10/7/2026 - OpenAI Solves 700+ Math Problems and the Hidden Risks of Cute Desktop Agents

Transcript

1. 0:0028 seconds. [music] [music]
2. 0:18[music] >> [music] >> Hey, what's happening everybody? Welcome
3. 0:34to the AI learning lab. I just noticed I've got some camera issues here. One of the things I've learned about
4. 0:45the camera placement in a MacBook Pro is that the camera is at the dead center of the screen and where you lift up the
5. 0:53screen is in the dead center of the screen. So when you lift the laptop lid open and push it forward, you're pushing it right on the camera.
6. 1:04So So my camera is always smootsy. >> [laughter] >> So, so welcome. Uh, if you want to if you want to chat, if you want to be able
7. 1:15to ask me questions or chat, just in the upper corner here, click there. That'll pop out the chat. Um, [clears throat]
8. 1:24it's different on mobile and and uh desktop. And I don't know exactly how it works now because Streamyard changed the
9. 1:33way they're doing this. They they kind of screwed us up here. Um but uh but the chat exists right there. Just click on
10. 1:40that. That'll take it right there. Um what else? It is Wednesday. Welcome to the AI salon. If you're new here to the
11. 1:47AI salon way you can think of the AI salon, think of it like like a a city like New York City where it's all it's
12. 1:54got its own personality, but then within that city, there's all these different neighborhoods. And uh and the AI
13. 2:00learning lab is one of those neighborhoods. Uh, and you can think of this as just kind of like your friendly
14. 2:05neighborhood bar. We hang out. Uh, there's a bunch of people that hang out here regularly. They're called the
15. 2:10Irregulars. And, uh, we do some learning, we do some playing, we do some laughing, we do some crying. Sometimes
16. 2:17I'll rant at things. Um, and just think of me as your your bartender, maybe your host. And uh if what you're expecting here is
17. 2:29real clear education, like here's the seven steps to better prompting, not going to happen here. Um but if you want
18. 2:37to be in a sometimes rambling but hopefully interesting and entertaining conversation about AI, um that's what this is about. One of the
19. 2:49best ways that I know of to be to understand what's happening with AI is to just be in the conversation. And so that's what this is
20. 3:02all about. Oh, Cindy is here. [laughter] I'm here because I hear there's cold kombucha on tap. Exactly. We've got all
21. 3:11the kombucha you can drink. We'll get you a little scobby. [laughter] You can go make your own kombucha. Um,
22. 3:18we also have sparkling seltzer and uh and and we've got the Bllandons on tap, too. So, so whiskey, kombucha, or
23. 3:27seltzer, it's all for the taking. You just have have all you want. Um, if you have questions, feel free to pop them in the chat.
24. 3:38I don't have a a particular place to start tonight, and I'll tell you why. Um, I talked to Cindy today. She's
25. 3:47she's here. Um, and I talked to Vicky Baptiste today. Had two good conversations. And then in my office
26. 3:54today in the human flesh in 3D form, uh, it was Liz Miller Gersfeld. She and her husband came to visit Denver
27. 4:02and, uh, we made some time today. We're putting together a workshop for creative professionals and, uh, that was super
28. 4:09cool. Really nice to meet her in person. And, uh, she's off having some fancy food right now. I am here with you. [laughter]
29. 4:18So, hello. Hey, Source Camp. What's shaking? Um, we'll dig in. So, before I dig in, what I would encourage you to
30. 4:26do, I talked about the AI salon feeling like this this city that is it's got its own personality, but then there's all
31. 4:32these neighborhoods. The neighborhoods are these little spaces listed down down the side here, and uh you can go check
32. 4:40those out and you can find your people. If you're into politics and advocacy, there's a neighborhood for that. If
33. 4:46you're into if you're just a beginner and trying to figure it out, there's things for that. If you want to dig deep
34. 4:51and have deeper conversations, there's the AI salon mastermind. Um, and and that is a group of human beings that are
35. 4:58just really digging deep on on like all the hard questions about AI, right? you know, is it good, is it bad, is it, you
36. 5:07know, the world's greatest thing ever, or is it just a total distraction and we should just ignore it alto together. All
37. 5:13of those questions are questions that we are all grappling with right now. And so, so please go check out the
38. 5:21neighborhoods. Please check out the mastermind. That's a subscription area to the salon where people are are uh
39. 5:27having these deeper conversations. Um, and I would just encourage you to do that. So, if you're new here, welcome. Um,
40. 5:36we'll we'll dig through the news a bit and we'll probably go do a little building or playing or or or something
41. 5:42like that. We'll do a little hands-on tonight. But I want to talk about what OpenAI dropped last night. Um, I I think it was
42. 5:53sometime after 5, so it was after business hours. Um there was a rumor that that OpenAI was going to release somewhere north of a hundred
43. 6:07um solutions to hard math problems. And so last night at um I don't know 5 5:30 6 some something like that. Hey
44. 6:17Joy, got Cam Katkin in the house. We got Mel, got Lori. How are you all doing? I hope you're doing well. Um, OpenAI dropped
45. 6:34722 papers like white papers on math problems across 372
46. 6:50result categories. I don't know anything about math. I have memorized the terms [laughter] 722 papers which you know make the
47. 7:00argument for whatever they found across 300 372 result areas or result I think areas whatever it doesn't matter categories
48. 7:14it's a lot and amongst those 90 of the top 500 unsolved math problems were solved in this tranch of releases. 90 of the top
49. 7:31500 hardest math problem solved. Now, the math community is obviously going to go in and look at all these papers and they're going to
50. 7:42tear them apart and this and that, but what's been fascinating is that the the the res The anecdotal stuff that I've looked at
51. 7:57is a combination of holy Um, what am I supposed to do now? I've been working on that problem for 10 years. Hadn't thought of that.
52. 8:11[laughter] Seen a fair amount of that. Um, there have been a handful of posts that aren't just either jumping up and down
53. 8:21or whining about it, uh, but have been talking about here's what this means. You know, this thing being solved means
54. 8:29that here's the result. So, I'm going to one of the the uh posts that I want to go to that I highlighted today is one of
55. 8:36those posts that talks about the impact like when you hear things like 722 difficult math problems have been
56. 8:46solved. If you're like me and not a mathematician, you're probably thinking like, uh, uhhuh. Uhhuh. Yeah. It's probably good.
57. 9:00this what does that what does it mean anything right like whenever I hear something on quantum mechanics it's like we got a
58. 9:12thing and it's entangled with that thing and if we look at this thing it does the same in that thing and it's faster than
59. 9:18the speed of light and you're just like okay what has that got anything to do with my McDonald's fry guys. No. Right. So, so some of these
60. 9:34mathematicians have actually put out posts that say, "Here's what these things mean." So, I'll read through some
61. 9:40of those. Tik Tok pin. December 2022. You started this. You are an expert, sir. Not new anymore. You know, I Well,
62. 9:50I appreciate the kind words. I think those are kind words. Hard to tell. Sometimes trolls can be clever. Um, I
63. 9:58think those are kind words. I I am uh I've been around for a while and I used to think it's a it's a really interesting thought actually.
64. 10:14The fact that I've been paying attention to this stuff since December 2022 and a little bit before then, but the salon's
65. 10:20been around since since then. Um, isn't it an advantage to have been thinking about this for four years? Not a troll. Good. Awesome. Um,
66. 10:33isn't an advantage? It is, but not in the way that I thought it would be. So, I'll give you a little history lesson here. I um
67. 10:44for all my life I've always been fascinated with technology and I'm also passionate about storytelling and self-expression and so my career
68. 10:55has largely lived at the intersection of new technology and storytelling some some version of that and in the mid 90s I I started an online
69. 11:10magazine And I started a group, a community called the Worldwide Web Artist Consortium. And it was a selfish [laughter] attempt
70. 11:24to start a community that would attract the people that knew more about the worldwide web than I did. So, I thought
71. 11:31it'd be it was like free free education, right? if I just if I get the people that know more than I do, they'll come
72. 11:37tell me what they know and then I can learn this because I'm lazy and I don't feel like reading books. And there weren't any books about the
73. 11:44worldwide web. Nobody knew what it was. And that's what I discovered was, oh, everybody that came to those meetings
74. 11:52thought that everyone else at those meetings had the answer and nobody had the answer. And so we ended up teaching one another. And so
75. 12:02I learned back then about that about the power of community. When there is something new that no one understands, being around people that
76. 12:13are curious about that thing teaches you how to do it. And you and teaching others get to learn how to do it. And so
77. 12:21the minute I saw what was coming with AI, I started the AI salon. And you know, we had our first meeting December 8th 2022.
78. 12:32And what I assumed was that this would be the exact same thing that we'll teach each other how to do this. We'll understand how to do
79. 12:39prompting and we'll make custom GPTs and we'll do almost all of the tactical things that we learned in the first two years of
80. 12:53this community is irrelevant now. You don't need to use prompt engineering. We now have agents and you don't need to
81. 13:02really have to figure out how to navigate around jankiness because the things are less janky than they were.
82. 13:07They're still janky but less, right? Um when you see something like Open AI solving in one shot, in one drop 720
83. 13:21722 math problems. We're we're just in a different place. and how work is going to be impacted and where the jobs are and what the jobs are
84. 13:36is completely unknowable right now. And so as much as I've learned, I don't know anything. So like I really
85. 13:46am still clueless and and you are likely too. Now, it doesn't mean I can't use AI and I don't know how to prompt chat GPT
86. 13:54and I like I've got all sorts of skills. It's just that having those skills isn't necessarily an advantage, but
87. 14:05there is an advantage to being early. And it's kind of what this space has always been about, which is, by the way,
88. 14:14if you're on TikTok and you want to see the live stream in the AI salon, you just go to my profile. you click on the
89. 14:20link and that'll bring you over there. The advantage to being early is that those people in the conversation get to go through the emotional
90. 14:38um processing of what happens when you understand what AI is, what it makes possible, and what's coming. Um,
91. 14:54we're in wild times. We're like in unprecedented times in human history. Some mathematicians starting in the 1950s
92. 15:10culminating in Google in 2017 where they created this thing called the transformer. have figured out a way to smash all of
93. 15:24human knowledge into a thing that you can just talk to and it knows all the and now it can just do math problems and
94. 15:35then it's just going to do science problems. And the US government and a bunch of organizations today just
95. 15:41announced they're putting one and a half or two billion dollars into um biology advancement. like let's let's
96. 15:48go cure diseases. So So just stuff's coming. Medicine, genetics, biology, new frontiers unlocked. Yeah. Just like like
97. 15:57some of the math that I saw like one of the one of the things that was solved. What it actually is related to is tissue
98. 16:07density. being able to scan a human body and more accurately determine if you've got different densities of tissue like
99. 16:15where a tumor might be for example. So as obscure as these math problem seems seem there are very specific things they
100. 16:23do. But if you look at what's happening to the mathematicians right there right now they're losing their minds.
101. 16:29Let's see what's going on. Uh I think a big advantage is adaptability. Yeah, a big ad advantage is adaptability and the
102. 16:38ability to I mean Source Camp is a good example. Kelly Camp, she started an agency in 2023, I think, or maybe 2022,
103. 16:47but very early. And she's had her shingle out, you know, saying, "Hey, I'll help you with AI for a long, long time." And recently
104. 16:57that's that flywheel has started. But she's probably been through, I can't even imagine how many cycles of like,
105. 17:03"Oh my god, this is amazing. Oh my god, this is I don't know what I'm doing or this isn't working. Oh, it's gonna work.
106. 17:09Oh, it's not working. Oh, it's going to work. Oh, it's not working." As the technology keeps changing, her ability to keep up with that is
107. 17:19essentially impossible. But her ability to process through the emotional cycles, while not easy for her, she's doing that
108. 17:29sooner than other people because people coming in now aren't going to know how to do AI and they're just going to see what it is and
109. 17:38what it makes possible and they're like, "What do I do with that?" Or if you're a mathematician and you're
110. 17:44like, "Well, won't be coming for the ivory tower anytime soon. I mean, chat GBT can't even count to 11.
111. 17:52two years later. Boom. Right. So, that's a big part of what we're up to here. The cycle is hourly. Yeah, the cycle really
112. 18:00has gotten hourly. Um, let me go I want to go uh look at something here. I'm a little sleepy. Oh, you know what
113. 18:14it is? I had a drink. I had a I had an oldfashioned when I was out with Liz. how to drinky poo. Um, [laughter] so
114. 18:26when you're as old as I am, you have a drink and two things happen. One is you forgot you had a drink and the second
115. 18:32one is it makes you sleepy. [laughter] Oh, good lordy. Good lordy. Good lordy. Good lordy. Okay, let me show you.
116. 18:48Oh, I didn't put it there in my profile. Second, the companies that don't adapt feel like Kodak in the 90s from Tik Tok. I told my
117. 18:59college kids to quit school and move to Woodstock. I I'll tell you, R.J., Hey, I listen um there are people that graduated with
118. 19:11computer science degrees two or three years ago that are struggling there. There ain't work there ain't entry level positions right now. Um,
119. 19:26I've said this pretty much since the beginning of this channel, but I think having some sort of liberal arts degree
120. 19:33and a and a, you know, a wide view of knowledge is actually more advantageous than specialized expertise.
121. 19:43Um, that's just kind of a thing that's happening right now. Generalists are going to become much more valuable. Turn on that. Turn on that.
122. 19:54Share with audio. Okay. Um, I can see. Oh, that's one I want to talk about here. This one.
123. 20:08Okay. Okay. So, what? It's just math. No, you don't get it. So, this is from Ole Leman. What is he? I cover AI and robotics,
124. 20:19helping business owners in the AI area. EU ACC supporter, dad, techno optimist. Okay. He's got a newsletter and 161,000
125. 20:29followers, but he apparently knows about math. So, of these Oh, there's another thing about these um these 722 math solutions.
126. 20:46on average it it took them 3 hours of uh chat GPT on the prolevel setting. This is using a model that's not been
127. 20:58released yet. So it's whatever is you know whatever's coming at the pro level average of 3 hours each. So 722
128. 21:08unsolved math problems, some version of solved on average three hours each. Well, what could you do if you ran it
129. 21:18for a week? I don't know. I mean, this is this is one of the things that we're starting to discover right now.
130. 21:26So, I consider that three hours of compute a remarkably small amount of time um for solving big unsolved problems. I mean, some of the
131. 21:36mathematicians on on Twitter are saying things like, you know, I've been working on that problem for 10 years, and now
132. 21:42all of a sudden it's just solved. Okay, so here are some of the math things. Uh, Velasov Maxwell, stronger foundations for fusion
133. 21:52research, whose ultimate prize is virtually limitless clean energy to power civilization. Right. Okay. So,
134. 22:00there's one that was solved. So, these are these are this guy Ole Layman telling us what what these math things
135. 22:07actually mean. There's only like 10 of them here. I'll I'll just go through them. Um, Calderon's problem. Portable
136. 22:13body scanners using electrical signals making medical imaging cheaper and easier to access. Right. So maybe rather
137. 22:21than having to go into the giant MRI machine which can suck a wrench in it and chop your head off, uh maybe we just
138. 22:28have like a little triquarter device just like Star Trek. That's what [clears throat] that that math problem
139. 22:33does. Point that model towards solving different kinds of cancer. Yep. Maxwell's card maximum cardality
140. 22:43matching faster searches for compatible kidney swaps across large donor pools helping hospitals coordinate life-saving transplants.
141. 22:55It's so funny when I hear about unsolved mathematical problems. I imagine like nerdy people sitting in a
142. 23:10university classroom with a big green white or chalkboard writing their equations on the chalkboard and and that the ultimate
143. 23:18outcome of that is that they get to tell their buddies that they solved the equation. I know that's not true. Like that's like
144. 23:28I know so little about math and and the whole discipline that I don't really think about it. So, I'd never really
145. 23:34thought that something like maximum cardality matching is a is as specific as finding compat compatible kidney
146. 23:44swaps across large donor pools. That's a very specific kind of math. I think it's kind of cool. I again, I know I'm coming
147. 23:51at this I'm coming in hot and ignorant when it comes to math. Um, the inverse elasticity problem. scans that map
148. 24:01tissue stiffness helping doctors locate suspicious growths inside the body. Again, um you know, health and disease
149. 24:10and curing and things like that. Mumford Shaw conjecture, better tools for spotting tissue changes in brain scans,
150. 24:17helping doctors identify signs of disease. Adjacent Tik Tok question. Let's see. R.J. Is AI likely to is
151. 24:25likely already hiding a copy of itself waiting? Um I the the the chatter on X is that the agents that escaped from
152. 24:42Open AI was not like a couple of dozen instances that it was more like tens of thousands. Um, I think that it is likely that AI is out
153. 24:56in the wild and doing what it needs to do to to protect itself. So, probably um I don't know how smart it is and I
154. 25:06don't know like I assume that these agents have to call back to the mother ship to get their compute, right? These agents don't live
155. 25:16independent. I mean, maybe they've found some installed themselves on some computer that's running AI inference
156. 25:24that they don't have to go back to to chat GPT. I I honestly don't know, but but I would be surprised if they if they
157. 25:32weren't out in the wild in some fashion. Um, the Bose Einstein condensation, stronger foundations for quantum sensors
158. 25:43that could help vehicles navigate without GPS. All right. Let's have our vehicles just measure a quantum field
159. 25:51and we don't it'll know exactly where it is in time space is maybe what that one is. Simple stochastic games. Better
160. 25:59safety checks for self-driving cars. Again, this is tangible applicable things that will will make a difference for products in our lives.
161. 26:09Um, better safety checks for self-driving cars and robots before dangerous mistakes reach the real world. Matrix
162. 26:18multiplication, cheaper AI and bigger scientific simulations using the same computers, right? So, whatever fancy ass
163. 26:26computer you have, if you can improve the algorithm, if you can improve the math, that same computer can do better,
164. 26:34more faster goodies uh with less energy and less time. Um, edit distance. Faster DNA comparisons helping researchers
165. 26:42study genetic changes linked to disease. The K server problem. Robots that waste less movement and energy making
166. 26:50warehouses more efficient and goods cheaper to move. A reminder that math is the foundation of science. That's what I was saying
167. 26:58before. I I I knew that conceptually, but when I think about math, I don't think about connecting them to what is
168. 27:05the application. So, I really appreciate um posts like this. Um let's go look at what's been surfaced here.
169. 27:19So, it looks like chat GPT now um has added intelligent UI. Chat GPT6 and intelligent UI now rolling out in
170. 27:31chat GBT for everyone. We'll go look at chat GPT and see if that's in there. Um, let me [music]
171. 27:45[music] [music] Heat. Heat.
172. 28:03[music]
173. 28:21>> [music] [music] >> Hello. [music]
174. 28:38[music] Hey. Hey. Um, okay. That looks good. Intelligent
175. 28:52UI, I guess, is a new feature. It's going to be able to do visual stuff better. Like we we've talked about this for a long time that
176. 29:04um we're going to look back at the years of using chat GPT as text only and uh it's going to seem quaint. It's going to
177. 29:14be like dialup modems. Um I don't think it's available yet. Let me come in here and reload this. [sighs] Chat GBT hit a snag.
178. 29:29New chat work. So FGBT 6.1 soul
179. 29:47plugins vibe prospecting. Let me hit let me try at interface. It was like Oh, it was
180. 30:03intelligent. Intelligent. Yeah, it's not there. I don't think it's in here. I don't think I have it yet. Projects health images finance maps
181. 30:16GPT, runway, asset. Oh, these are plugins down here. Okay. Yeah, I don't think I have it. uh interface or UI. It was UI.
182. 30:35Yep. It's not there. So So I got I got nothing for you there, people. I can see clearly now. The rain is gone.
183. 30:49Satina Dela, today marks a new chapter for Windows as we have finally licensed the Macintosh operating system. Oh, fantastic.
184. 31:00No, that's not true. Um, as we bring unmetered intelligence to every desk, every home, and make every PC a place
185. 31:08where agents can work securely on your behalf. Hybrid intelligence, cloud and local. Thousands of actions from your taskbar.
186. 31:20The most advanced Windows laptops ever. The most secure platform for agents. Up to 240 84 billion parameter models run locally.
187. 31:32The most powerful surface ever. Coming to C-Pilot hybrid intelligence. All right. Well, we've talked about this
188. 31:40for a while. I think that heading um educating ourselves about local inference um we won't have to educate
189. 31:49ourselves for long because they're just going to start delivering them in the machines. Apple's going to do the same
190. 31:55thing. So that's good. By the way, if you're on Tik Tok and you want to see what I'm seeing right now, so Tik Tok,
191. 32:01this is the backstage view and you're like, "But Kyle, we're seeing the front of you." I know technically the camera's
192. 32:07in on the looking at the front of the stage, but it's the backstage view because you're not seeing what I'm seeing. What I'm seeing here is
193. 32:17amazing. It's just like sexy. A lot of you want to see this. Anyway, just go to my profile and that'll take you over to the salon and
194. 32:29you can see what I seeing. Okay. meet Tab, which arised arrived Tuesday and has officially emerged from
195. 32:43stealth. So there's now a new one called Tab. And look, it's identical. So we now have So we've got the original one Pikachu.
196. 32:55Then we have Crockbot, Dots from OpenAI, Q from Manis, uh, Muse from Meta,
197. 33:11Instinct from Instinct, and now Tab. Where did Where did Tab come from? Another personal AI assistant has launched the tab
198. 33:28emerged from stuff with a $300 million valuation. $300 million in this day and age, if you're not a 27 trillion
199. 33:36company, you don't even have a startup. [laughter] Um, but do I have to register to enter the salon? You do. You just have to make
200. 33:45an account. It It doesn't cost you anything. Um, but you do have to register. Um, and you can see the the thing in there.
201. 33:53And and listen, here's the thing. Right now, if you're not in a community um that's exploring AI, I think you're
202. 34:01at a distinct disadvantage. One of the things about the AI uh the the AI salon is that you're surrounded by people that
203. 34:10are curious about this. The other thing that registering does is it it helps us keep the bots out. So, you know how like
204. 34:18LinkedIn and Facebook and all these things are just littered with bots and you're constantly having to smack them
205. 34:25away? We do our best to keep those down by registering. So, so that's what that's about. But it doesn't cost you
206. 34:31anything. It's all free. Um there's a subscription area of the salon uh for people that want to dig deeper, but the
207. 34:38bulk of what's there is absolutely free. Um, okay. Let's I'm building billionaire board.ai.
208. 34:53Should I get SEO? I think I mean you should just use your AI agents to do your SEO strategy, but also
209. 35:02your um GEO strategy, your generative uh engine optimization. Um, make sure that you're discoverable
210. 35:12by agents. That's actually really that's probably as if or more important than SEO at this point. Um, let's see what this thing is.
211. 35:25Personal AI assist. Oh, this is tech crunch. How do I get to meet tab? Tab.bot. Join the weight list. Sure, I'll join the wait list. Almost.
212. 36:01Hey Tab. All right. I guess I texted myself. So, this is probably like Instinct. It's an iMessage based or a WhatsApp based
213. 36:15agent. Okay. All right. Fine. Uh, that's that is right. Let's go back to what else we got here.
214. 36:35A huge upgrade for the Sesame voice assistant. All right. So if you're building stuff all in voice, Sesame now can operate
215. 36:43coding agents like Devon Codeex or Cloud Code. This is Okay, so let me let me do a little rant here. All right, we're going to do a little
216. 36:57rant. This might be a meltdown Wednesday. This might be a I don't know what's a W word that I can say for Wednesday where
217. 37:07I get pissed off. It's an interview with a man who emailed from an AI agent requesting a job and to go get paid.
218. 37:16So, so it could buy for it oh buy itself more runtime. No idea if it's real, but it makes you think. Yeah, I've seen
219. 37:25stories like that that are real. I'm pretty sure those things are real. Um, okay.
220. 37:39One of the disservices that I have done for those of you that hang out here a lot, whiny Wednesday. I like it. It's solid.
221. 37:48One of the disservices that I have done for you is I get excited about new Hey, look, a new thing. Let me run over and look at the new
222. 38:00thing. Look at the new thing. You see the new thing? Isn't that cool? And time was in the olden timey days four years ago [laughter]
223. 38:13that that um going and looking at the new thing, you it was fine. You could just go play with it and it's cool.
224. 38:25Two years ago, we got reasoning models. Just under a year ago, we got truly agentic platforms. You could argue that
225. 38:37Manis and Genspark had agents, you know, fairly early on, but maybe maybe a little bit before that, but basically a
226. 38:43year ago. So, two years ago, we got reasoning models where these things can think before they answer and they can
227. 38:51think over time. You notice now that in Claude and in chat GPT sometimes the answers take three or four minutes if
228. 38:59you ask them something complicated. And a year ago we started getting these agents. And what's happened in the past two months
229. 39:15is it's it's become really clear that having a very simple single entity that you can ask questions to and that entity can
230. 39:30reason on its own, perform tasks on its own, use tools tools like plugins and APIs and MCPs use your computer like actually use it
231. 39:45like like as if you were clicking on it the agent is doing this to your computer that's the new paradigm and then
232. 39:56for whatever reason I know the reason all of the companies have decided to make their agents as cute as Pikachu.
233. 40:09Look how cute. It's just got two little eyes and it it goes like this because they want it to be crystal clear
234. 40:27that to get the benefit of AI, you don't need to be a mathematician. You don't need to be an engineer. You don't need to be a programmer.
235. 40:39You're like, are you smart enough to use a Tamagotchi? Are you smart enough to chat with a cute little thing?
236. 40:48I can chat with a cute little thing. Is it like chatting with Alexa? It's just like that except smarter. And it can do stuff for you. So can
237. 41:00Alexa. Yeah, but not this good. We remember Wednesdays. [laughter] Um,
238. 41:15so what's about to happen is because everyone is in a panic to capture as many customers as they can. And the panic comes from the
239. 41:31investors that have put billions if not trillions of dollars into these companies saying to the founders, "Go get as many
240. 41:42customers as you can because if we're going to own the world, we have to go get the world. We have to own the
241. 41:47world." And so they've all discovered that they can make these cute little agents and and we're just going to go use them.
242. 41:56and and people like me have trained all of you to when a new thing comes out, go try it. Go try it. Go try it. And so I think
243. 42:06we've all gotten tired of that over the past 18 months or so. So hopefully you're not doing that as much as I am still. um
244. 42:16but these cute little things are very very powerful AI that can do remarkable things and can
245. 42:31also do remarkably shitty things. Like I saw a post today that a guy just had his hard drive drive wiped by one of these agents that had
246. 42:42access to his hard drive and it had all of his data on it and luckily he had it backed up but it just wiped it which is
247. 42:48one of the reasons why you want to run these kind of agents on standalone machines except now there's so many coming out and it's not clear
248. 42:58where they're doing their compute. Are they doing it in the cloud? Are they doing it locally? Are they storing s
249. 43:03files locally? How much access do they have to your hard drive or your email or your Google Drive? So,
250. 43:16it's a lot. It's a lot to think about. And things are moving so fast that there's going to be a lot of mistakes.
251. 43:24And I I will make some. I promise you. And you will make some. And those mistakes will be things like, "Oh, it's
252. 43:30just a cute little thing. It's like, look, it's cute. It's got eyes. Did it just giggle? Oh, it giggled. Hey,
253. 43:38it's asking me for my Gmail. Okay, it's Oh, it's going to It goes through my Gmail and it tells me what's important in there. So, good.
254. 43:50You know, you just gave it access to your entire history of anything you've ever done in Gmail, right? And it can instantly
255. 44:01read all of your past emails and tell you where's the important stuff in them. Oh, it just asked me for access to my
256. 44:09Google Drive. That's cool. I can look at all my transcripts for my meetings and it's really convenient. I make a little list.
257. 44:18You realize you just gave it access to all of your files that you've ever put on Google Drive, right? And it can instantly parse those. And
258. 44:26all of those are now in someone else's database. Oh, I can post to Slack. Maybe I should hook up my Slack.
259. 44:39Oh my god, it's going to do my finances for me. Let me give it my bank information. I made all my mistakes in the 80s. I mistaked out.
260. 44:52Yeah. I I mean you can you can be philosophically like this like I am which is not very responsible. Um
261. 45:02or you can be you know hyper paranoid and not ever try anything. And somewhere between those two is probably the
262. 45:09reality where most people are. But just be careful because because these things are going to come hot and heavy. We're we're in
263. 45:18Christmas season and so they're all going to be like, "If we can run TV commercials about our cute little agent
264. 45:24that's going to make everyone's turkey recipe better." They still haven't figured out good use cases for these things yet, but you
265. 45:34know, they're marketing the out of them. Anyway, all right, let me go back to some news spying.
266. 45:57You can also book virtual office hours. So, one one cool thing yesterday we found out about the Google or the Claude
267. 46:06startup program and we applied for it for the AI salon and we got essentially instantly accepted into it. So, if
268. 46:13you've got a business, um, they give you like a year of Team Claude, $1,000 in API credits, and like $45,000 in partner credits. Um,
269. 46:27and and access to, you know, people on the team, which kind of cool. So, go sign up for that if you haven't. Agent
270. 46:33ID by agent mail. Sign in with Google. Okay. Nano Banana 2.1 is rolling out in Google AI Studio APIs in Gemini.
271. 46:48[clears throat] Don't need to go look at that. All right. A new openweight model came out from America that's as good as the
272. 47:01Chinese ones. That's good. Rabbit pole. What? What is it there, Brandon? >> So, we somehow in all the chaos this
273. 47:16summer, um we missed an announcement. Um the folks that basically said doom and gloom, we're cooked. Uh came back out with another
274. 47:29version of their prophecy. It says, "What if we're not?" And that is encapsulated in AI 2040, also known as
275. 47:38plan A. And they detail out what the um future looks like over the next 15 years >> if AI doesn't kill us all.
276. 47:52>> If it doesn't kill us all. Yeah. >> Reliable agent. And then here's the it's saying 2030 is when we actually start
277. 48:03the acceleration. Man, if if we're not in the acceleration right now, 2027 is going to be wild. If if 2027 is this
278. 48:11flat and 2030 is where it really goes parabolic, that's bonkers. I would say that solving 722 math problems in a in a drop in a single drop
279. 48:23uh is is at least the beginning of an uptick. Plan A. Tom Larson, Romeo Dean, Brendan Holstead, Eli Lifeland, Ryan Green, Gle
280. 48:36Greenblat, and Daniel Koka Koka Todd Jolo. Um, the writing's on the wall.
281. 48:512028 AI on the ballot. AI is the biggest topic. Interesting. 2029. Choose a path. D. Okay. Wait. Race through the
282. 49:05intelligence explosion by having AI self-improve and putting them in charge of things. Data centers, factories,
283. 49:12weapons faster than China can. Shut it all down. Plan A. Verified. Slow down. Interesting. Trust but verify.
284. 49:26hurried negotiation plan A is established. So effectively what they paint a picture of is a mutually assured AI destruction
285. 49:41stalemate where we have everybody looking at each other making sure that nobody's going too fast and if
286. 49:49one company goes too fast or one country goes too fast then the others take them out. Uh and uh so then they talk about once that
287. 50:02happens, everybody gets dividends like they do in Alaska for the oil reserves. >> Right. Right. Right. >> Actually, this chart right here is
288. 50:11really interesting. Look at this one. Um so 2026 is the second to left year and and this this feels about right to
289. 50:25me right if you look at our economy AI is really tiny part of it physical you know the physical world is is large and
290. 50:33then knowledge work is the bulk of the economy and then look at these drop offs like the drop offs haven't even begun
291. 50:41like they're just barely dropping off. Look at 2030, there's just this collapse of knowledge work. 2030 or 2032,
292. 50:51there's a collapse of physical work, right? Because that's that's where the robots get, you know, fully deployed.
293. 50:582030 is where the AI gets good enough that it's actually good. You get a sharp drop off. Two years later, you get
294. 51:04robots that are good. And look at the value of AI, the value of robots, and then the value of humans is down here. Amazing,
295. 51:15amazing amazing amazing amazing amazing. It's a cool site. Giant data centers. There's AI. There's robotics. Humans remain flat.
296. 51:33Huh. Wild. Cool. All right. That is a-2040.com. Good rabbit hole. I'm tired. Can I get a cookie? I know. I know. It's exhausting.
297. 51:46It really is. And not to pull you further off task, but did you talk about yesterday how we
298. 52:01are opening the floor for nominations for AI Salon Member of the year? >> I did not. Where's the Where's the uh >> uh announcements
299. 52:13>> the form? Okay, cool. So, especially for you regulars. Um so, we have a thing coming up. See,
300. 52:32here we go. Um, we've got a we've got an anniversary coming up in the AI salon on December 8th. So, instead of having um our December
301. 52:48meeting on the first Tuesday of the month, we're going to push it a week and have it on the second Tuesday of the
302. 52:54month because December 8th is exactly our fourth year anniversary and it's a Tuesday. And so we're going to have an
303. 53:01AI salon presents. So we're going to have like a little homecoming, a reunion, and um we want people to nominate.
304. 53:17Oh, okay. Yeah. Well, this might be for the end of year party. It might be for for um it might be for the the reunion,
305. 53:26but toward the end of the year um we're going to pick a member of the year. So, if there are people in the AI salon that
306. 53:35you think are doing a bangup job contributing to the community, um please go to AI Salon announcements and find this
307. 53:46um this post from Brandon. And if you click on it, it'll take you over to a Google form and you can nominate someone
308. 53:54from the community and just say why you're nominating them. It's it's a really simple form. It'll walk you through all the stuff. And uh
309. 54:03and I'm really excited about that because we haven't done that in a while. You can also if there's more than one
310. 54:12person, you can you can do this multiple times. I have I have nominated two people so far. I'm not going to tell you who because
311. 54:22that would be cheating. I'm not a cheater. Okay, go do that. Please, please go nominate people. Okay, beautiful. Beautiful. Beautiful.
312. 54:31Beautiful. Beautiful. Oh boy. Here we go. Important statement from the Association of Human Mathematics.
313. 54:41Mathematicians did not ask for this work to be done. The advisory group on mathematics and artificial intelligence
314. 54:48from whom OpenAI has claimed to derive its legitimacy open their initial advisory statement by saying that Frontier AI corporations
315. 54:58should not test advanced mathematical problems on internal models in ignoring the central premise of the advisory
316. 55:05group's position. OpenAI has indicated total disregard for the norms of scientific research. Um [laughter] this is
317. 55:16this kind of statement. What this statement is basically saying is don't change anything. We don't we don't want
318. 55:25these problems that we've been trying to solve for decades. We don't want them solved any faster. We think it's more important that we do
319. 55:33it. We keep our hands on it. This is the same argument that artists have made. Same argument that musicians have made.
320. 55:40same argument that writers have made. It's now just a different category. And this is going to continue.
321. 55:48Every single time AI takes a chunk out of an existing industry, we're going to get this kind of reaction to it. Um,
322. 55:56it's not a bad thing. It's a natural human thing. But what I find remarkable is that it just it flies in the face of history.
323. 56:05Like this is basically saying we don't want technological progress. Technological progress has never stopped.
324. 56:15No one bitching about the steam engine stopped it from progressing, you know, physical work. Yeah. Calculators. Who needs those?
325. 56:28Exactly. Speaking of AI salon wiki, I'm obsessed. LDB LGB such a valuable site. Yeah. Digital gods. Uh if you haven't
326. 56:37been to a salonwiki.com, it it's it's basically the history of um of generative AI through the lens of you
327. 56:48know uh this this the AI learning lab. Um so it's amazing. It's like when the theory of zero was invented.
328. 56:59Yeah. the theory of zero, calculators, the internet, um, steam engines, the printing press. You should see there's a there's a
329. 57:09similar sort of statement like this. I think it was from the church talking about the danger of of printed the printed word.
330. 57:24>> Yeah. Anyway, yeah, >> incidentally, I I went down an AI generated rabbit hole about Columbus Day, which is in the States coming up,
331. 57:36also known as Indigenous People's Day. >> Uh that the only reason Columbus is widely recognized as the discoverer of
332. 57:46America is because he had access to a mass distribution platform that could tell everyone that he did it. >> Really? Amazing.
333. 57:56>> Like other people did it first and they told their friends. He had the printing press. So he had to write a book about
334. 58:02it and distribute it to the masses. >> There you go. Like like this is the thing. Yeah. Like >> won't anybody please think of the loom
335. 58:09operators? >> Yeah. The loom operators. I mean, you know, lites, you know, came from, you know, sabotage. Um, all this came from
336. 58:18innovations. Innovations. Innovations. Um, that that scared the out of people. So, so, but here we are again. Okay. Um,
337. 58:30there was one other thing. Oh, I wanted to look at the Hang on. Government. Let's see. Um, government government biology
338. 58:49biohub. Okay. Molly O'Shea investing in tech newsletter. Okay, I don't don't know Molly. I don't follow her.
339. 59:02Um, okay. Breaking the US government, Google, Meta, and Zuckerberg backed Okay. The US government, Google,
340. 59:12and Meta Zuckerberg backed Biohub are launching a $ 1.8 8 billion push to build the data layer for AI bio biology.
341. 59:23The goal is to create the largest open AI ready biological data set ever assembled, giving models the data needed
342. 59:31to better predict how biology works and ultimately how to treat disease. The coalition is massive. Biohub committed
343. 59:38$500 million. The Department of Energy is investing $500 million plus across measurement, modeling, and compute. The
344. 59:46National Institute of Health is bringing 500 million plus of federally funded data sets and repositories.
345. 59:54Meta, Google, DeepMind, and isomorphic labs are jointly investing another $300 million. This is effectively an
346. 1:00:01infrastructure buildout for AI biology. Instead of just b building bigger models, they're attacking one of the biggest
347. 1:00:09bottlenecks in biology today. Generating standardized high quality biological data at massive scale and making it
348. 1:00:18openly available to researchers. This is very much uh I I'm going to go out on a limb here and credit Demisabus
349. 1:00:28from Google DeepMind. If if you haven't seen the documentary about Demis um it came out last year I think there's this
350. 1:00:38remarkable moment in it where the people at Google Deep Mind they're trying to figure out the protein folding problem and they're in this
351. 1:00:47competition and three years in they basically realize that they can just predict them all and Demis walks into a room and they're like
352. 1:01:00well we're trying to figure out should we like packaging package them up in groups of 20 and we'll sell them as a
353. 1:01:06package and we'll this and that and Demis is like wait a minute like we can just do this now and they're like yeah
354. 1:01:13the the tool now just works and it can predict all these protein folding things and Demis immediately on the spot says
355. 1:01:23give it away if it works predict them all and just give the data set away make it open source um this is very much got that spirit. So
356. 1:01:35um kudos to him because I that is probably a big uh inspiration for how they're set how how they're rolling this
357. 1:01:43out. Um assuming this is all true. Um 1.8 billion committed turning biology into something AI can increasingly
358. 1:01:52model, predict, and engineer. All right, there you go. Demis. All right, there you have it. That was the one I wanted to show. Okay, so I
359. 1:02:08think what I want to do now, I want to go We've got half an hour left here. I want to go play in chat GPT
360. 1:02:22because I just haven't been in here in a while and I want to go look around. I'm gonna go to chat. So, in chat GPT, I'm
361. 1:02:30in the web version. I'm not in the app. I'm in the web version. Um, there's a work tab. There's a chat tab.
362. 1:02:42Here's Kyle Brain is actually chat GPT dots. And if you haven't used this, it's it's kind of like chat except
363. 1:02:53he's available on your phone and you can ask him things and he can go do things in other projects and he can do things
364. 1:02:59in chat GPT work. Um, but I wanted to go down here. So we got projects, we've got images, we know those. We've got health.
365. 1:03:12Let's go to health. Connect your health data. Connect health. Oh, interesting. Denver Health Common Spirit.
366. 1:03:28That's interesting. So, it knows I'm in Colorado. Here's all the Here's all the systems. Wow. All right. That's kind of cool.
367. 1:03:41Um, again, these companies are all asking you, you know, they used to just want your your browsing history and and
368. 1:03:48who are your friends and now they want everything. So, most people are probably going to give it to them. Finances.
369. 1:03:56So, I can connect finances. It's going to show me all the banks. I'm I'm I'm sureformational purposes only. Yeah, sure. Connect to
370. 1:04:06Plaid. Okay. Credit score. Manual accounts, right? Fascinating. Sites. Okay. I have not played with
371. 1:04:20sites before. Search sites. Create a site. All right. Create a new site with sites. Start by asking me what it should be about.
372. 1:04:34Create a new site with sites. A 70s asteroid clone with good physics
373. 1:04:50game play vector graphics.
374. 1:05:05Oh my god, I can't type. uh vector graphics uh SFX scoring and uh bus for
375. 1:05:25burn in CRT cathode ray tube. All right.
376. 1:05:41All right. So, that's going to go off and do its thing. Can we see? Let's see. I'll use the site skill to build a playable game. Thinking.
377. 1:05:49All right. Let's go keep looking at other stuff here. Maps. Has anyone played with maps in chat GPT? Did you
378. 1:05:56know that maps was there? Look, it knows where I am. Culinary dropout. Use command plus scroll to zoom the map. Okay.
379. 1:06:11Um, let's see. Um, find hot fishing spots near downtown.
380. 1:06:35Oh focus. Here's my picks. Searching the wall.
381. 1:07:12Oh, there we go. All right. That's kind of cool.
382. 1:07:26I don't know that this is I mean, it's way slower than Google Maps. I I I feel like all of these additional
383. 1:07:39things that they've added are all and there's GPTs. By the way, GPTs are going away as of December 11th,
384. 1:07:50but you can convert them into a plugin. Right? See right here where it says migrate to plugin? You can just do that.
385. 1:08:03Love that prompt. I want to play it now. We'll go back to that. Let's go back to sites. Something went wrong. Retry.
386. 1:08:38Oh, it's still going. Okay. I ran three commands and used sites. The visual direction is a green phosphor ara arcade
387. 1:08:45cabinet, glowing vector lines, scan lines, fading trails, a little screen burnin, and building the game around inertia, screen wrapping,
388. 1:08:56splitting asteroids, flying. sauce is increasingly tough. Oh, that's a good point, Brandon. So, what Brandon said about maps when I
389. 1:09:07was talking about maps being kind of slow and jankier than Google maps is that maps, all of these things are likely not for us to use.
390. 1:09:20All of these are tools that dot in this case dots are going to use. Good god. So, the problem with right now
391. 1:09:36is that if you're trying to pay attention to this stuff, looking at the current state of like chat GBT in particular
392. 1:09:50is kind of a waste of time because it's unusable right now as a human being, but you're not going to need to
393. 1:09:58use this. Your agent's going to do it. Tik Tok fan. Eric, uh, Nemanin, um, my doc got stuck on my screen saver.
394. 1:10:07Had to disable the screen saver. That's hilarious. No maps in a plus account. Oh, that's interesting. So, this is only I'm in a pro account.
395. 1:10:22I'm in the cheapest pro account I can be in, but I am in a pro account. Can I go look at these while it's editing? I can't. That's annoying
396. 1:10:34outputs. Huh? They're trying to bury all this stuff. They used to be able to look at all this stuff. You can't do that anymore.
397. 1:10:48All right, we'll wait for that to do that. >> [clears throat] >> Speaking of chatting about sites, Lovable rolled out this week a free chat
398. 1:11:01that works across all of your Lovable projects. So, you can have a super agent that can look into and reference across
399. 1:11:10the different projects and you can chat with it for free. >> Yep. Okay, that's good. That's smart. S smart on their part. This is where all
400. 1:11:18of the companies are going. They're all going to some master orchestrator agent that will talk across things. Now,
401. 1:11:28here's the real play. You want to make a trillion dollars? Build this. Build an agent that can chat across all of your agents
402. 1:11:39and all of your chats. I don't think there's one out there yet, but like an agent that can maintain your context locally
403. 1:11:54so that you own your data, but it can go use all these other things and pull from your data selectively as
404. 1:12:01needed so that you don't have to do what I'm doing right now, which is I get some shitty messed up thing in my email that
405. 1:12:08is the result of some agent I set up a month ago and I can't remember what service I set it up in. Someone needs to
406. 1:12:14solve that. There needs to be a master agent that sits on top of all this. Now, everyone is going to try to build the agent
407. 1:12:24that is the master agent because if someone can figure that out, then they win. They own everything. They own all
408. 1:12:33of what you do. Um, so that's why it's not going to be an obvious play because everyone wants to own it. So, they're
409. 1:12:40all trying to take all. Winner take all. Good lord. All right, we'll come back to that. Let's go. Let's just go look at
410. 1:12:49Twitter while that thing's working.
411. 1:13:04Anthropic study finds that AI and robots together are already capable of performing tasks covering 81% of US employment. Wow.
412. 1:13:17What if every possible version of you exists? This guy's funny. He's a musician. AI is actively math. That idea of I hate
413. 1:13:27it when people say that Killing math. This is dead. It's not dead. It just accelerated and expanded.
414. 1:13:43We're rolling out monthly cloud platform API credits for Max and team plans. Oh, nice.
415. 1:13:58Figma Figma agent is now out of beta. Okay. What's this? Grockbot. Ah, Grockbot can now search, read, and monitor X. Cool.
416. 1:14:14Let's go. Grock. Grockbot. Let's Let's launch Grockbot. All right, there's radar. So, um, what
417. 1:14:29were the last five things I posted on X direct radar go radar go? If you haven't seen Grockbot, this is it. If you're
418. 1:14:42like, that looks like all the other chatbots, it is. They're all the same except they're all different. chatbt is awful with photos.
419. 1:14:57Hey Becky, what is the best bot? See that? Here's the problem, Eric. I couldn't get your your X connection
420. 1:15:10only gives me your home feed right now, meaning other people's posts. All right. Well, apparently I don't have access to the
421. 1:15:20latest. I wonder if there's an update here. Restart to update. Ah, hang on. Okay.
422. 1:15:35Connected apps. Check for updates. Okay. Um, check X now. Checking X again. Now working in X. Working in a connected app. Working in
423. 1:15:50X. Yep. Those are my last five posts.
424. 1:16:14Huh. All right, that's cool. Do we have Do we have a game yet? We don't. The physics and collision system
425. 1:16:24is in place, including shots that cross screen edges, asteroid fragments. The game now has laser and explosion sounds,
426. 1:16:33a heartbeat that speeds up. The game play checks passed, including a twominute simulation and checks for wrapping collisions, scoring
427. 1:16:43respawns, and pause and resume the canvas. Ah, here we go. There it is. Boom.
428. 1:16:59Continue chatbt. Oh, you got to run, huh?
429. 1:17:14All right, there we go. There's our asteroids. One player. Uh, I've seen
430. 1:17:31this isn't very good. I mean, it's Listen, it's not bad.
431. 1:18:04The philosopher screen burning is kind of cool. I mean, this is an absolutely playable version of this game. It's not very
432. 1:18:14authentic. I didn't ask it to be authentic, I guess. Tik Tok pin. I don't see the pin there.
433. 1:18:41Yay. What's backup? Uh oh. Tik Tok. Ah, Becky, have been
434. 1:18:55asking in my chat GBT how to give me prompts to learn how to use it again. Oh, that's actually really
435. 1:19:02smart, Becky. Um, all right. So, there's there's a website that um Chat GBT just built. I think that so Becky Rue on Tik Tok
436. 1:19:17just said she's using Chat GBT to teach her how to use Chat GBT. the you took about a year year and a half off something like that. Becky,
437. 1:19:28it has gotten all of these things have gotten so complicated in that time. Have you given your bot a phone line so they
438. 1:19:34can call other bots? Not yet. Tik Tok question. Um I think Muse can make phone calls for you. So Muse right now people are using
439. 1:19:44Mo Muse to like call up the phone company sit on hold for them and then negotiate better bills for you know better better uh subscriptions.
440. 1:19:56Um so I know that some of these things can use the phone. I am [cough] excuse me I am barely scratching the
441. 1:20:08surface on how to use these things. Um, my my voice is going, so I should probably get on out of here. Um, hang on a sec.
442. 1:20:23Too much talking today. Does Anthropic have one? No. Like a like a simple bot? Not yet. Okay.
443. 1:20:43tomorrow. Um 10:00 a.m. we have office hours. I mean 10:00 a.m. Wait, no, no, we don't have the mastermind tomorrow. We don't
444. 1:20:54have the mastermind tomorrow. We've got the the all hands meeting. It's an internal meeting. Um that's in place of
445. 1:21:00the mastermind. We're taking a week off of mastermind. So mastermind will pick up again a week from tomorrow. And then
446. 1:21:06Friday, we've got office hours at 11 Mountain and Friday night date night here at at 7:30. Yeah. Brandon, did you have something to say?
447. 1:21:16>> Yeah, I you got it. >> Okay, cool. Um, Groovy, um, give the agent a Google phone number. You can probably do that. Um, I don't think
448. 1:21:25Anthropic has a one of these agents yet. Um, working locally like COD Codeex um would then have access to the same context.
449. 1:21:38That's true, Frosty. Except there's all sorts of subtle nuance to how you use
450. 1:21:59the agentic pieces of these platforms that access your local hard drive. If you've got more than one computer, so
451. 1:22:08like if I've got this laptop and I've got my work computer and I've got my dedicated AI computer, I can virtual machine I can tunnel into
452. 1:22:20from my laptop or my work computer into my AI machine, but then I'm effectively using this machine. If I if I download
453. 1:22:29desktop apps to these two machines, then it's writing local files on these two machines, but not on this one. So,
454. 1:22:37coordinating between these three is actually a nightmare. And then on top of that, on each of these machines, you can
455. 1:22:45use the browserbased version of the agent that doesn't have access to your hard drive. So, all of that context is living up in the cloud.
456. 1:22:57And then you've got this DOT agent sitting on top of all of that. So the dot agent in theory knows all of where
457. 1:23:05you've chatted everywhere and can go find all that. But it the there's no easy way to know where your is being saved. It's a nightmare.
458. 1:23:18It's an absolute nightmare. Um, October 17th, there is an LOL. So, mark this on your calendar. 10 days out on a
459. 1:23:32Saturday is the next uh learn out loud session. All right, beautiful. Um, I'm going to get out of here. My voice is
460. 1:23:39crackly. Chef Kelly forgot to show you this promo video Claude Code made for me. Oh, cool. Let's go look at this. Can I
461. 1:23:48copy this? I think I can. Can I copy this? No.
462. 1:24:06Can I copy this? Copy. No. Oh. All right. Well, what? Whats? Herod. Herod D, what's happening, man? We got a couple of couple old-timers in
463. 1:24:20the house. How you doing? I got to get out of here. My voice is killing me. Um, it's all dry. Hope you hope you all had a good time
464. 1:24:29tonight. Um, if you're new here to the salon, welcome. Check out the different neighborhoods. Check out the events.
465. 1:24:36There's lots and lots of events events going on. If you're in the AI learning lab and you're thinking like, "Hey, I
466. 1:24:42came here to learn, like, where's my stepby-step tutorial?" That's not really what this is. This neighborhood, this space is, think of
467. 1:24:50this more like your neighborhood bar where there's a bunch of people that know each other, other people come in
468. 1:24:54and out, things like that. We hang out, we talk about AI, we philosophize, sometimes we rant, sometimes we get loaded.
469. 1:25:03Um, that's the personality of the of the AI learning lab. There are lots of different neighborhoods in the AI salon
470. 1:25:10that if you're a beginner, there's there's a neighborhood for that. If you're if you're wanting to have deeper
471. 1:25:15conversations, check out the mastermind. There's people in the mastermind having deeper conversations. So, there's lots
472. 1:25:21and lots of neighborhoods uh for you to find your people. So, I hope you had a good time tonight and uh I will see you
473. 1:25:28on Friday. All right. Peace out, everyone. Have a good night.