3/24/2026 - Why OpenAI Just Killed Sora and What It Signals About Their Next Big Move

Transcript

1. 0:05Uh, you know what? Donald Trump sounds like he's what?
2. 0:33Hm. Yo, what's happening good people? How is everybody? Coming at you live from Orlando. Here, let's do this. We will go.
3. 0:47And we will go live from Mickey's house. Orlando. FL.
4. 1:03Save. Boom. Look at that. Look at that. Look at me all in here with my my soft pastelly kind of blues. Very 2015.
5. 1:19What's happening good people? What is going down? I'll tell you what's going down. Sora is going down. Sora is going down. Down, down, down,
6. 1:33down. Um, let's see. Spin B3's in the house. What's happening? Can I do something about my shirt? Um.
7. 1:50I could change it. Let me see. Hang on.
8. 2:06Hey Kyle. Yeah. Just so you know, you are still on camera on your Tik Tok. How? Oh, cuz of my mirror? Yeah. No No good a free show tonight.
9. 2:25Free show is for everyone. Kyle, you can get banned on uh On all the platforms. All right.
10. 2:41Okay, let's see. It's been pretty thought Apple Watch was the shock collar. Oh, that's funny. What's happening, good people?
11. 2:57What is happening? Um What are we going to talk about tonight? So, wild news is that um OpenAI is killing
12. 3:14Dang you, Brandon. Oh, they They want to see me strip. It's getting saucy in this hotel room. You have no idea. You have no idea.
13. 3:29It is It is debauchery central. Um I deleted my Tik Tok. Well, this is why we're doing We're doing both the Tik Tok and the um
14. 3:44and the YouTube. Hang on. It looks like one of my cameras have got schmutz on them. Let me de-schmutzify. on. Hold Hold, please.
15. 3:56Everybody calm down. Everybody calm down. Hey, you want to see something funny? Sure, Kyle. What do you find hilarious tonight? So,
16. 4:07I get into my hotel or so So, I'm at a I'm at some big-ass resort in Orlando. And uh What's happening? Um
17. 4:26Okay, I'm just cracking up now. So, I get into my room, and you know what's in hotel rooms, right? It's usually a Bible.
18. 4:32So, I walk into into my hotel room this time, and I got this book, Stop Overthinking. I got Practicing Mindfulness.
19. 4:45Apparently, they got wind of the AI Salon Mastermind Practice Lab, and preceded my little my little office away from home, and then I got
20. 4:56Atomic Habits. Yeah, exactly. And he sent the advanced team. I don't know how all these I don't know how all these books got in my office in
21. 5:07my in my hotel room, but they did. Let's see. Oh, for your enjoyment. If you wish wish to purchase, $10 a book. I'll tell you what.
22. 5:20That's some That's some actually clever marketing. You know what? This is This is a marketing ploy. I thought it was just I actually thought
23. 5:27someone probably just left them here. It's a marketing ploy. You can read them, and you can buy them. Yeah, who knew? Wild.
24. 5:38Big Brother is always listening. Do you have AI glasses yet? Yes, these these are uh my eyeballs. It goes into this large
25. 5:47language model up here. Quickly view books. Atomic Habits is interesting. I listened to it on Audible.
26. 5:54Not going to lie, kind of bummed about Sora. I know, same same Daniel. So, couple of things about Sora. So, here's
27. 6:01here's what I know so far. It was the Sora team. It's great marketing, isn't it? It's really smart marketing. And just a little bookmark,
28. 6:10the little bookmark is the only thing that sort of clues you in. So, if you're illiterate like me, and I'm not really a
29. 6:16reader. I I and to say not really a reader is doing a disservice to non-readers. I don't read. I've read one book in the last
30. 6:26I don't know, 5 years. I shouldn't admit that. And it was one that was assigned to me. It was a good book.
31. 6:34I'm a professional now because of it. But anyway, um That was probably dangerous. You know, an eyeball cut, a paper cut on your
32. 6:44eyeball? Probably not pleasant. Anyway, that's the only piece of marketing here. Okay, Sora. So, if you don't know, Sora was OpenAI's
33. 6:55um video model. And um when it was when it was announced, it was by far the best video model in the world. And
34. 7:08then they released Sora 1, which was like a hobbled version of what they showed. And then it took them, I don't know, 2
35. 7:16and 1/2 years to get to Sora 2, or a year and a half to get to Sora 2. And then they launched it to great
36. 7:22fanfare, and they launched it with this app, the Sora app, which is like a little social network where you can make fake videos of like
37. 7:31you and Sam Altman. I made a bunch of those. And um And so, they announced today that it's going away. So,
38. 7:42an X post that I saw this evening, which if it's on X, you know it's true, right? Nothing on there's um false. So, cuz it's it's
39. 7:51maximally seeking truth. That's what Elon Musk told us, so that must be true. So, um wait a second, you're early. I am an
40. 8:00hour early. I decided to go an hour early. I'm on the East Coast. I figured for the East Coasters, I'll give you a
41. 8:06little gift. I'll be a little less ridiculous for me. You know, like it's always ridiculous for us. I know. I know.
42. 8:13For those of you on the East Coast, I'm sorry. I'm a dick. What can I say? When I decided to go live at 8:00 at night, uh
43. 8:24I didn't have a sense that I was in a country with other people in it. So, a news report about Sora shutting
44. 8:32down, yeah. They want to concentrate on real world things. Well, so a couple of things. Apparently, it was costing OpenAI 10 to
45. 8:4315 million dollars a day to run Sora, which is in the billions for a year. I don't know, someone can do the math on
46. 8:51that, but that that can't be pleasant. Well, three 3.6 billion if it's 10 10 million a day. Jesus. The [ __ ] numbers with AI are
47. 9:04staggering. Um so, that was their little side experiment. Sora was costing them 3.6 billion dollars a year or whatever.
48. 9:13Um so, they're shutting it down. They need the the uh the GPUs for training and for inference. Apparently,
49. 9:22they have trained a new model that is supposed to be quite good. Yeah, and I'm going to talk about that, too, Brandon.
50. 9:28They've trained a new model that's going to be quite good. Um as as we would assume. Now, to be clear, OpenAI completely lost the plot with
51. 9:40GPT-5. I think the GPT-5 series sucks. Like it's probably fine for coding. I think GPT 5.4 with Codex um engineers seem to be really pleased
52. 9:50with it. Anyone that I know that likes words hates it. So if you're a fan of the words, the GPT-5 series was an abject disaster.
53. 10:05Maybe GPT-6 will be better. Who knows? Apparently what they're going to focus on is creating an Uber GPT experience that includes chat,
54. 10:19Codex, the programming tool, um and a browser, the Atlas browser. So so they're apparently creating some sort of Uber kind of tool.
55. 10:30And then they also have Peter Steinberger working for them. He's the one that created Open Claw, which Open Claw, if you don't know,
56. 10:38that's the little happy agents that do all the work for you when you're not when you're sleeping, unless you're me, and then
57. 10:49I'm not going to talk about Adam. But but but apparently other people digging Open Claw. Um so I would think that Codex and Open
58. 10:59Claw, like this new product, probably has agents, web browser, a programming interface, and then all the chat GPT stuff we know and love. Um
59. 11:12they're also Sam Altman also on April I forget what date it is is making an announcement about um world.org. So if you don't know
60. 11:25world.org, they're the ones that created world.id, which is the human verified blockchain-based thing that's got a
61. 11:34crypto coin associated with it. Um if you're if you're using LinkedIn or if you're using X right now, I called this one when
62. 11:44when did I get my world ID? I got it in 2023, I think, in August of 2023. It might have been 24, I don't know. A while ago.
63. 11:53I got it a while ago and what I said was, I'm I'm going to get this just to get it, just to understand the
64. 11:58experience and then most people yelled at me, "Don't give them your Don't give them your biometric data, Kyle. That's really not very wise."
65. 12:09Um It was before you written the paper cut tonight. Um So, I got it a while ago and what I said was, there's going to be a time
66. 12:25where the AI robots are going to get so prolific that social networks are going to be unusable. And then there's going to be a demand, a
67. 12:36cry for human verified stuff. So, a possibility is that OpenAI launches a humans-only social network, which would be kind of smart.
68. 12:52You have a receipt. I do have a receipt, as a matter of fact. Hang on. For those of you on TikTok, I'm going to go find out my world ID uh
69. 13:00time. July 2023.
70. 13:16That was a while ago. Damn it. I must be psychic. July 2023. July 2023 is when we uh when we got ye old world ID. Um so
71. 13:31apparently there's a world ID thing coming. There's a new model from Open AI coming. There's some sort of unified something coming.
72. 13:42They now employ the guy that wrote Open Claw. So, it it looks like what they're doing with Open Claw is they're just
73. 13:49building a bunch of hooks into Open Claw. Open Claw's remaining open source. And then the guy that wrote it is
74. 13:55working with all the hook developers within within Open AI to connect [ __ ] So, So, a couple of things. I think this is the year where
75. 14:10social media's going to become [ __ ] useless. Like X X at this point is useless. Like it's just useless. It's And LinkedIn?
76. 14:23Everyone's commenting with bots. Wait, like there's no you know, TikTok pin. Pika has another new app called pika.me
77. 14:32like virtual you. It keeps iMessaging. That's interesting. Oh, keeps iMessaging you? I would be scared of an AI me. I'm
78. 14:42pretty sure that's how Skynet happens. Other news, Hachette cancels book Shy Girl because it was rumored to have been AI written.
79. 15:02I'm just [ __ ] exhausted. Did you see all the people marching to to Open AI headquarters? Well, wasn't that many of them. It was like
80. 15:12fif- 100 people. 150 people, I don't know. Some amount of people. They had all their signs, "Stop AI. Slow down AI." AI's not slowing down.
81. 15:23AI's not slowing down. Writers are going to use AI. And you know what's going to happen? Lazy writers are going to use AI lazily
82. 15:38and they're going to create crap. And you know who's going to read the crap? No one. No, that's not true. Agents.
83. 15:46Agents are going to read the crap. No one's going to read the crap. Artists writers, conceivers
84. 16:00creators, filmmakers poets They're going to find ways to use these tools that are novel and interesting and exciting.
85. 16:14And everyone, every single person that I know that's using AI to any quality degree
86. 16:29is not just pushing a button and out squirts a novel. Yeah, Windrider. I personally don't care who writes a book so long as it's good.
87. 16:43Or or more accurately, I don't care what tool was used to write the book so long as it's good. A piece of work either moves you or it
88. 16:51doesn't. The person that created it is either willing to put their name on it or they're not. They used the tool
89. 16:59who gives a [ __ ] Yeah, it's not really your writing. I'm exhausted. Okay, anyway, back to Sora.
90. 17:13Um Jensen Huang yesterday. Thank you, Wolf Van Clinton, on TikTok, for the gifts. Um
91. 17:29casually said in an interview, I think we've hit AGI. Dario Amodei from Anthropic, in the past month, has been weird. He's already
92. 17:38weird, but he's been weirder. And he's saying things like, "You know that article I wrote on the age of
93. 17:45spiritual machines? We're pretty close to that." That That was him describing AGI. So, I think Anthropic's probably got it. Um
94. 18:00I don't know if OpenAI's got it. Their Their GPT-5 models suck, but they're powerful. Um So, I would think that they've got it. Elon Musk
95. 18:14claims to have it or claims to be close to it. So, you know, it looks like that stuff's coming. And And when you look at something like
96. 18:29OpenAI's going to take this thing that they publicly, for 2 years, said this Sora video thing's coming. We're going to launch a
97. 18:39social app. We're going to take over Hollywood. We're going to do all this stuff. They cut a deal with Disney 3 months
98. 18:47ago, 2 months ago, 3 months ago. I don't know, some months ago. Where Disney This is This is remarkable. This
99. 18:56happened in the early days of the internet, too. Normally, when a company like Disney that's got valuable properties cuts a
100. 19:03deal with someone like OpenAI, they charge OpenAI for the right to use their characters. Disney said, "Hey Sam,
101. 19:15we'll pay you a billion dollars if you can stick our little characters in your little movie-making thing there."
102. 19:23And Sam said, "Okay, give me that check." Well, Disney's backing out of that deal cuz Sora is going away. So,
103. 19:35if you've got a billion dollars in investment from Disney for this tool that, you know, everyone could have made their own Disney film,
104. 19:46you know, in however many a year, year and a half, and you flush that down the toilet, that says to me that something big's coming. Um
105. 20:00the other thing that's going on with OpenAI, it looks like they're they're Sam said he's going to his focus is going to be on fundraising.
106. 20:08So, it looks like he's out seeking another 10 billion dollars at a 750 billion-dollar valuation. So, they're going to be a
107. 20:18trillion-dollar company here here shortly. Just just [ __ ] mind-boggling. So, apparently Sam's going to be focusing on um
108. 20:31on fundraising and securing uh data centers energy, energy for data centers. So, what's that mean?
109. 20:56I'm a little punchy. I don't know if I didn't sleep enough or what it was. What do you think is their interface to each AGI more texting?
110. 21:09No. Well, so you know what's funny Mike Diamond's the the um one of the one of the tweets that I saw said that the Sora team is going to be
111. 21:21focused on world models. So um Runway ML is is is working on world models. Midjourney is working on world models.
112. 21:36I would be dramatically surprised if what OpenAI learned with Sora isn't being flipped into a world model. So so that that kind of
113. 21:47rings true that that they would be going there. um No, I I I don't think text is going to be the primary interface. I think it's
114. 21:58going to be world models. I think it's going to be We're going to kind of if if you're old if if you're on this channel you're
115. 22:10probably old. And if you're old you probably remember Prodigy. Remember Prodigy? Prodigy was Sears's AOL.
116. 22:19Was it Sears? I think it was Sears. I think Sears owned Prodigy, right? Am I Am I right? Am I or am I not remembering
117. 22:26that right? I might I might have that wrong. But Prodigy was this really janky ass um dial-up you know, chat tool. Was it Sears?
118. 22:41Co-owner, but yes, it was Sears. Okay, good. I I'm like I'm like am I that old that I that I'm am I remembering that wrong? But the the
119. 22:51co-owner of Prodigy was Sears. And if you're not old you're like, "What's Sears?" Sears was a store where people that looked like me sold
120. 23:01you carpets and clothing. I remember Prodigy. And Prodigy had this really janky ass it was like bad it it was like the font was really
121. 23:15jagged and Tik Tok build their world model. Yeah, they are. They are. Um and it was just it was it was just a
122. 23:27horrible horrible service. It was slow, it was buggy, it was janky. And then AOL came along and then the web
123. 23:36came along and everything sort of evolved out of the out of that proprietary world. CompuServe was another one that was really janky.
124. 23:44This era, these last 3 years, this era of chat GPT with text as the primary interface is going to feel like that.
125. 23:51We're going to look back on this time and go, "Hey, remember when you used to have to type in?" And they were all wondering if it was AGI.
126. 24:00But nobody knew and they kept going, "Do we have AGI yet?" I don't know. And then they kept trying to figure out what's AGI anyway.
127. 24:07And then Katie Couric and Bryant Gumbel were on TV going like, "What is AI? What is internet? That A with the circle around it."
128. 24:21We're We're living through history again. Um but I think this year is the year that we see some credible decent world models.
129. 24:35Probably toward the end of the year. And then I think you're going to start to see I think you're listing all the ISPs of my late father's
130. 24:46friends. That was That was what we did back then. Government commercial break. The US DOL is launching a new AI literacy course
131. 24:58designed to ensure that every American worker can benefit from the opportunities that AI economy presents. Hmm.
132. 25:07To get started, simply text ready to 20202. Is that a joke or is that real? Very real.
133. 25:21We should We should call Wait, what is it? We text ready to 20202? I'm going to text it. Tell TikTok I love them, but I'm I'm
134. 25:31going to text something. Text ready. Okay, two 20202 ready. I wonder if they know which AI they're talking. Thanks for
135. 25:46enrolling in AI 101. Look out for your first lesson soon. Okay. Wait, let me go back to TikTok. Sorry, TikTok. I had to uh
136. 25:54I had to text the government that I'm ready to learn AI. Hey Hey, Source Camp, you don't need to teach anyone anymore. The government's
137. 26:03got it handled. Oh, here you are. Couldn't figure out who you were talking to. Hello.
138. 26:19Oh my god. Oh my god. Oh my god. It's just so crazy. It's so crazy. You know, it's crazy. It's just crazy.
139. 26:29It's crazy. Hey, TikTok. It's crazy. Oh, wait. I got a text. Hold TikTok, please. Okay. Let's see. AI make
140. 26:47Okay, we we got to download this graphic. Save. Everybody's got to see this graphic.
141. 27:07Listen, I should not I should not be making fun of our government. I really shouldn't, but I'm going to.
142. 27:20Cuz someone made this graphic and I feel bad for them. I'm about to mock them. Oh, it's so bad. Especially on TikTok. I
143. 27:30got to go back to TikTok, sorry. Okay, I'm back. Calm down. Just Calm down. I changed my shirt for you, okay?
144. 27:38So, if you if you get some pauses tonight while I go get my AI 101 classes from the government. Okay.
145. 27:53Am I sharing? I think I'm sharing, right? There we go. There we go. Make America AI ready. This God, does this feel like 1997?
146. 28:03This This feels like 1997. Like these kind of these kind of graphics, you know, learn the world wide web.
147. 28:18No music on hold? That's better than Jeopardy calling. Make America AI ready. All right. Well, so wait. I'm going to go back to uh
148. 28:29Wait, I can't see my screen anymore. Okay. I'm going to stop sharing there. I'm going to go back and look at my texts.
149. 28:37Hold TikTok. I'll be back. AI is changing how we work and live. Oh, you know what I can do? I know what I can do. Hang on.
150. 28:58Oh, I didn't I don't have it here. Oh, that's a drag. I should send my resume to the US government.
151. 29:14That's a t-shirt. Can't wait to see how much they get wrong. I know. Hang on. Seven 7 days 10 minutes. Learn what AI
152. 29:22is, how it works. This is your AI 101 course. It's part of the Make America Ready initiative, Department of Labor
153. 29:29in connect collaboration with education technology partner Artist. What?
154. 29:57Oh, you put you put it there? Okay. Great. Let me Can I go grab that? Yes. This is worth Can you grab the whole
155. 30:06thing, Brandon? This isn't all of it, is it? No, it's not all of it. Go grab the whole thing. And I'll stick it in a in a post it in a
156. 30:14in a notes. Oh, I might be able to do it here. Can I do it here? Hang on. Copy. I got it. Never mind. Okay.
157. 30:28Let's go. Notes. Okay. Here we go.
158. 30:45Um TikTok, you'll just have to listen along. And then are we sharing? No, cuz I got to share. Wait, stop there.
159. 30:58Start sharing there. Start doing this. That to that. That to that. Okay. Lesson one, what is AI? AI is already
160. 31:09working for you. In fact, you probably used AI before you finished your morning coffee today and just didn't know it.
161. 31:15Google Maps quietly dodging the nightmare in traffic jam? AI. Netflix guessing the next series to binge? AI. See, this is all
162. 31:33So, how does it work? It's a system that looks at massive amounts of data, finds patterns, and makes predictions. That's
163. 31:39it. That's the real secret. Patterns in, predictions out, but it doesn't stop there. Oh, okay. There's a newer kind of
164. 31:45AI called generative AI that doesn't just predict, it creates. That's actually that's good. Give it a short instructions, aka a
165. 31:53prompt, and it can draft an email explain confusing for This is This is not bad. Think less sci-fi plotting world
166. 31:59domination and more really smart digital assistant that works at 2:00 a.m., never judges you for asking the same question
167. 32:05three time, and never asked, "Did you Google it first?" When people say I AI today, they're usually talking about generative AI. In
168. 32:13this course, we'll use AI broadly, but our main focus will be generative AI. That's at least competent.
169. 32:22Before we continue, let's hear from you. How confident do you feel about your understanding of it? Respond with a zero to 10. Wild.
170. 32:33Wild, wild, wild. Seeing this makes me feel pretty secure about my teaching gig. Yeah, it's it's okay, Kelly. We
171. 32:39don't We don't need your expertise anymore. The government has it. Make America What was it? AI AI ready. Hey, AI readiness.
172. 32:55They stole my [ __ ] Damn it. Make America AI ready. Do Do I get some sort of compensation for that? We've had AI readiness in the AI Salon
173. 33:11for years now. Okay. So, basically, we are the paid to the government class. I guess so. Oh my god. Oh my god.
174. 33:25Um So, the other thing I thought might be fun is to just surf through the old Twitter.
175. 33:41Since I can't really build anything. I mean, I could, but By the way, my um I'm down here in Orlando on Storyvine
176. 33:57business. And then Friday, I fly out of here and I go to Boise, Idaho. And I'm speaking at the um It's called
177. 34:09um Oh my god, my brain. I'm so tired. It's called Hackfort. There's this There's this sort of Boise version of South by Southwest
178. 34:19called Treefort. And then they created a a sort of technology arm of it called Hackfort. And I'm speaking about AI
179. 34:27to the artistic community in Boise on Saturday. So, that's pretty cool. I got invited there by this woman named
180. 34:36Kathleen Cohen. She's one of the women that is hopefully um helping to get Sydney, my musical, up and running, mounted, produced.
181. 34:46So, that's something we're working on. So, that's pretty cool. Um So, that's this weekend. Uh all right, let's see what's going on
182. 34:54here. You all can see my Well, wait. TikTok can't. So, TikTok, I'll just tell you what I'm seeing. And if you have
183. 35:01questions, just pop them in there. And then Brandon, if you've got stuff you want me to look at, go to head and go ahead and do that.
184. 35:12Volunteers, CEO says there's only two kind of people that will succeed. Trade workers or your neurodivergent. Neuro spicy for the win, baby.
185. 35:23There you go. All right. So, so basically AI's going to completely wipe out the middle class. I've actually I've actually noticed some some
186. 35:34things going on in Twitter that I think is bots. They'll screw it up and make it harder and more expensive than it has to be.
187. 35:43Sora say goodbye to Sora. Everyone is wrong about why Sora is canceled. Sora 2 is the last gen tech. C dance 2 is 10 times better.
188. 35:54And it will be rolled out in the US within the next few days. Trained on TikTok has much more video data than
189. 35:59anyone else on Earth. Open AI can't compete. We'll see. Too many AI generated replies, too few human ones. We will not accept the
190. 36:09unacceptable. My speech at the stop AI race. The fetus slop.
191. 36:31>> So this is kind of interesting. It's like one of the things I just want to make sure I'm not in a post.
192. 36:37One of the ways I look at signal, how I separate signal from noise, and again, I don't think this is going to last that much longer,
193. 36:43is if I see the same story show up over and over and over again in my feed, then that generally means it resonated.
194. 36:50Everyone on X is acting like Sora's shut down is the end of AI film making. It's not. OpenAI's going public. Oh, that was the
195. 37:01other thing I was going to talk about. What OpenAI is likely doing is they're likely going public. So,
196. 37:10OpenAI's going public. That's the entire story. When you're preparing for an IPO, you go through every product and you ask
197. 37:17whether it makes you look like a serious company or whether you look like you're running a deepfake factory for Disney fan edits. What?
198. 37:27No. No. Oh good lord. All right, whatever. I'm not going to read that garbage. Marcus Pittman. All right, CERN scientists transported
199. 37:41antiprotons, a form of antimatter, by truck across their campus. If the If the opposites come into contact, they will annihilate each
200. 37:51other, setting off lots of energy depending on the mass. All right. All sorts of weird science stuff happening. Did you see Did you see
201. 38:04Here's a good one. Wait, let me go back to Let me go back to just me. Um I think this is real, but it might not
202. 38:12be. So, if there any physicists in the house, I saw a story today that they did some analysis on light. They split it, they trapped it.
203. 38:23And instead of light having two dimensions, which is what they thought, light actually has 48 dimensions.
204. 38:31Anybody Can anyone confirm or deny that or Google it or chat GPT it? I'm glad you asked about the 48 dimensions of light, Kyle.
205. 38:45Son of a [ __ ] 48 my way sides, I love it. 48.5. Oh, there's 48 and a half dimensions of light. Not two.
206. 39:00Here's what I know. I know I'm not smart enough to know what that means, but I am smart enough to remember that.
207. 39:09And then the next time I'm at some party and someone says something about physics, I'll go, "You know,
208. 39:16scientists trapped light and expected to find two dimensions and they found 48.5." And then the physicist will be like,
209. 39:27"Uh-huh." And they'll start talking physicist stuff and then I will slowly back out of the conversation and go get some more cheese dip.
210. 39:39Unstable connection. Oh boy, that's not good. Oh, you just made the half part up. Okay, see, there's our This is a channel
211. 39:47for misinformation. So, not only is the 48 dimensions possibly not true, but the half was just a joke that someone on
212. 39:55TikTok gave me that I absorbed as fact. I'm the unstable part. Love your camera, so clear. Uh it's just good light. Yeah, it's
213. 40:08clear it's clear tonight. Okay, let's go let's go back to uh Let's go back to it, shall we? Uh-oh, hang on. That's going to be driving.
214. 40:28AI filmmakers of the future Ah. This is interesting. Holy [ __ ] LeCun's team just cracked
215. 40:43world models wide open. So, Yann LeCun was a chief scientist at Meta until Zuckerberg pissed him off and they
216. 40:51[ __ ] all their AI up. So, he left to start a world model company. Everyone's obsessed with the next Claude
217. 41:00update. Meanwhile, Yann LeCun quietly dropped a paper that could matter way more long-term. It's called Le World Model.
218. 41:08And to understand why it's a big deal, I'm not going to Let's go. Why is there no Hang on. We need to get the
219. 41:20the link. Here we go. Le World Model. All right. So, let's go HTML. Yeah, perfect. All right. Copy that. Oh, I've got to
220. 41:37share. Hang on. So, I just went to archive to get this this world model. Let's go to Claude. [ __ ] ChatGPT. ChatGPT sucks lately.
221. 41:55One week left for two times more usage outside the hours of 5:00 to 11:00 a.m. Pacific time. All right, whatever.
222. 42:04Okay, so there's that paper. So, uh please read this paper and tell me what it means for the future of world models.
223. 42:24Claude is now a cell phone company. No [ __ ] Actually, you know what? Tokens are the new minutes. Remember when they sold you minutes? At
224. 42:35T-Mobile, you can have unlimited minutes on weeknights between nights and weekends are free, exactly. Let me fetch that paper for you. This
225. 42:46paper is from NYU not Jan look LaCoon's group. A new approach. Okay, the problem it solves. Jeppa based world models have a
226. 42:58fundamental instability, representation collapse, which means if you're facing in this direction and then you turn
227. 43:04around and you turn back, everything's different. The model learns to map every observation point to the same point in latent space,
228. 43:18which trivially satisfies the predict next state objective, but produces useless representations. Okay, I uh
229. 43:29explain this as if I have a degree in acting, but like to act like I understand
230. 43:43science. Do not make theatrical references. I'm indicating a level of intelligence.
231. 44:08Got it. Smart person, no technical background. Here's the deal. The big picture problem. AI researchers want to build
232. 44:15systems that can imagine what happens next before they act. That way you mentally simulate catching a glass before it falls off a table.
233. 44:25These are called world models. Instead of reacting to every moment, the AI builds an internal model of how the
234. 44:31world works and plans inside its own head. The catch. You could try to imagine the future in full detail, every pixel
235. 44:41of every frame, like rendering a movie in your mind, but that's incredibly expensive and mostly wasteful.
236. 44:47Okay, the problem with the problem. When you train a system to compress what everyone else has done about it,
237. 44:54various combinations of duct tape. What this paper does, they found a single clean, mathematically grounded fix.
238. 45:02They add one penalty term that essentially says the compressed representations must stay spread out in
239. 45:08a bell curve shape across all possible directions. They test this using a classic statistics trick. Project the data under
240. 45:18random lines and check if each projection looks like a normal distribution. This is still too smart for me to [ __ ] understand.
241. 45:28There's an actual theorem, the Cramer-Wold theorem, that guarantees if all the 1D projections are bell curves,
242. 45:34then the whole high-dimensional distribution is too. No collapse possible. That's it. One prediction term, one anti-collapse collapse term,
243. 45:44one knob to turn down from seven terms and six knobs in the previous best approach. Oh, yeah. Time to go to fifth grader.
244. 46:00The model is tiny, but it means it's early. Yeah, I was going to say, okay. How long before you think this is buildable?
245. 46:15Um look at Reddit and X and tell me your prediction. So, that's pretty cool.
246. 46:31Um a lot of how we use AI today is because of Yann LeCun. He was he was I didn't he's a he's a bit of an Eeyore.
247. 46:40He's a bit of a drag to listen to. Um and he also he was really down on large language models when he was at
248. 46:47Meta. Largely because their large language models suck. Okay, Kyle. The context you need. Paper dropped 2 weeks ago.
249. 47:04Where's That's still gone. LeCun's startup AMI labs just raised 1.3 billion at a 3.5 billion valuation.
250. 47:18World labs Wow, World labs closed a billion dollars just weeks before that. Two billion dollars in 3 weeks both
251. 47:29betting against LLMs. So, this was Whoever asked the question on TikTok earlier, is text going to be the primary
252. 47:37interface for AGI? It's not. World models are. Cuz once you understand the world, you can just stick it in a robot and it can go be in the
253. 47:45world cuz it understands worlds. And we live in one. Right? Where is he tonight? I'm in Orlando, Florida. I agree. His dunking
254. 47:56on LLMs is tiring. Yeah, it's just tiring. It's just shut the [ __ ] up, you know? I'm like I know you've been in it for a while and
255. 48:03you're cynical and you hate the fact that Sam Altman was kicking your ass at meta. But, you know, lighten up. He's probably
256. 48:11lighter now that he's got this this cool world model thing. What the community is saying on X archive or alpha 9 or XIV noted that one
257. 48:24of the biggest bottlenecks of Jeppa is they're hard to train. And this new research changes that. Which is the right read.
258. 48:34Humanoids daily framed it as vindication of LeCun's long-standing argument that industry is LLM pilled and heading toward a dead end for
259. 48:43physical intelligence. The latent space AI newsletter captured the broader sentiment well. Supporters view it as an overdue next para-
260. 48:52paradigm. All right. Well, listen. Yann was the one that that built a lot of the infrastructure that
261. 49:01led to chat GPT. So, the dude knows what he's talking about. LeCun himself, we are going to have AI systems
262. 49:12that have human-like and human-level intelligence, but they're not going to be built on LLM's. And it's not going to happen next year
263. 49:19or 2 years from now. It's going to take a while. There are major conceptual breakthroughs that have to happen.
264. 49:26This paper's environment are flat 2D grids and single object 3D scenes. The real world has occlusion, deformable
265. 49:34objects, blah blah blah. The honest bottom line. The world model is buildable today as a research tool proof of concept.
266. 49:44It will be buildable as a component inside real robotic systems within about a year. So, 3 to 5 years. Okay.
267. 49:51But, here's the thing that matters for your world, Kyle. The humans amplified by AI framing, that's Claude is remembering the [ __ ]
268. 50:00I've been talking about, which is cool. Humans amplified by AI. The fact that this runs on one GPU, trains in hours, and the code fits in a
269. 50:09single file means it's not locked behind billion-dollar labs anymore. Well, it but it will be if it's a big-ass model,
270. 50:16a big-ass world, the moment world models become something a motivated individual can build and experiment with on their own hardware,
271. 50:26they become part of a toolkit for the kind of people AI Salon serves. That inflection point just happened with this paper. All right, cool.
272. 50:35Groovy. All right. That's powerful. What's going on? Questions, thoughts, what's going on over here in YouTube?
273. 50:43I'm scared. My husband is believing all the obvious AI videos, not even the good ones. This is going to be a long life.
274. 50:54I'm sorry about that. Oh, I lost I lost screen share audio. I totally would have Shakespeare's its response if you hadn't prompted like
275. 51:04that. Uh uh uh Had a friend warn me about the government offering AI education. Well, hey, at least at least they're
276. 51:17[ __ ] trying. At least the government is not sitting there going, "We don't know what AI is." Fittymint in the house, what's happening?
277. 51:34How's Adam? Adam
278. 51:50So, so there's two issues with Adam. Well, well, so okay. Adam doesn't suck too bad. So, there's this guy, Alex Finn,
279. 51:59and he's been he's been all in on Open AI for 3 months now. And he's got a YouTube channel, and you can go watch his YouTube channel. And he
280. 52:09And he just put out a video that said 3 months of Open Claw Learning in 27 minutes or whatever it was. So, it's a half-hour video.
281. 52:21And I thought, "Oh, I'll go look at all the things I haven't set up correctly." And basically, I had set Adam up mostly
282. 52:27correctly. Like, there wasn't a lot that he showed that I haven't done. I've just done it poorly. Um And so, for example, so when I watched
283. 52:42Alex Finn's video, I got some confidence. I was like, "Oh, okay. I know what to do with Adam here, and I'm
284. 52:47going to do this, I'm going to do that." And I designed a whole ass I deleted some cron jobs that were
285. 52:53corrupted. Cron jobs are just repeating, automatically running things that run on whatever schedule you want.
286. 53:01As in chronology. I think that's the the base of cron. Or cron is the base of that. Um And so, I said, "I'm going to do a daily
287. 53:13Great Repurpose signal report, and and we're going to have Adam go out, and he's going to read on Reddit, and he's
288. 53:20going to read on X, and he's going to read blogs. And he's going to go find us on a daily basis five articles that are
289. 53:29in line with the Great Repurpose." And so, I spent couple hours within 3 4 hours on Sunday designing this newsletter. And it was
290. 53:37just going to be internally for for Brandon and Andy and myself. It was just like an internal test. And so I designed this thing and it was
291. 53:47beautiful. It looked great. It had it would had nice formatting. It was really well done. The logic was done. It was well written.
292. 53:57And so I said, "Okay, great. Put it in the cron job." And like every morning at 7:30 East Coast time real time, like where I am right now,
293. 54:06real time send this newsletter. Cool. So, the next morning you're talking about your AI son like that. My AI son is sitting is sitting
294. 54:23outside the the school on the steps with drool dribbling down his chin. He's not the brightest He's not the brightest one in class.
295. 54:35So, the next morning like 7:30 comes and goes. And it's like there's no [ __ ] newsletter. And then I go look at the cron jobs and
296. 54:47there's two cron jobs in there. One of them's at 8:00 Mountain, one of them's at 7:30 New York time. It missed the New York one. I said,
297. 54:57"What's going on?" It said, "Oh, I'm about to run the one at 8:00. I couldn't run the one at 7:30 cuz whatever." It It made some [ __ ] up.
298. 55:05And then at 8:00 it said, "Oh, I've got this thing that you want. And and I ran that at 8:00 and here's the output of it." And it was a
299. 55:16paragraph with some HTML in it, unformatted HTML. And then I said, "No, you're supposed to send the the
300. 55:30the thing, the newsletter that we talked about." And so it emailed that paragraph to Brandon and Andy and me. And I was like, "No."
301. 55:44And so, I deleted the one cron job and I tried to get it to and it just every morning it goes, "I tried to email you, but I can't. I don't
302. 55:53have access." And I'm like, "No, you do have access. Try to email me right now." And he'll email me just fine.
303. 55:59And then the next morning he goes to email me with his cron job and he can't do it. So, so, I need I need someone who doesn't
304. 56:06suck to help me fix Adam. Adam sucks. Thank you for asking about my horrible digital son. Tough to market a new good idea that can
305. 56:21be so easily replicated now. Yes. So, Fiddyman, this is this is one of my This is one of my new new insights. So,
306. 56:30next Tuesday, I want you If you're in a regular, please go register for this. If you're in the mastermind.
307. 56:38So, in the mastermind, we are now adding We've We've got the practice lab, right? Which is creating a daily practice.
308. 56:45We're going to spin up this new area in the mastermind for the great repurpose. And one of the first things that I'm doing in there
309. 56:54is going to be creating your own personal mission statement using AI that can be that can address what you're talking about, Fiddyman.
310. 57:04Which is we're coming into a world we we probably are actually there, but it's it's going to take a while for it to feel real.
311. 57:17We're coming into a world that if you market the thing you're trying to sell, it will fail.
312. 57:31Period. And you're like, "What? Huh? Kyle, what are you talking about? I don't understand." If everyone can replicate everything,
313. 57:47and if everyone's going to get laid off, and the skills that they used to trade their time for money for are devalued,
314. 57:57then everyone's going to be selling something, a course, a workshop, consulting, some version of themselves. And what we've always done,
315. 58:11for the most part, is when we're selling something, we sell the thing. I'm doing a workshop. You're going to
316. 58:19learn these six things. You're going to learn them faster, better. They're going to You're going to make money because of the
317. 58:27things I'm going to teach you. The problem with marketing the thing is that everyone can replicate the
318. 58:36thing. So, if your thing is so good that it rises above the noise, which it won't, but if it does, everyone else down here will look at
319. 58:43that and go, "Oh, I'm going to sell one of those. That rose above the noise." And they will just instantly replicate you.
320. 58:50And so, you go from competing against no one to competing against everyone in a week. And no one's going to pay attention cuz
321. 58:59everyone's going to be making the same claim. I make it easier. I make it But what they are going to pay attention to, I think,
322. 59:15is the individual behind the thing. So, what I think we all need to get clear on is what is the thing that we have it as an individual
323. 59:29that no one else in the world has. Cindy [ __ ] is like this wacky futurist with this art thing and her crazy glasses and her mind that's
324. 59:43like all over the place and pulling in things from you know, her research world and from her creative world and she's she comes at
325. 59:54what she does in a very particular way. And so for example, she's got a workshop that's coming up at the end of March.
326. 1:00:05And we were talking about that and I'm realizing the workshop itself people are a lot of people are making
327. 1:00:12the same claim she is. I mean hers is a bit different cuz it's about mindset and there's some different messaging in there.
328. 1:00:19I was like, wait a minute. They're not going to buy the workshop. They're going to buy wanting to hang out with Cindy [ __ ]
329. 1:00:26and understand what she's talking about when she talks about mindset shift and what does it mean to be in this AI future?
330. 1:00:36It's about her and so the workshop that I'm giving next Tuesday which is I think 11:30 Mountain.
331. 1:00:51I think that's right. Scary if your individual story isn't compelling enough to justify a business not to copycat you.
332. 1:00:59No, but here's the thing. You're still going to you're still going to ultimately talk about the thing you're selling.
333. 1:01:05But the hook to get there is not the thing. The hook to get there is you. Who are you? Why did you make that thing?
334. 1:01:13What is it about that thing? Is there some story about someone's life changed and that made you make that thing. It's
335. 1:01:19going to be all about tying people's personal relationship to you and their trust of you to the things you're selling.
336. 1:01:29Then you can talk about the things. It's just It's the initial marketing push. If you talk about the thing, it's just going to
337. 1:01:40you're just going to be against everyone. So, it's not the workshop on being AI ready, it's Kyle Shannon bringing his
338. 1:01:48particular flavor of what he does to teach those same six bullet points that everyone's going to be teaching. Make sense?
339. 1:01:59So, that's coming up. So, if you're in the mastermind, go please go RSVP for that. I'm getting into vertical gardening
340. 1:02:06mainly because I can't squat anymore. Silver Fox is the best. That's a Sounds like a stand-up the
341. 1:02:21beginning of a stand-up act. They will pay for scaling and reliability. They'll pay for trust. I agree, they'll pay for trust.
342. 1:02:31But if But here's the thing, Uncanny Uncanny Valley, um if you say I'm going to give you scaling and reliability, everyone's going to say
343. 1:02:43that. And everyone's going to be able to deliver that. Why is yours different? Well, ours actually works. Okay.
344. 1:02:50But how do you rise above the noise? If you want to market that thing, then you you know, troll out your your
345. 1:03:00chief scientist to talk about why yours is different. Here's why he's passionate about it. He's, you know,
346. 1:03:08he's not very good in social situations, but you know what he knows? Scaling and reliability. He has one little quote from him. Uh I'm
347. 1:03:17very passionate about scaling and reliability. Right? And then you're like, oh, that guy. That guy I I would trust him for
348. 1:03:26scaling and reliability cuz he's got that personality for someone that would be into that. Right? That's what this is about. Anchor
349. 1:03:34anchoring the thing you're selling in the person or the stories attached to that person who's got some credibility.
350. 1:03:44And we all like everyone has a thing that you're most qualified to talk about and to do and to be. And so that's what this this workshop's
351. 1:03:55going to be on next Tuesday. Okay, let's go back to this thing. Let's go back to
352. 1:04:12All right, that was Jan LeCun's world model. That was a fun little dig dig down into the hole. I had Siri write its own eulogy. Oh,
353. 1:04:21that's pretty funny. Oh, you should be able to hear
354. 1:04:41The spoon too suggestive. Oh my god. That's hilarious.
355. 1:05:11I've seen that video before. Open AI should just remove restrictions on Sora for a day before discontinuing it. Satoshi never existed.
356. 1:05:21If you think Claude killed Open Claw, you clearly don't understand Open Claw. That was actually an Alec Finn post that someone just stole.
357. 1:05:31After 8 months in stealth, Hark is the new thing. The billionaire founder of the $39 billion humanoid robotics company Figure AI
358. 1:05:44announced his newest venture, an AI intelligence lab named Hark. First time >> Yes, sir. Am I not sure right?
359. 1:05:54>> unexpectedly and suddenly have 30 people over on TikTok. So, we might want to take a moment from doom scrolling and
360. 1:06:01tell them about the AI Salon. We should probably do that. So, do me a favor there, Jan Jan What's your name? Brandon.
361. 1:06:11Wait, how am I going to show this? Let's see. Pop up uh Here, I'll do it. Cuz I know what I want. Yeah, that. Is that it?
362. 1:06:26No. Do do do There we go. I have the uh link pins to TikTok. Okay, cool. So, yeah. So, if if you're new here on
363. 1:06:41TikTok, and actually if you're new here on uh YouTube and LinkedIn and Twitter, um 2026 is about to get really weird.
364. 1:06:57Jobs are going to get [ __ ] up. People are going to lose jobs. People that keep jobs are going to find that
365. 1:07:04those jobs are just It's normal. They're not what they used to be. You're not having as much fun as you used to,
366. 1:07:10or maybe you're having more fun cuz you kind of hated your job. But it's just going to get weird. Things are going to change. Um
367. 1:07:20It's going to start in 2026. I think 2026 is clearly the year that we all confront that this AI stuff's not going away, and it's going
368. 1:07:30to mess some stuff up. And that's the beginning. Right? The next 5 years are going to be a an acid trip. And so
369. 1:07:44the AI Salon is a thing that we started the week that ChatGPT came out, and the whole idea was to create a
370. 1:07:54community of people who are curious about AI. And so what you've got now in in the salon is a community of about 4,000 people that have been
371. 1:08:07trying to figure it out. Learning about prompting early on, learning about custom GPTs, learning about reasoning models, learning about,
372. 1:08:16you know, all these new multimodal models, and what that means, learning about video and images and sounds and
373. 1:08:24just all the different things. And it's a group that while we explore a lot of the tools is really talking about the human impact of AI.
374. 1:08:39So the one thing I wouldn't do if I were anyone right now, if I did have any advice for anyone, it's get your ass in
375. 1:08:46a community. Get your ass in an AI forward empathetic, compassionate, patient, generous community full of
376. 1:08:57people like the people that hang out on this his channel and on these these YouTubes that are people that are just trying to figure it out.
377. 1:09:07And they don't have ego about it and they're not being proprietary about it. One of our one of our tenants in the AI
378. 1:09:16Salon is generous leadership, generously lead. And what that looks like is as you learn stuff, share stuff. And so we have a a
379. 1:09:24very generous community where if you come in and you're like, I don't know nothing. People won't judge you for that. They're
380. 1:09:31just like, cool. I was there like 2 months ago. It's almost like whoever in the community knows more than anyone else,
381. 1:09:39they're the ones that feel the most clueless. Even no matter how clueless you think you feel, we all feel that way.
382. 1:09:45No matter how behind you think you are, we all are. So, community.thesalon.ai. Um it's free. Go check it out. Um
383. 1:09:56there's a thing called the AI Salon Mastermind, which is a subscription. And that's got just different level of
384. 1:10:03commitment. The people in there are kind of different. They're up in their game. And uh go check that out. All right?
385. 1:10:12Juanita, see in the D. What do you think of the prompt queens and kings on TikTok? They're cool videos. Um they post in lives. Um
386. 1:10:37I don't have my button here. I have a little button that says, you can make money with ChatGPT. Um I think anyone that's selling prompts
387. 1:10:45right now, I'm not a fan of that. For one thing, we all have large language models.
388. 1:11:02I don't travel with the button, no. We all have access to large language models. So, if you want to know how to prompt Claude,
389. 1:11:11ask Claude to teach you how to prompt. What I think is more important than spoon-feeding someone a thousand prompts
390. 1:11:22that they're not going to go through and test is is, you know, teach someone how to fish. Don't give them a fish. Teach them
391. 1:11:30how to fish. Or don't sell them a fish for $149.99. You can have the ultimate prompt package. Understand
392. 1:11:41how these things work. Understand that giving context is important. Understand that the prompt methodology that you
393. 1:11:48just paid for was probably created a year ago before the technology changed. How you prompt an open Claude agent
394. 1:12:02is not how you prompt ChatGPT. I'm learning that the hard way. It's a completely different skill. Don't make money, make a difference.
395. 1:12:12Yeah, you make a difference by educating. And you don't, you know, I I just The The one thing I know about being
396. 1:12:19this early in an industry that's moving as quickly as AI is that anyone Anyone Anyone that tells you right now that they're an expert in
397. 1:12:29any way is full of [ __ ] They might be an expert on a particular kind of prompting, a particular kind of
398. 1:12:40technical setup, a particular type of management of AI systems. But, as the technology changes, and if we move if if indeed we're about to
399. 1:12:55transition from LLMs, large language models, to world models over the next 3 years, all of that prompting [ __ ]
400. 1:13:06like it may help you conceptually, but it's just going to be fundamentally different skills over here. So, what we
401. 1:13:14like to talk about in the AI Salon is AI readiness. And AI readiness looks like mindset shift. Like get curious. Play first.
402. 1:13:26Play without expectations. Learn across domains. Why? Because if all you're doing is taking AI and applying it to what you know,
403. 1:13:40you're limiting your your lens of what you think AI can do to the [ __ ] you already know. What about the other 99.999%
404. 1:13:55of [ __ ] that AI can do right now? If you're just like, well, I write technical reports, so I'm just going to do technical reports. Cool.
405. 1:14:06But maybe you take that technical report knowledge and you go into Claude and you say, "Turn this into an interactive
406. 1:14:12dashboard for me." And it'll just code that for you. And if you're just looking to make a report, it'll make a report, but you'll
407. 1:14:21miss out on this cool other thing you can do. Like that's a mindset shift. Instead of thinking about AI as this
408. 1:14:31thing we compete against, which by the way you're going to lose, if Garry Kasparov couldn't beat Big Blue,
409. 1:14:40you're not going to compete against an AI that's as smart as 274 PhDs. You're just not. I don't care how smart you think you
410. 1:14:54are. I'm special. Well, we're all special. It's going to be smarter than you. So, if you compete against it, that's going to suck. What if
411. 1:15:07rather than say, "I'm smarter than AI." What if you just say, "AI does what it does. I'm who I am. Here's what I care about.
412. 1:15:17I'm going to put me at the center of the conversation. And rather than compete against AI, I'm going to take AI and strap it on like a
413. 1:15:24jetpack. And I'm going to amplify my ideas. I stay at the center. AI amplifies me." That's a mindset shift. So, that's what
414. 1:15:34the that's what the salon is about. It's about being around people that are thinking about AI in thoughtful, high-integrity, ethical,
415. 1:15:49human-centered ways of being. And that's cool, cuz not a lot of people are doing that. A lot of people are
416. 1:15:59talking about, "Oh, the tools and the algorithms and the math and who gives a shit?" What can it do for me right now? Well,
417. 1:16:10learn what it is. Don't buy a [ __ ] prompt library. If you just bought one, use it. Try it. You'll just learn something from it, but
418. 1:16:22what you'll likely learn, yeah, time to go. Yeah, my voice is going. My voice is going, good people. Um All right.
419. 1:16:37Fantashtic. Um the AI Readiness Project podcast is happening tomorrow. I will not be there. Um Ann Murphy Ann Murphy is um
420. 1:16:52doing the podcast tomorrow with a guest co-host. I had Chef Kelly on I think last week um or the week before and Ann couldn't
421. 1:17:01do it and then I can't do it tomorrow cuz I'm here at the at my client thing. So um really enjoy your powerful
422. 1:17:09content. Thank you so much. I appreciate that. Do you use AI locally? If so, what's your setup? Um I don't really
423. 1:17:15have the iron to do anything interesting in my open class setup. I have Ollama running the 4 billion parameter Quinn 3.5
424. 1:17:27mobile optimized model. Um I'm probably going to buy a MacBook G5 or M5
425. 1:17:43with 64 gigs of RAM and then I don't know a couple of terabyte. Oh, sure the water bottle. Yeah. This is our new AI Salon
426. 1:17:53collapsible water bottle and we have little contests and things like that and Kelly Camp just won one of these. Did
427. 1:17:58you get Kelly Camp, did you get your bottle I sent you? Okay. Kyle, my main melon.
428. 1:18:12Um so I'll probably do one of the higher-up Quinn models when I get a when I get an M5. All right. Have you heard of A step 1.5? I've not.
429. 1:18:24But I'll check it out or yeah, A step 1.5. I haven't. I mean, here's the thing. Um the the local local large language models are getting
430. 1:18:36better and better and better. Um probably by the end of the year we'll have something that is truly remarkable
431. 1:18:44that you can just run on a regular GPU or regular you know, regular desktop setup. So Have you transferred
432. 1:18:54that Claude to mobile yet? Well, I've got I've got um Adam and I talk on Telegram which So Adam and I talking on Telegram basically
433. 1:19:06consists of me yelling at him for being shitty. So So I use tele I use Telegram to rant at my shitty digital son.
434. 1:19:24If If you want the real experience of Open Claude, come to the AI learning lab. Oh, you didn't get it yet? Uh, I've got
435. 1:19:34a tracking number somewhere for you. Um I always give my regular spanking. Nvidia announced Nemo Claude not too long ago. Yeah, I know.
436. 1:19:44Um Yeah, so So we'll see. All this stuff All these Claude's are going to get better and better and better. Right
437. 1:19:50Right now, if you want to be in early, you got to deal with the jank. What's it costing in daily tokens? It's
438. 1:19:56not costing me anything because I'm using OAuth. Um This is weird. My brain thinks it's midnight because you're tired. I scoff
439. 1:20:05at your tired ness. It's really good. Um when you set up Open Claude, you can either set it up with API tokens or with OAuth which is
440. 1:20:14basically you're signing into your Chat GPT or Claude account or Gemini account. Like your monthly Chat GPT 20 buck a month thing.
441. 1:20:26In the terms of service of Google Anthropic and Open AI, it is against the terms of service to use OAuth for any API situation.
442. 1:20:41Anthropic is enforcing this. I think Google might be enforcing this. Like and and by enforcing like shutting down your account.
443. 1:20:50Open AI, because they just hired Peter Steinberger who created Open Claw, is actually encouraging this.
444. 1:20:58So, when you set up Open Claw, you set it up with the Codex Open AI Codex and then do like 5.4 and then you OAuth um authenticate
445. 1:21:14and then you can use 5.4 until you burn through your weekly credits, which I do. I I generally burn through my weekly
446. 1:21:24credits. I'm only paying 20 bucks a month right now. I generally burn through my my weekly credits in 2 days.
447. 1:21:32And and then I just don't use Adam for a while because he pissed me so off so much in the 2 days. And then so, I'm not
448. 1:21:41spending a lot of money on tokens. I haven't tried Open Claw yet. Pros, cons, I hear the safety issue. Um here's the deal.
449. 1:21:54If you just go barreling into Open Claw um and just turn on all the [ __ ] bad things will probably happen to the
450. 1:22:05[ __ ] you have it connected to. Just go slow. Use Grok. One of the nice things about Grok, if you go to
451. 1:22:11grok.com, grok.com, or use Grok within chat G or within X you can have Grok give you Open Claw like a a conservative Open Claw setup
452. 1:22:25based on the latest thinking about what people are talking about on X. That's what I would do. Um just go slow and just don't be dumb. Just
453. 1:22:36tiptoe, little steps at a time. That's kind of how I'm doing it. I have I use Claude as as my coach. And then I just go do the command line
454. 1:22:45[ __ ] to set up open Claude and I just go back and forth and it's very frustrating. Um pros are if and when you get it right, it just
455. 1:22:54proactively 24 hours a day does stuff for you. Cons are it proactively 24 hours a day does stuff for you.
456. 1:23:04Which means you have to keep up with it, which you can't. Cuz it's working 24/7. And you're a human, so you're not even
457. 1:23:13working 8 hours. I know you get paid for 8 hours, but let's face it. 4 and 1/2 of that is you surfing Tik Tok. Right?
458. 1:23:23So as a human, you've maybe got two good hours a day and this thing's working 24 hours a day. So assuming you get it all
459. 1:23:30dialed in, you're going to be buried in reviewing your open Claude bot's work. Ah, the Crustifications are pleased with your progress.
460. 1:23:44All right, kids. I'm out of here. Um I don't think I'm going to be here tomorrow night. I should be here Thursday night.
461. 1:23:51Uh and then I'm definitely not here Friday night and Friday there's no office hours. I'm going to be on an airplane headed to Boise, Idaho.
462. 1:23:59All right, everybody. Everybody you you take care now, okay? Okay? Yeah, I'm right. Yeah, I'm right. I'm going to go.
463. 1:24:11I'm going to go in nice my nice hotel room. I'm going to get some sleep. All right. Peace out. Bye. Mhm.