:12the for APIs the application programming interface where you can basically hook up your large language model whether
400. 1:12:21it's claude or chat GBT through the MCP protocol to any piece of software that you know makes their software accessible
401. 1:12:31through the MCP protocol and like once it's hooked up let's say it's Salesforce for example you can just say to claude
402. 1:12:39or chat GPT hey go into Salesforce and find me this that or the other record and through the MCP it'll go into that
403. 1:12:46software well today Anthropic announced MHS the model hardware standard and so it allows for Claude or Chat GBT through
404. 1:12:58any piece of hardware that follows the MHS standard or I guess whoever writes a connector in MHS format um to talk to
405. 1:13:07any piece of hardware. And so here's the launch video for this. This is this is really astounding. If you've used MCPs,
406. 1:13:13they're pretty amazing where you just say words into chat GPT and it goes and uses software and comes back with a
407. 1:13:20result. This now does that for hardware. Now, experiments that took weeks take days. That can enable scientific breakthroughs much quicker.
408. 1:13:32>> For the past year, [music] we've been putting AI to work inside labs and manufacturing facilities. We kept hitting the same wall.
409. 1:13:42There's no common way to connect a model to [music] physical equipment. The model hardware standard changes that. MHS
410. 1:13:48gives any AI model one standard way to connect with and operate devices. If you wanted your instruments to work
411. 1:13:55together, say a camera and a stage of a microscope, you had to build a custom [music] software integration between
412. 1:14:01them. Every device you add complicates the picture for complicated experiments. That's weeks of work. With MHS,
413. 1:14:10>> wait, I just want to I just want to stop right here. So, so imagine every one of those lines is custom software
414. 1:14:19development. So, if you're running a research lab and you've got, you know, a centrifuge and a oscilloscope and all
415. 1:14:27the different things you're writing custom software or trying to integrate custom software to get all of these
416. 1:14:32things to talk to one another. So, watch what MH what is it? MHTT, whatever it is. Watch what this protocol does.
417. 1:14:41With MHS, each device connects once through one interface. The setup is accelerated by the agent. Any device
418. 1:14:49that speaks MHS can talk to anything else that does. Devices communicate at bare metal speed. The agent has access
419. 1:14:57to this context and can control operations. >> Here we're seeing Claude controlling a Leica microscope on its own. It's
420. 1:15:05focusing the microscope, [music] looking for a bacteria, and deciding what to capture next. At Genentech, a scientist
421. 1:15:13designed an experiment in a PDF, dropped that PDF into Claude, then Claude autonomously executed that experiment.
422. 1:15:20When things went wrong, it was able to recover even overnight. The iteration cycle is much faster and the focus for
423. 1:15:28me as a scientist shifts to actual scientific questions. Once everything is connected, running an experiment looks
424. 1:15:34like you ask Claude to start it. Here we're looking at images of neurons in the brain. With MHS, I can instruct the
425. 1:15:44microscope to move around, go deeper, or even take side views all in real time. If you can test hypotheses faster, you
426. 1:15:53could create general technologies faster, um, like finding new materials, also solving important scientific
427. 1:15:59problems. This is how a century of progress can compress into a decade. >> Crazy. Right. So, um,
428. 1:16:23I mean, it's kind of mindboggling, right? We we're we're we're going from software where we have to learn the
429. 1:16:32interface and we have to do the things to chat GPT where we have to prompt it and it will give us answers to agents
430. 1:16:45where we can just give it a goal and it will go off and do stuff and it will go off and use tools software
431. 1:16:55to go off and do those things. And now this is bridging us into the into the real world, right? So as the models get
432. 1:17:02good enough to come up with a hypothesis and design experiments, now this thing the this protocol hooked to the right machines will be
433. 1:17:14able to just go do those experiments. crazy. Our jobs are now managing the bots. Our jobs now, Kelly, are to have the ideas.
434. 1:17:31Like the the scientist said it in the in that piece. The scientist said, "Our job is now to ask the interesting scientific questions."
435. 1:17:43Our job now is to ask the interesting questions. What do I want? What do I want to do for the world? For
436. 1:17:53my community, for myself, for my family, for my doggy. What do I want to do? What do I want to do? That's our job.
437. 1:18:07Doesn't seem like that right now. Right now, it seems like I've got to learn all this Just take a breath. Calm down.
438. 1:18:18Look at where this is headed. Look at the trajectory of where these lines are headed. Where how these lines are going up. The up means
439. 1:18:28the machines are going to be able to do most of it. So again, back to the question, if tomorrow you had 10 employees
440. 1:18:44that could operate at a world class level, what 10 employees would you hire and what would you ask them to do?
441. 1:18:55So on Friday night date night, one of the things that we do is we assign weekend homework. So, here's your weekend homework.
442. 1:19:06And you can take this as seriously or not. You don't have to do it. I don't give a if you do it. I'm not going to grade you.
443. 1:19:13But the homework is come up with some answer to that question. If tomorrow I had 10 worldclass employees that worked for me
444. 1:19:23247, seven days a week, forever, what would I build? What would I do? What would I ask those employees to do?
445. 1:19:35Come up with an answer for that. And here's the thing about this exercise. There's actually no wrong answer. The
446. 1:19:44only wrong answer is not to have one. Not having an answer to the question, what do you want to do? What do you want
447. 1:19:53to create? What do you want to put in the world? What do you want to have these employees do for you? That's the only wrong answer.
448. 1:20:00My surgery was all the result of using LLMs to research my symptoms and the doctors agreed. Amazing. That's wild. Wild. Yeah,
449. 1:20:13something's up. Hey, 10 employees, help me figure out what the hell's was going on with me medically. Hey employees,
450. 1:20:20help me start this business. Hey employees, help me come up with a creative fundraiser for the church.
451. 1:20:29and I want it to be world class kickass. Okay, so that's weekend homework. Um, bonus homework. Every month we have a a
452. 1:20:40uh a theme. The theme for August was taste. And in challenges and competitions in the AI salon, um, there's a there's a competition in
453. 1:20:52challenges and competitions for the theme of the week, taste. So taste is the kissing cousin of what do you want
454. 1:20:59to do? Right? What I want to do is X. AI goes and does that stuff. It's your job then to have taste of did it did those
455. 1:21:08employees do it well. So these are all related kind of themes. Okay. The other thing mark your calendar. Next Tuesday is AI Salon
456. 1:21:19Presents. Uh we have this amazing guy named Price who lives in the San Francisco Bay area. He worked as a
457. 1:21:26product guy for years, realized that AI was going to take his job and rather than become the victim of that, he said,
458. 1:21:34"What is the job that is the least likely to be impacted by AI immediately?" And he said, "Roofing." So, he started a roofing company. Um,
459. 1:21:46but he's really smart and he's all in on AI and he's starting a roofing company. So, if you want to come here, if you
460. 1:21:53want to come experience the future of work, AI Salon presents next Tuesday 5:30 to uh to 8 Mountain Time. 5:30 to 8. 5:30
461. 1:22:07to 7, sorry, 90 minutes. Um, next Tuesday. Um, it's going to be a really good one. Uh, Price Fulier is his name.
462. 1:22:17It's going to be really, really good. All right. So with that and then Monday we start the new format. So Monday's going to be
463. 1:22:24play first. So we're going to start playing with some stuff and we're going to have the secret super secret back
464. 1:22:29channel which I'm kind of excited about and I hope you'll all enjoy. With that I think we're out of here. The pre-show
465. 1:22:37starts a little late and only on TikTok. Yeah, exactly. I don't even know if we'll do a pre-show. I might I might
466. 1:22:42drop the pre-show alto together. That we'll figure out. But we're trying some new things and uh yeah, I hope you're well. Okay,
467. 1:22:51this chat is really messing up for me. Okay, beautiful price folder. There's the link. So, go to the events in the AI salon. Please
468. 1:23:02RSVP for it so you get reminded when it's there. And that's it. Beautiful. Hope you had a good time tonight. Uh,
469. 1:23:11happy Friday night. Have a beautiful weekend, everyone. Um, go think about the hypothetical. If you had 10 worldclass employees tomorrow,
470. 1:23:24what would you have them do? Who are they? All right. Peace. It'll be empowering. Have a good night, everybody.