6/3/2026 - The Power of Vibe Coding: Building Functional Apps Without a Tech Team

Transcript

1. 0:00All right, I'm here. [music] [music]
2. 0:15[music] >> [music]
3. 0:30[music] >> Welcome to the AI Learning Lab live. It's live. We're live here. If If I weren't live, could I do this?
4. 0:45No. No. But we're here. We're doing it. It's Wednesday. Wednesday is the night we do stuff. It's the night we make
5. 0:53stuff. So, what am I going to do tonight? Not that. Why is it because I'm just a nerd do? Well, a scoff law.
6. 1:06It's part of it. It's part of it. Nobody puts baby in the corner. We're not a slave to our format. Producer Brandon's like, "We should have
7. 1:18a format. we should make it slightly predictable for me so that it's not just pure panic and chaos night after night after night.
8. 1:27And I said, "Sure." I agreed to that for an episode or two. [laughter] Hey, producer Brandon, what's happening? [laughter]
9. 1:44Oh my god. Okay, >> I'm gonna I'm actually gonna hijack uh the the our own format. Um >> Oh, beautiful. Go ahead. >> Because we've just had
10. 1:55>> what wasn't the last uh episode of the four-week lovable sprint >> because the four-week lovable sprint has
11. 2:03now become a fiveweek lovable sprint because >> there's just too much good stuff to fit into four weeks.
12. 2:10>> Well, you know, it's funny funny you should say that. A similar thing happened on the way to the uh four-part
13. 2:16um separating your identity from from your job workshop in the mastermind in the great repurpose um where Andy has
14. 2:24agreed to continue that as a weekly um you know great repurpose mastermind um through the end of the year. So so
15. 2:32every two weeks that one's continuing too. So [laughter] well but so it's a good point. So I do want to hear about it. So, I want to
16. 2:41tell tell me all about it. Um, but but I think this is this is an important thing that we're in the middle of living in an expansion
17. 2:52>> and and and when you're living in a contraction, you know, you can teach things in sort of tight little contained repeatable
18. 3:00packages. When you're living in an expansion, just keeping up with what's happening, and I think Lovable is
19. 3:05probably a good example of this, is actually really hard. So, uh, tell tell me what happened. Let me I got to go
20. 3:10change something, but I'm listening. >> Yeah. So, what we did this week was focused on That's not me. Uh what we did
21. 3:19was we focused on publishing and polishing. And we talked about the difference between uh an MVP, a minimum
22. 3:27viable product and a fullyfledged web app that is sitting on GitHub and has software developers working on it. We
23. 3:35talked about the challenge of vibe coding something and having technical bloat and technical debt and talking
24. 3:44about refactoring code and changing and streamlining as you go. But then we also talked about how uh at the end of the
25. 3:53day if we get get one deliverable product um it's a win. It doesn't necessarily me we're not the success is
26. 4:02not measured by how many people come to visit your site because although lovable will make it Google friendly now it's
27. 4:11still Google and there's still 10 billion websites out there sponsored listings AI overviews and everything
28. 4:17else competing for that real estate so some of those old school marketing tactics >> are um you know still valid in terms of
29. 4:28trying to make your app a reality. I see it with vibegreeting.com. To me, it's the coolest thing to be able to make
30. 4:36cards and have them generated and sent out in the postal. Um, but until I tell somebody about it or somebody shares a
31. 4:44tweet that I make about it and the right person sees it, it doesn't really gain traction. And so that's where our
32. 4:51recommendation was to start local uh and work with um your chamber of commerce and other people in your local sphere to
33. 5:00boost it locally and grow from the ground up. So, we've launched soft launched um a happy habits adventure for my um
34. 5:14for my web app uh and for my kids. And >> uh I was going to show we were just working with uh the 11 Labs uh key to
35. 5:27get the stories to read back to us. And wouldn't you know, we ran out of time and I ran out of credits on 11 laps [laughter]
36. 5:39making the entire show last week in AI. Um, you know, but we are working on fixing that. So stay tuned next week.
37. 5:50But, uh, yeah, we're we're really excited and I'm really excited next week to see what everybody else has been working on behind the scenes.
38. 5:57>> That's beautiful. That's really good. Really really good. Yeah. And I also I think the other thing I think local's
39. 6:01really interesting, Brandon, and then and then you also said it. I think that the personal touch of you as a founder
40. 6:09of, you know, that software, whatever that site is, whatever that thing is that you built, talking about why did
41. 6:15you build it? What does it mean to you? Like it's your kids, you know, calling them the tidbits. All of that
42. 6:20storytelling about who you are and why you did it. That is the one thing that no one else can copy, right? they can
43. 6:28copy everything else. They can copy the app, the look and the feel, the functionality, every piece of it, the
44. 6:34marketing, the advertising, right? If you're doing it all really well, that's just going to be swiped. But the one
45. 6:39thing that they can't, you know, copy is is you and the connection that people have with you. So, I think that's right.
46. 6:46Start small, start personal, and then blow it out from there. >> Yep. And Andy's gonna vibe code uh the site for humans 2.0. I love it.
47. 6:55>> There you go. That's beautiful. There we go. And and I know you'll appreciate I'm gonna see if I can figure out what's
48. 7:01going on with my 11 labs. Maybe I'll hop back up at the end. We'll take a look at it. But Gemini, >> but um I know you'll appreciate my
49. 7:08shirt. I was huge on Google+. [laughter] >> Remember circles? >> Yeah, >> that was Plus, right? Circles plus. Yeah.
50. 7:19>> Wow. That's a that's a little time capsule. All right. Cool. Well, thank you, producer Brandon. >> Yep.
51. 7:25>> Welcome, welcome, welcome. So, here's what we're going to do tonight. Um, you old lunar man. Um, here's what we're
52. 7:32going to do tonight. Um, I'm going to go I'm going to go start it's going to be a bit of a commentary night, kind of like
53. 7:38Monday night. Um, we might go play with some stuff, right? I'll probably go play a little bit with Idog, but I I let my
54. 7:46subscription there lapse, so I only have really slow production. Um, ID at this point, ID for a while was the
55. 7:56single image generating tool that could do text and then they kind of disappeared for a while. Um, and so now
56. 8:02they've ca they came back today with 4.0, their new 4.0 model. Um, which you can also run locally which is kind of
57. 8:09cool. Um, so we might go play with some stuff like that. Um, but there's just there's like there's something in the
58. 8:16air. Something big is coming from OpenAI tomorrow. Um, I think it involves codecs. I think it probably involves a new model 5.6. Um,
59. 8:30my instinct, my strong instinct, do I even want to say it? Let me feel it in my body.
60. 8:45I think that what they're going to do is they're going to collapse chat GPT, their web browser, and codeex into a single app.
61. 8:57I don't know if it's going to happen tomorrow. It could, but I have a feeling that those three things are going to
62. 9:03collapse. So you'll have a web browser that is also chatgpt that is also an agentic platform that can do really cool
63. 9:10[ __ ] locally on your computer. Um they are going hard after professionals that want to do all the things like all
64. 9:21the stuff we're talking about here tonight, right? We're we're doing seminars and we're vibe coding sites and
65. 9:27we're doing all this stuff, right? And they want you to when you go to your computer, the window that you log in
66. 9:34from, they want it to be chat GPT. Just like Google wants it to be your default search engine. I think OpenAI is going
67. 9:41to want you to be, you know, logging into their stuff and and them owning all of the stuff, which scares the crap out of me. And
68. 9:51that further reinforces that I think um being able to run models locally is actually really important and and having
69. 9:59your own data locally. Um I also want to go look at Hang on, let me let me throw a thing up here just so I don't forget it. >> Ness,
70. 10:08>> I'm back. >> Hello. I I fixed the uh lovable credit error and I and I promised the uh the lovable crowd that came to the lesson
71. 10:18that if they came over here to learning lab because we were literally out of time uh that I would I would show them
72. 10:24the fruit of our labor um and as we went. So I've got this up here and so let me go ahead and share my screen.
73. 10:33And so we've got this is on lovable now the uh bedtime but magical. So it's geared for ages 5 to 8. Incidentally my
74. 10:43kids are five and eight. Uh but you pick the guide and then you run the quest. So you try to check the boxes of morning
75. 10:51and afternoon. Uh and uh you Mary Chester, we recorded it. So don't worry, it'll be up there shortly. Actually
76. 10:57recorded it this time, not AI generated. [laughter] But then it generates an AI bedtime story. And so if we come over
77. 11:04here, uh, we ran through the demo version. And so Naomi, uh, had her morning quest and she had to wake up and
78. 11:12have a big stretch, get dressed, brush teeth, eat breakfast, shoes, and backpack. Uh, and then at night, we have
79. 11:19five that we put on pajamas, brush teeth, tidy room, read the story, and lights out. And then at the end of it,
80. 11:26it does a story. And if we say read to me, uh, it should wake up my friend. and it should read this to me. And um that's what
81. 11:38>> Captain Coral's cozy coat. >> Naomi snuggled deep into her bed. The soft blank feeling like a warm hug
82. 11:46beside her. Captain Coral with his kind sparkly eyes smiled. "Well, me Hardy," he rumbled gently. "What a day you've had." >> So obviously
83. 12:00it's a good for a demo. But there's so much more. This was the other thing we talked about. There is so
84. 12:05much more that I want to do with this app. Like I want to be able to customize the routine. I want to be able to change
85. 12:11the voice. He should actually sound like a pirate instead of the guy that's letting you watching telling you to mind
86. 12:17the gap at the subway platform. You know, there's so much more that we can do. And the the vibe coders job is never
87. 12:24done because when you're just thinking things into existence, you never stop thinking. So you never stop building and at some point
88. 12:34>> you kind of get to that pens down moment but this is just for right now an MVP but it works. >> It's beautiful. It's it's it's awesome
89. 12:42to see and it's also listen I mean I think that you know if you've been in product development before if you've been an
90. 12:50engineer before I mean that is kind of how it goes but normally that's a two to five to 10 person team that are all sort
91. 12:58of vying for hey here's the priorities here's what I wanted to do and then that's got to go through the approval
92. 13:04committee and then you've got to have the tech team tell you that well that's way more complic that that simple thing
93. 13:09you want is way more complicated than the thing you think is complicated. So, we're going to change your priorities
94. 13:14based on technical require
95. 13:30>> true to form. One thing that I learned at the Zenus conference was that the CEO took something that had been on their
96. 13:35road map for four years with like a hundred up votes in the community and just took an hour and vibe coded a solution. And he >> the CEO did
97. 13:46>> Yeah. the CEO. He's like, "We're going to fix this. I'm going to fix this. It's been long enough. We're going to fix
98. 13:50this." And so everyone's like, "Well, when's it going to be released?" He goes, "Well, it has to go through
99. 13:54testing. you can't just they're not just going to let me soft launch this thing into the community without stress
100. 13:59testing it against security and everything else. But the point is that it got unstuck and it got built and um
101. 14:06you know so moving from idea to implementation there there is no road it's just >> there go yeah do it.
102. 14:15>> Yeah it's beautiful. Congrats. Okay so if you're part of the AI salon mastermind if you'd be so kind Brandon
103. 14:22pop up the the mastermind graphic. Um if you're not a part of the mastermind you you need to be a part of the mastermind.
104. 14:29Um, if you're feeling overwhelmed, if you're feeling stuck, we've got the practice lab, which by the way, that's
105. 14:34tomorrow at 10:00 a.m. Mountain time. We've got these workshops. Um, and and so next Wednesday is the fifth out of a
106. 14:44four-part series. It's the fifth of a four-part series on Vibe Coding with Lovable. All of those recordings of the
107. 14:51first four are in the mastermind. So, um, if you haven't joined the mastermind, you should join it. If you
108. 14:58have joined the mastermind, um you should take advantage of these programs that are expanding and going deeper. Um
109. 15:06and it's it's a it's a it's powerful powerful thing. Okay. So let's I don't know about the practice lab. Okay. So let me tell you about the
110. 15:18practice lab. One of the things that started happening in the community was our members were
111. 15:33cheerleed by me getting increasingly overwhelmed with trying to keep up with AI. You know, I'm gonna go live seven
112. 15:41nights a week. We're going to learn all this crap. Every time there's something new, we're going to learn how to do it.
113. 15:44We're gonna learn how to do it. We're gonna tell you how to do it. We're going to put links to all this [ __ ] in the
114. 15:49community. you're going to learn how to just come here, we'll learn. And then and then about a year and a
115. 15:58half in, it became clear you couldn't keep up with that [ __ ] anymore. And yet people in doing all that
116. 16:07learning had discovered, oh, you know, there's there's things I can build and there's things I want to build. And that
117. 16:12anxiety level stayed as as people gained mastery of some of the tools, that anxiety level stayed. and people were
118. 16:20starting to get burned out and frustrated and didn't know what to do and didn't know where to face and we're still trying to keep up.
119. 16:29And so, um, so we were looking at, um, ways that we could support the community and especially people that were, you
120. 16:39know, really doing this. And Leah and I when when Leah Fon was still here, we started the mastermind. I don't know when it was, maybe a year
121. 16:47ago. right around this time, a year ago. Um, and we didn't really know what it was. And, um, yeah, it was about this
122. 16:55time. It was about this time last year, we started the mastermind. And then in August, Leah took a road trip. She drove
123. 17:00across the country. She dropped her kids off at their college and she decided to move to San Francisco and she stepped
124. 17:06stepped away from the AI salon. And we brought on Liz Miller Gersfeld. And Liz had gone through this similar
125. 17:14cycle of like gotta learn, gotta learn, gotta learn, gotta build, gotta build, gotta build, got, you know, a little
126. 17:21little ampied up. And then and then through her own daily practice started to figure out how to integrate this into her life. And so we were
127. 17:33talking about um that and her daily practice and and um it inspired the mastermind practice. And so we created this structured um
128. 17:46this structured way to look at a daily practice of yours about how you use AI that essentially follows the cycle of AI
129. 17:54readiness. Play first, create excellence, generously lead. And so it's a it's a 10-week cycle. So, the the
130. 18:01mastermind practice lab happens every week on Thursday at 10 a.m. Mountain time, and it's a group of people who are
131. 18:10doing a daily practice using AI where we get together and we walk through this framework. And so, it's a a 10-week
132. 18:16cycle. Um, and we're on our third cycle right now. We're right about midway through it. We're coming up on the
133. 18:22fourth or fifth week. Um, and what's powerful is we've had a number of people who've now gone through it two and a
134. 18:30half times. And to a person, what's happening is that the people that are practicing, doing a daily practice on how they use AI,
135. 18:44they're slowing down. They're stepping away from the computer on a more regular basis. They're getting
136. 18:50much more intentional about, well, wait a minute. what do I actually want to do? Right? So, they're not just in the rat
137. 18:55race of I have to do AI, they're starting to get to what do I what do I want? What do I care about? And so, the practice lab is just that
138. 19:05weekly check-in to be with a group of people who are going through it. Because it's not always easy. You can ask a question like,
139. 19:15"What do I want more of in my life?" It's a really easy question to ask. It's actually a really hard question to
140. 19:21answer. Like I'm in the middle of it right now. It's a mess. So anyway, all of those recordings of the mastermind
141. 19:28practice lab, they're all recorded. They're all archived. So if you end up joining um the the mastermind, you have
142. 19:35access to all those those past meetings. Uh and then you're welcome to join the the weekly meetings. So that's it. And
143. 19:42it's uh Sid Harrow and myself uh are co-hosts of the the practice lab now. So that's once a week. All right, let's go look at some some
144. 19:51news kind of stuff. And again, I might play a little bit tonight, but there's enough going on that I think I just want to talk about stuff.
145. 20:00Um, what do I want to do here? Do I want to do that or do I want to do window? I think I want to do window.
146. 20:18go here. Okay. So, Grock Imagine 1.5 which was just released. So, that we could go play with.
147. 20:34There's Idog. I did a muscle car shot. This is This is the new ideagram 4.0. It's not bad. I mean, it's I feel like all these models at this
148. 20:48point are decent. Um, I don't think this is any better than uh than what I saw with the new image gen from Chat GPT. But let's um
149. 21:04let's go look at Grock Imagine 5.0. So, we're going to go to grock.com. gr.com. I think you need to have a paid X uh whatever you call it
150. 21:20subscription to have access to this. I think I pay seven bucks a month or nine bucks a month. Comment from Sharon.
151. 21:30All right, let me go look at this. I asked chat my I asked my chat GPT to help me on my path to retirement and
152. 21:39what it looks like to me. Not really retired but life path change. It was cheaper than therapy. Yeah, exactly. [laughter]
153. 21:50Listen, I I mean one of the things here's the thing about large language models. They have all the psychological
154. 21:58education in them. All the knowledge is in there. Um they're infinitely patient. um they they they tend to be a little
155. 22:05sick of fantic. So so they're always going to tell you you're swell for the most part. Um but if you want to process
156. 22:12through some stuff, they're actually really good at it. Okay, so let's look at image quality. Are you going to go speed or quality agent?
157. 22:27Well, that's agent. I don't want agent, do I? I want image. How do I know what model I'm using? All right, let's do we'll do we'll do same
158. 22:40thing. 1970s here. You know what? In fact, what I'll do is I'll go back to ideoggram. I'll get the exact prompt that I used for this.
159. 22:58All right, here we go. One thing that I have noticed about Grock is it's incredibly fast.
160. 23:18See, I think that's better. I think this is better. Wait, why did that do that?
161. 23:37Muscle is the future. They're all Challengers. They're all 69 Challengers. That's kind of a That's a hybrid. That's sort of a Camaro
162. 23:53Camaro looking thing. And then what's cool about these is I can take that and just say um car starts
163. 24:12wait why can I not type and rolls away
164. 24:3723%. So, this is pretty fast. If you haven't played with with Grock, imagine um it's getting pretty good. Here's
165. 24:47where I see XAI is that they are very much in fourth place. They're not really considered one of the leaders.
166. 24:59Um, but they're getting better quickly and I could see them. They just launched an agentic platform called Grock Build.
167. 25:08They've got models that are pretty good. They've got these image and video gen that are pretty good. Um, they're not
168. 25:14ridiculously expensive. They're pretty fast. Um, so they're worth paying attention to if you can stomach Elon.
169. 25:21And if you can't, I understand that. But, you know, it's at least worth knowing what's here and what's coming. [laughter] All right. 90%.
170. 25:32Let's see what it does here. There you go.
171. 25:54It looks like the car needs needs uh needs a tuneup. That's better. That's usable. That's actually, you know, the wheels look like
172. 26:10they're rolling like physics. Great job this morning, Kyle. What do you have on tonight, Kyle? Okay, so let me jump back here for a second.
173. 26:23So, first off, thank you for that. Um, this morning I was on the uh the Daily AI show uh with the crew over
174. 26:32there. It was the first time I did that. And uh Gareth was in the hood and Andy was there and who else was there? There
175. 26:39were a couple of people there that I recognized. So, thank you for everyone who came and hang out, hung out. Um that
176. 26:45was an oil burner for sure. Yeah, exactly. They uh there were some gasket issues in that in that resto mod. Um so,
177. 26:53so what I'm going to demo tonight is probably not it's it's going to be a bit of a poperri tonight and it's not really
178. 26:59a demo night. I know Wednesday is demo night, but tomorrow we're getting something from OpenAI. And I think as a
179. 27:07um people trying to jump on that bandwagon, there were lots of announcements today as well as the OpenAI stuff. Um, the other thing, I
180. 27:16don't know if Lord Digital Gods is in the house tonight, but I finally got a chance to go play with the wiki that LDG
181. 27:25put together about all of the AI learning lab recordings. It's amazing and it's right now it's open to everyone. Um, so I'm going to
182. 27:36talk to him about that, but he's using it as a source of truth um for some other AI stuff we're doing. So rather we
183. 27:43used to have basically just a PDF that he would maintain. Um he's now turned that into this functional wiki. It's
184. 27:50it's really amazing. In fact, we can go we can go look at it right now. Um is this it here? Yeah, this is it here. So
185. 27:59let me go back to the top. Let me just make sure I'm sharing. I am sharing. Um, so what this what you can do here is
186. 28:13you can search for so if I want to search for weird Mary and I can search what whatever fields all the fields or I can say let me just
187. 28:26search the chapters um and I can do sort by newest. I'm going to say sort by oldest. So 5224 creating weird Mary from Cedar Hill. And
188. 28:39so I can just click on this and it'll take me straight to that chapter. Start the start the process to write weird Mary. Two-hour process.
189. 29:11No audio on the YouTube. What the [ __ ] Hang on. I must not have turned on my uh the share audio. I didn't turn that on. Share. Okay.
190. 29:34Play all the songs I've created. So, I don't know if you remember this night, but this was like a twohour
191. 29:40process of making this song Weird Mary. This is [ __ ] This is cool. >> This is [ __ ] This is cool. Four working. Two with not quite right
192. 29:53lyrics. Two with the right lyrics. So, we're we're going to go bridge. Next thing we'll do is a chorus. [music]
193. 30:03>> All right. We got to go fix some things. So anyway, so so that's the um that's the AI learning lab wiki, the YouTube
194. 30:13wiki. So you can search for stuff. You can just scroll through stuff. Let's preview home. Here's videos. There's live streams. The
195. 30:26goop phase. Let's see how many times I mentioned goop. Goop. [laughter] We'll do oldest first. [clears throat] So, I mentioned Goop.
196. 30:40Let's see. 511. May May 11th was the was the oldest of this year. Look, the goop phase,
197. 30:58there's there's lots of ways to participate, right? One is you can just lurk. You can just watch what other
198. 31:04people are posting. You can come to turn into a butterfly. They actually have these enzy phase. So this is a remarkable
199. 31:14remarkable resource. Um so if you've been if you're a regular if you've been here for a while and you remember all
200. 31:20the wacky hijinks we had you know I can look up the word janky and go fields. Let's just look in again. We'll just do chapters and then
201. 31:34we'll go by oldest first. September 3rd, 2023. The leading edge but janky
202. 31:55of those is fine because they're coming from a place of knowledge rather than ignorance. So, all I'm trying to battle
203. 32:01is the the ignorance and and um the URL of the wiki is a salonwiki.com. AI salonwiki.com. How was the wiki made? Um Lord Digital
204. 32:16Gods is a genius and he loves data crunching and building. How many reference [laughter] How many references for
205. 32:30March? Good question. All right, let's see. March. [laughter] Let's see. Marge. Marge. Marge. Marge. Where's Marge? Here we go.
206. 33:00Where's Marge? [laughter] What will we do with you? Oh, it's time for me to learn this. Wait, what was that? Who put that one in
207. 33:12there? Yes, I saw him, too. I checked. He wasn't there for me. That's awesome. So cool. I mean, what's funny about this is
208. 33:24Google just announced um YouTube, ask YouTube, I think they're calling it, where you're going to be able to just across YouTube look for the
209. 33:35best video clips that say a certain thing. But the ability to just mine an existing channel like this is this is
210. 33:43this is really quite something. So anyway, so thank you to Lord Digital Gods for this. this becomes the
211. 33:49foundation for all sorts of cool stuff we can do with this. So, it's pretty exciting. Um, okay, let's go back here to So, so
212. 33:58Grock is worth playing with. Grock one, imagine 1.5 is the latest image and video model from them. So, it looks like it's pretty decent.
213. 34:09Um, [clears throat] all right. hours a year just on food shopping, searching, scrolling endlessly, delays,
214. 34:19substitutions. Gouff changed that. Now, in partnership with SpaceX AI, we have rebuilt the shopping experience itself.
215. 34:27Introducing Go, your personal [music] shopping genius. Powered by Gro, the most advanced voice AI ever built. [music]
216. 34:37>> Go. I'm running late. I need my go-to breakfast. >> Got you, Sarah. I have your energy drink, protein bar, and banana ready.
217. 34:44>> Sarah doesn't search, scroll, and waste time building a cart. >> That's perfect. Send it to my office.
218. 34:50>> By the time she's at her first meeting, >> that'll be ready in 15 minutes. >> Breakfast is already at her desk. Go
219. 34:57understands your environment, your local [music] weather, the game tonight, what's trending where you live. We
220. 35:03co-developed a custom image experience that turns those signals into a moment made for you. It's 65 and sunny barbecue.
221. 35:12>> Yeah, now we're talking. >> So anyway, so that that thing was created with with uh with Grock. Um I do want to talk about SpaceX AI for a
222. 35:22second, but before that from Gareth, you said my name 69 times. [laughter] Oh my god. Baby face Shannon. I know. I
223. 35:37know. It's you. You can tell you've been doing this for a while when you bring up a video of yourself and it shocks you
224. 35:43what you look like. 632, hold, please. Um, yeah. So, if you want to put together AI Learning Lab trivia, it's
225. 35:51probably now going to be easier to to mine all that data. So, anyway, so that's a that's a that's an LDG masterpiece right there. It it's
226. 35:58actually really really valuable. Um, just just as a like one of the things that I like to to think about with with all of us is
227. 36:10like we are going through history right now, right? You know, and we've been since we've been doing these lives uh since April of 2023.
228. 36:22Um, and that's that's that's some history right there and you know the different phases that we've gone
229. 36:28through. And so those are all captured. So all those moments are now really easily discoverable. Which listen, as a data hoarder,
230. 36:37I know we've got some data hoarders in here. We have we have we have the neurospices that love to hoard their data. As a data hoarder,
231. 36:44the ability to actually go back and mine it is is really quite a remarkable remarkable thing. That's what I was thinking. The
232. 36:53historical sign significance of what LG has put together is amazing. Yeah, great work, LG. In fact, you know, it like
233. 37:00this this might this might even be the kind of thing that that is worth putting together like a press release. Maybe we
234. 37:06can get, you know, get an article on it or something like that. Um, and uh, you know, when they're like, "How did you do
235. 37:12that?" I'm just point to Glenn. Glenn did it. He did it. [laughter] Okay. So, let me tell you a cool stat I
236. 37:22heard about Space Xai. So you remember a few couple of weeks back couple of things happened. Two things happened.
237. 37:34Well, this was this was early May. Early May and probably in April, Anthropic was getting a real bad reputation
238. 37:47that they basically didn't have enough compute. [laughter] And so they didn't have enough compute. And so people were going back to open
239. 37:58AAI like people had when when when Opus 4.6 came out or 4.5 4 No, it was 4.6. When it came out in December,
240. 38:07everyone was like, "Oh my god, Claude's the way to go. If you're a coder, you've got to use Opus 4.6. It's the promised
241. 38:13land." And so a bunch of people left OpenAI and went went over to that and and a bunch of uh businesses started
242. 38:19taking anthropic a lot more seriously. And then they basically ran out of compute. Their demand went up. They ran out of processing chips.
243. 38:28And then they announced this weird deal. I think it was right after right after the judge threw out the lawsuit between Musk and Sam Alman.
244. 38:40Um for you know statute of limitations reasons. They announced this deal between Anthropic and SpaceX where SpaceX took their Colossus One
245. 38:49supercomput, this big data center and they they they licensed it to Anthropic and then they rolled SpaceX and XAI together. Well,
246. 39:03they're getting ready to go public. And apparently the deal with Anthropic is Anthropic is paying something like $1.5 billion a month for
247. 39:18um for for Colossus for the compute. So they're paying SpaceX SpaceX AI $15 billion a year and I think SpaceX
248. 39:29generates 12 billion a year or it's like it's it's slightly above or slightly below the 15 that they're getting from
249. 39:37XAI from from leasing this Colossus supercomputer to Anthropic. So they've got guaranteed revenue. They're going to
250. 39:45go public. It's going to be a multiple of that revenue. So they're literally just taking this thing that they built
251. 39:50to go build AI. they're turning it into revenue. They're going to go public on the back of that and just basically
252. 39:56print money off that thing. So absolutely wild and insane numbers like just the speed with which these deals
253. 40:03are happening and you know to go from zero in revenue to to 15 billion dollars a year in revenue with one data center.
254. 40:13Holy crap. So if you if you don't think that that the building and installation of these data centers is is is going to happen, it's
255. 40:24going to happen. When there's that much money at stake, it's going to happen. Anyway, all right. I thought that was
256. 40:30fascinating. Learned that the other day. All right, let's see what we got here.
257. 41:07Look closely. There's more in the showcase. Ah, look here. Showcase submission. MD. I wonder if that's marketing
258. 41:23show. Yeah, I bet. Okay. Yeah. So, so Open AI is dropping all sorts of little hints. We'll we'll we'll just keep we'll
259. 41:31just keep scrolling here and playing. All right. Let's see. And Colossus Colos Colossus isn't even the best model.
260. 41:38Well, no. Colossus isn't the model, Danielle. Colossus one is a data center. So, I think Colossus one had 200,000 GPUs in it
261. 41:50and and then and then they built Colossus 2 and Colossus 2 is going to have more than a million. And so they basically
262. 41:58um built used Colossus one to train whatever the current Grock models are and and now Colossus 2 is I think it's
263. 42:06rolled out at this point or they might still be building it. But anyway, the the the uh Yeah, it's crazy.
264. 42:20The other thing we can ask LDG is I assume that that wiki probably has an RSS feed. So, you might be able to
265. 42:28actually call that with with AI stuff. If you want to build AI stuff to tap into that that library, there's
266. 42:34something there. Okay. Um, six planets will align and put on the show of our lifetime. Okay, that's cool, but not why
267. 42:42we're here. Reeve marks the beginning of highly controllable image generation and editing. By creating images from code and
268. 42:53translating existing images into code, we bring software level precision to anything from photo realism to graphic design.
269. 43:07I I don't know what Reev is. Reev is the best 4K image model in the world. Let's go look at it.
270. 43:27Start creating. This is probably one of those where I can't I assume I'm going to get zero credits here.
271. 43:43Try it out. See new album work. I guess I've been here before. Create um 70s. Oh, wait. I should still have it in
272. 43:58my Yeah, there we go. Do I use the right model?
273. 44:21Oh, that's interesting. Look what that's doing. It's showing you the structure of the image before it makes the image.
274. 44:33I wonder if that's part of the diffusion model or if it's I bet it is. I actually love how this is looking.
275. 44:48I mean, I love I love how it's generating. Cars look a little plasticky. They're still rendering. That one looks
276. 45:00pretty good. Still looks a little fake. Whoa. The [ __ ] was all Oh, look at this. Oh, so this is layered. Can I move [ __ ]
277. 45:20Oh, I can. Apply. Am I not sharing?
278. 45:40>> No, you're sharing. We just weren't sure what tool you were using. >> Sorry. app.reve.com. Reve No, Emmy. I found me on 112923. Oh,
279. 45:53that's so cool. app.reve.com. No, no, that's not like Christopher Reef. R Reve, not reeve.
280. 46:12Wait. So, wait a minute. So, this is actually quite impressive if I if I'm if I'm picking up what they're putting down
281. 46:23here. I'll show you something here in a second that I think we can do.
282. 46:48Okay. So, can I go into the little wheels here? Let me see. No. Okay. But I can go here and then I can open this banner text.
283. 47:05And I'm gonna say, yeah, I can edit this. So, I'm gonna say, um, sweet, sweet
284. 47:20muscle, and I'm going to say apply. So, what this is doing is it's
285. 47:35it. The way diffusion models work is you put in a prompt. It goes and looks at, you know, all of the images that were
286. 47:44somehow tagged with the [ __ ] that was in your prompt. It amalgamates them together into a single image and then
287. 47:50dnoises these noisy images. This noisy image that was sort of this amalgamation of all these images. What they're doing
288. 47:58here is they're identifying the elements that need to be done in the image. They're then writing prompts for each of those elements and
289. 48:11creating the uh the segmentation I think it's called where you can identify those things and you can prompt the elements of your
290. 48:19image individually. This is quite really quite powerful. It's really good. Mine was Kyle telling me I was in the
291. 48:27right place and to stick around. Little did we know. Ah, Silverf Fox. That's so cool. Is this Groc Imagine? No, this is Reeve. R reve
292. 48:54still better than most. Mary Carol. Yeah, I know. We're all just sweet, sweet muscles. So, it did that. That's so cool.
293. 49:02And now, so the starburst. So I can I do both of these? I bet I can highlight those two and say make these um ex make these um like
294. 49:17let's go um turtles. Let's see if that works. swapping the muscle car out for a turtle
295. 49:32hero in that same gritty factory. So, I guess I guess what I'd have to do Stop. Let me stop that.
296. 49:49Wait, how do I Oh, there's a bug here. Let me reload this. Okay. So, if I go to this banner, so I'll make
297. 50:03this banner multi-pointed starburst. So, we'll say um turtle graphic.
298. 50:18And then this one will go um instead of multi-pointed um rounded cartoon
299. 50:33explosion starburst and then we'll say apply. And so this should just change out those shapes on the banner.
300. 50:45Mary Carol, I've never heard of Reeve. Neither have I. I Well, I obviously have because I had an account here and I've got credits here.
301. 50:56Um maybe it was called something else before and now it's called Reeve. Um but this this level of of item level control
302. 51:11pretty slick and it's not bad. That's doing spelling pretty damn good. I love their diffusion um their diffusion
303. 51:27algorithm. It's It's cool looking. It's cool looking at all the structure. I'm going to the website now. Thanks.
304. 51:34app.reve.com. Yeah, that's it. It's always been Reeve.art. Oh, interesting. Yeah, it's Reeve.com now. I
305. 51:42believe Gareth told us about Reeve back in the day. Cool. Google is doing AI lender a Google is doing AI layer renderings including text and
306. 51:52segmentation work like Reeve. I don't understand how it's different from OpenAI images. You can edit those as well. Oh, I don't
307. 52:01know. It's just newer, but look, there's a turtle and there's didn't really do rounded graphic, but all right. So,
308. 52:07there you have it. You can move [ __ ] around. You can do segmented. Oh, let's Yeah, that's cool. All right, cool. Next.
309. 52:18Let's keep moving. Let's keep moving, people. We don't have a lot of time. There's a lot of [ __ ] to talk about.
310. 52:29The biggest cinema screen in the world. If you're in the trenches building AI agent systems right now, you're
311. 52:37incredibly early. If you're if you're sitting on this live, you're incredibly early. If you're if if someone heard about this live, if
312. 52:46someone's just playing with AI, you're incredibly early. Starting to be suspicious of any word any post that uses the words genuine or
313. 52:56honest. I agree with that. Magica, formerly Galaxy AI. We didn't downsize the creative team. We replaced it. Okay,
314. 53:04I don't care about that. Let's see. You're giving me stuff to talk about tomorrow on the Daily AI show. [laughter] That's good. Gareth,
315. 53:17I checked but found three references to Christopher Ree. Um, [gasps] the zipper. Do you remember the zipper?
316. 53:40The zipper, baby. Remember that crazy ride in the 70s and 80s? [laughter] The zipper, man. That's a [ __ ] death machine. [laughter]
317. 53:55Oh man, your pizza's so good. You should give it to the cat. Oh, by the way, there's a new feature in in the iOS version of the
318. 54:08X app where you can take any tweet and do a video response to it. And so, there's a lot of video responses to
319. 54:15tweets that I don't know if I like it or not yet. Scientists injected gold into eyes. Vision restored naturally. gold nano particles.
320. 54:42Huh. Interesting. All right. Starship in slow motion.
321. 54:58Oh, they did some cool quantum [ __ ] with light. Oh, so here's another this was another one that that was worth pointing out. So, Gemma
322. 55:07is the new Google open-source models and they they just launched a 12 billion parameter model. This model can run on a 16 gigabyte
323. 55:18MacBook Pro. So, like my M1 MacBook Pro that I that I stream on, I've been streaming on for it's like I don't know four or five years old.
324. 55:28That's it's got 16 gigabytes of RAM. This large language model can run on on a 16 gigabyte machine. So, if you've got
325. 55:3816 gigabytes of of VRAM on a on a on a uh a GPU on a PC or if you've got a 16 gigabyte MacBook or you know Mac,
326. 55:51this is something you can run 12 billion dense 256 256K context window built-in reasoning um vision OCR
327. 56:03user interface plus charts native audio speech translation function function calling for agents and then text plus
328. 56:12image plus audio interle runs local get started now. Oh yeah, Mustafa Solomon also for Microsoft they launched seven different
329. 56:27Microsoft AI models yesterday. So, um, Microsoft was they're doing all sorts of crazy things, but now they've
330. 56:36got their own models. We don't know if they're any good yet. I haven't heard anything about them, but they launched
331. 56:40those. So, anyway, so if you're playing around with the local model game, um, the uh, the gem, in fact, why don't I
332. 56:50let me change how I'm sharing something. I'll go share something else. Um,
333. 57:13Why did I not get my screen share window? Yeah, because I guess it wasn't running.
334. 57:37Okay. So, let me flip over to LM Studio. So, here I am at LM Studio. There's my little This is my little uh If you
335. 57:45haven't seen this, this is the Codeex Pet. And you can I I had to make a little mini version of me because my my ego knows no bounds. Um,
336. 57:57but what we're going to do here, wait, why do I have two Why do I have two embedding models running? Eject one of those bad boys. Okay. Um,
337. 58:14so here's the new model. So the the Gemma 12B parameter model and then are there oh there's quantize stuff here.
338. 58:25So I guess we'll do the quantise 8. Okay. So so what you're looking at here I don't know if you can all see this. So
339. 58:32I'm in this thing called notebook LM or not notebook LM sorry LM studio. Let me make this bigger. Hopefully can you still see this
340. 58:41Brandon? You can still see what I've got going on here. You see codeex. Oh,
341. 58:57okay. Hang on. Wait. You don't see you don't see LM Studio here. Okay, you do. Okay, got it.
342. 59:12Okay. So, this thing, this is just the codeex pet. It just it basically sits on top of all the other apps. So, so, but
343. 59:22what you're looking at here is not Codex. What you're looking at here is LM Studio. What was the app that I was doing before
344. 59:30I moved to LM Studio? That was Reeve. We We all need Kyle Codex's pet on the screen. I I'm It makes me very happy.
345. 59:40The way this thing works is once it's set up and it displays correctly. Like on my M1, it wouldn't show up, but this
346. 59:46is we're looking into my M5 right now, so it shows up there. And when you're using codecs and you're talking to it,
347. 59:53this thing like gets a little talk bubble underneath it that says what the what the thing's up to. So anyway, um so
348. 1:00:01in this mode that you're looking at right here, um I've got Quen 3.6 six 35 billion parameter model loaded and I can
349. 1:00:11just, you know, talk to an LLM just like chat GPT like um give me a sonnet about bananas.
350. 1:00:27So that's pretty quick. Upon the vine and tropical light, they hang a curved crescent. the emerald shade. Their peel
351. 1:00:34is bright, vibrant, and range. Before the earth, its heavy fruit is laid. So, you know, it knows what a sonnet is. It
352. 1:00:40can write it. It's pretty quick. Um, and then this developer page, this is where I'm running. These are the models
353. 1:00:51that are running. So, instead of me using the interface like chat GPT, these are now running as a server. You see it says
354. 1:01:00server settings at the top here. And I've got the Quen model running. And then I've got this text embedding Nomic model
355. 1:01:09that allows um my Atom, my OpenClaw um instance can run all all of its compute is being done locally on
356. 1:01:20this machine. It's not doing anything up in the cloud. And then there's this search thing here and I can go get new
357. 1:01:27models. And so um so this Gemma model, this is the new Google model that's small and fast. And then you see the download options. There are
358. 1:01:39an a 4K quantized model, a 6K quantized model, and then a an 8 zero quantized model. And you'll notice that this one's 7 1/2 gigabytes, this
359. 1:01:54one's 10 GB, and this one's almost 13 GB. So, the reason that you would get smaller versions of this is depending on
360. 1:02:03your RAM. So, if you've got a 16 gigabyte machine, your operating system is going to take up some of that 16
361. 1:02:09gigabytes. And then you have, so let's say you've got 12 gigabytes left. If you put a 12 gigabyte model or even a 10 gigabyte
362. 1:02:20model that might not run very well, but you could get this, you know, 4K quantized version that will live in in
363. 1:02:27that 12 free gigabytes. So that's part of what you look at when you look at these models. One is the model name
364. 1:02:34itself and then one is the quantized versions of them. So I'm going to go grab the this version of that one and then we'll
365. 1:02:41we'll just go look at it. But anyway, so if you're playing in this in this uh in this local large language model game,
366. 1:02:53if you're playing with LM Studio, um that's a new model worth looking at. Okay, so let's go back here. Gemma 4. Get out of that post.
367. 1:03:17In the race between problems and solutions, solutions are win winning. Peter Diamandis, this this guy's sort of
368. 1:03:24the ultimate optimist for this AI stuff. Um,
369. 1:03:43Elon Musk. This is made by Grock. Imagine 1.5. >> Trojan ground.
370. 1:03:57For 10 years, I have buried my own heart beside the bodies of my brothers. [music] Today, I bury that grief in Trojan ground. [music]
371. 1:04:13>> Listen to me. >> We will not yield. [music]
372. 1:04:32There you go. Not bad. Still a little plasticky looking, but um I want to get to the news that's related to that 4.8. Let's see.
373. 1:04:55Wait, what's this? This looks like a cool chart. Local bandwidth, local AI hardware capacity times bandwidth time software stack.
374. 1:05:27Huh? Yeah, it's complicated. You got to be a nerd to play with this stuff right now. All right. Um,
375. 1:05:44it's fascinating watching the generation visualization on Reeves interface. I agree with that. That's pretty cool. Um,
376. 1:05:51looks good for trailers. The problem would be your feature would have to match the trailer. We are nerds. Okay. So here
377. 1:06:06this is excited to share the first short brand film for Codeex. Catch it airing during gave game one of the NBA finals tonight.
378. 1:06:18Lost your share. Really? Oh [ __ ]
379. 1:06:33Okay, so this came out today and I apparently this is airing on the uh during the finals
380. 1:06:48tonight, the NBA finals. >> You going to come out with us? >> Namaste. >> All right. Have a good night.
381. 1:06:59It's past [music] bedtime. Off you go. >> This is mission control. >> Actually, it's a good idea. >> You are cleared for launch. GCMC.
382. 1:07:12>> Steep tight. >> Ignition sequence started. >> Let's build a companion app. >> The code awaits your command. 2 1
383. 1:07:28>> engines full power. >> You are about to enter the edge of all possibility. >> Initiate second stage.
384. 1:07:44>> We'll be listening for signals from the other side. From the limitless worlds you awakening,
385. 1:07:58>> we wish you courage on your journey.
386. 1:08:20H wild. Um, there's a couple of interesting things in here. Let's go look [snorts] >> full power.
387. 1:08:48Use imagine to create spite sprites for the planet. So it looks like it looks like maybe a new version of codeex and then
388. 1:09:02so these all look like components and it looks like there's some sort of game or simulator whatever she was building the
389. 1:09:08launch dashboard or whatever she called it. So it looks like it's got pretty cool user interface kind of stuff that it can do.
390. 1:09:18And then there was another thing I saw here that someone pointed out that I thought was actually a pretty cool
391. 1:09:25little Easter egg. The other side. So these paper clips floating right here. So, one of the this is a bit of a
392. 1:09:43a [ __ ] you to the to the doomers [laughter] where if you give if you give um I forget I does anybody know where the the
393. 1:09:54paperclip problem came from? So when they talk about um AI alignment, one of the things that they talk about is is
394. 1:10:02the paperclip problem that if you optimize an AI to just make paper clips, it will ult ultimately kill all of
395. 1:10:10humanity because it just wants to keep making more and more paper clips and and it will it will optimize for just that.
396. 1:10:16And so, so this is probably some sort of, you know, [ __ ] you to that that what's he got going there? I don't know.
397. 1:10:27Some sort of dashboard to the >> from the limitless worlds. You awaken. >> We wish you courage on your journey. [music]
398. 1:10:54That's a pretty good ad. I mean, I think most people are not going to know what the [ __ ] it means, but you know. Um,
399. 1:11:02the other thing, let me go show you another thing that's in codeex right now.
400. 1:11:21All right. Did that download using a new chat? Pick a model. Gemma 12B context window. Crank it. 131,000
401. 1:11:38tokens. Load model. All right. So that's Did it load? Failed to load. Why? Download completed. Load model.
402. 1:11:50Failed to load model. Copy error to clipboard. Oh, I probably don't have enough RAM. Oops.
403. 1:12:07Wait a minute. What happened? sharing stuff again
404. 1:12:25and I restarted before before this here live. Okay. Pick a model. Gemma 4.
405. 1:12:39We'll go 64K. Load model. All right. Whatever. Um, but let me show you something going on in codeex here. Am I in codeex?
406. 1:12:55Yeah. So in codeex there there one of the things that open AAI announced today or was it yesterday? It doesn't matter.
407. 1:13:11Um so there's there's these plugins that you can install in codeex. So what codeex is if you haven't played with it
408. 1:13:18it's kind of like chat GPT. you're like, "Hey, go do this thing for me." But it can it can go off and assign itself
409. 1:13:26tasks autonomously without you having to do it. And it can use tools. And so like the first tool here is
410. 1:13:33computer use. And this other tool is Chrome. It can open a Chrome web browser and surf the web. It can use Google
411. 1:13:40spreadsheets and and presentations. Today they added data analytics, product design, creative production, sales, investment banking,
412. 1:13:52and public equity investing. So these six um they're not even really skills. They're they're like collections of skills.
413. 1:14:05So, I was saying earlier that um that I actually thought it was a pretty big deal that some new smart glasses came out.
414. 1:14:25So, here's Monaco Glass, the world's first wearable Linux computer in glasses form. run clawed code, codeex, and any coding agent
415. 1:14:37anywhere. So, they're positioning these glasses not as you can do cool [ __ ] and you can take pictures of
416. 1:14:47people, but as if your coding agent is off doing your coding for you and it's running independently. Oh, wait. Am I not sharing again? No,
417. 1:14:59wait. Am I sharing wrong? No, this is just a general note that you should use codecs to fix your screen sharing issues.
418. 1:15:09>> Oh, yeah. Yeah, I should. Um, all right. So, let's let's watch this video. So, so this is the the Linux computer in glasses form.
419. 1:15:28In the age of artificial intelligence, building has never been easier. I used to grind at the keyboard, but nowadays I
420. 1:15:35just tell the computer what I want and the AI builds it for me.
421. 1:15:51Introducing Monaco Glass, the world's first heads-up display smart glasses designed for Vibe productivity. For
422. 1:16:00running your cloud code and code >> Wait, can we just can we have a hype check, please? For for Vibe productivity. [laughter]
423. 1:16:09Listen, LDG, going into the the What is the date today? The June 3rd, 2026 AI learning lab transcript is Vibe
424. 1:16:18productivity. You heard it here first. Oh my god. >> Designed for Vibe [music] productivity for running your cloud code and codeex
425. 1:16:27terminal and for vibe coding hyperpersonalized apps that fit into your [music] everyday workflow. Now, in
426. 1:16:34order to build in the real world, we had to solve the problem of input. We've integrated a bone conduction microphone.
427. 1:16:42It sits on my nose and listens to vibrations coming from my nose bone. If you're at a rave party or a very loud
428. 1:16:54coffee shop, my AI listens to me and only me. >> That's actually really slick. So, you know how they have bone conducting
429. 1:17:02earphones? This is if you're in a noisy environment, microphones are bad. So, >> chop, my AI listens to me and only me.
430. 1:17:10Hey Monaco, do a deep research on this topic and draft a PowerPoint presentation. Make sure you visualize trends with data and graphs.
431. 1:17:21We've integrated a vision engine. So when I raise my hand, the menu shows up. Can choose the app I want. Let's try
432. 1:17:33music. Can tap it. Go to the bottom of the playlist. Play this song. I can reduce the volume and increase the
433. 1:17:41volume again. But the big story is actually hyperpersonalized software. Now imagine you're a student. You're sitting
434. 1:17:49in a classroom and your professor is writing math equations on the blackboard. You can simply say this generate.
435. 1:18:11Wait, hang on. I think I think we have to jump over and and celebrate Chef Kelly. Oh, I wouldn't look cra
436. 1:18:19crazy swiping in the air all alone like that in a coffee shop. Did do you remember the what was the Tom Cruz movie
437. 1:18:27where he's like doing all this stuff and we made fun of him? Yeah, like you're going to do that someday. This is the day. [laughter]
438. 1:18:35Sound. Yeah, let's go. We'll go finish this up. Let's look at this uh rest of this video. I assume they've got sound.
439. 1:18:41It looks like the back of the ear earpieces are are big. There's a music app. You'll have to stop every 45 seconds
440. 1:18:52though and uh restart the screen share for the glasses. >> Good lord. Good lord, Macintosh.
441. 1:19:06Delete it. Delete it. Delete it. Delete it. Delete it. Delete it. Delete it. Delete it. Delet it. There are thousands of ways of how you
442. 1:19:19can integrate Monaco glass into your actual day-to-day work through the operating system that's running inside
443. 1:19:25of the glasses in the cloud and on your personal work computer. We believe in the future all types of
444. 1:19:39productivity will move to a form factor that's as small as a [music] regular pair of glasses. And it starts here
445. 1:19:47today with Monica Glass. Hope to see you there. >> Okay, they didn't mention sound and it looks like this is probably a laser
446. 1:19:55display. And notice how it's only green. Um, my suspicion is this is to do this in full color would require three lasers.
447. 1:20:07Um, and that probably can't fit in this form factor is my guess. There's got to be a better way. Yeah. No, I agree. I think this is this is
448. 1:20:16kind of crap, but this is the first time this use case of using these glasses as an interface for vibe coding. Like
449. 1:20:24rather than having to pull out your phone all the time, if you just were wearing glasses and then it would goz, you know, your shit's ready,
450. 1:20:34only one color, forget it. I'm I'm uh it it may or may not have one color, but nothing that they showed here was
451. 1:20:41anything but green. So I am I am chat TMZ speculating that the reason it's green is that they had to use a single
452. 1:20:49laser and not three different lasers. And also I think blue lasers wasn't blue yet. Yeah. Um blue lasers were the the
453. 1:20:58most recent one that were really they're still really expensive. Um let me let me look up another thing here. Um Scorsesia
454. 1:21:05I want to talk about this uh before we go. So, as you know, oh, score se as you know there's been a lot of um
455. 1:21:29am I still sharing? Yes. There's been a lot of anti-AII sentiment recently and a lot of that is coming
456. 1:21:38from the film industry and artists and things like that, but the film industry in particular really really loud and and
457. 1:21:45sort of increasingly just angry and angry and angry. Um, and this came out today. >> So, you think Flux actually will storyboard for me?
458. 1:21:54>> Well, let's try it and we'll go from from what you think. >> I need a place that doesn't feel modern, a town, not a village, not a city,
459. 1:22:01almost medieval. Even the the streets are narrower, cobblestoneed. The main road through the town is twisting and
460. 1:22:08turning. Put the camera higher looking down. Deill would have his production designers do oil paintings. This is that
461. 1:22:16in a sense conveys a cinematic a cinematic intelligence. So you think flux actually will storyboard for me.
462. 1:22:24>> Well, let's try it and we'll go from from what you think. You need a place that doesn't feel modern. A town, not a
463. 1:22:29village, not a city, almost medieval. Even the the streets are narrower, cobblestoneed. >> Okay, we already saw it. Okay, so um so
464. 1:22:39[clears throat] Martin Scorsese Scorsesi is arguably the most famous popular filmmaker, you know, in history, living
465. 1:22:48filmmaker in history. He's 82 years old. He has been through all of the technological changes, right? He's been
466. 1:22:54he's been around for for ages. He he you know has gone from film to digital to nonlinear editing to special effects and
467. 1:23:04you know digital special effects and things like that. I noticed an immediate an immediate split in the anti- AI crowd because of
468. 1:23:17this video. So he's he's an adviser for Black Forest Labs and which which does AI video and he's basically saying I'm
469. 1:23:25just going to use it for storyboards. He's going to use it for storyboards until he can use it for film, right?
470. 1:23:30He's a storyteller. Basically what he said is um in the end it's about storytelling and we use all sorts of
471. 1:23:36different technologies for storytelling. I saw an immediate split in the anti-AII crowd around this where one was one was
472. 1:23:44basically like, you know, [ __ ] him, you know, that's really disappointing. You know, they're sort of sticking on the
473. 1:23:50bandwagon. But but there was another group that was basically like, ah, [ __ ] I guess we got to deal with this now,
474. 1:23:57right? And listen, I under I understand the sentiment of being very frustrated that everything's about to change. Um,
475. 1:24:08but this seems significant to me today. And I think what we'll start to see is we'll start to see things like medical
476. 1:24:16breakthroughs, people that we respect in in specific industries where you're like, well, that industry is never going
477. 1:24:22to do it. And then someone in there will do it. Um there there have been four or five decently major filmmakers, Jim
478. 1:24:28Cameron probably being the most famous, that have have been kind of pro- AAI, but Marty Scorsesei is, you know, on a different level.
479. 1:24:38And so if he's basically embracing AI, then, you know, that's going to win over some people. I don't think it's going to
480. 1:24:45stop the vitriol, but it's [laughter] gonna um I think it is going to make a difference. Um, Bradley [clears throat]
481. 1:24:54Collins, if you're a large developer, so this is about the glasses, I think. If you're a large, if you're a dev on a
482. 1:25:00large repo, this isn't going to get you the results you need. Vibe coding large code bases requires, in my experience,
483. 1:25:05detailed prompts to avoid drift and land changes without tons of debug. I agree with that today. And I also like some of
484. 1:25:13the comments about these glasses being not the right form factor. If they're not in color, I'm not interested. [cough and clears throat]
485. 1:25:21I think we're probably two to three years out from these [clears throat] agentic platforms being truly seamless, but
486. 1:25:31we're only two to three years out, right? So, if you're doing some sort of informal um some informal development, um these might be useful.
487. 1:25:43They're probably not going to be useful for the next year. Um, but then at some point there's going to be a tipping
488. 1:25:49point just like Opus 4.6 was a tipping point that said, "Oh, this this thing's good enough now to do a bulk of my code.
489. 1:25:57It's going to be the same with these with these platforms." Scorsese's also seen Kodak and Blockbuster come and go. Yeah, exactly. Exactly.
490. 1:26:06Um, all right, we are at the top of the hour. [clears throat] We're going to get something from OpenAI tomorrow. that
491. 1:26:14that commercial kind of made it seem like Codeex is going to remain a standalone thing. I have a feeling at some point they're just going to
492. 1:26:20collapse all these apps together because it doesn't make sense to me to have a chat GPT browser web browser chat GPT that you can look
493. 1:26:33at in a browser and then codeex which is this standalone app that does all this other stuff. It just seems to me that at
494. 1:26:39some point those have to come together. Um, tomorrow is the AI Salon uh mastermind practice lab. So, if you're in the
495. 1:26:50mastermind or you want to join the mastermind, come to the practice lab. That's tomorrow at 10:00 a.m. Um, not
496. 1:26:56tomorrow. Yeah. No, tomorrow. Yes. Uh, tomorrow, Thursday at 10:00 a.m. is the mastermind practice lab. And then on
497. 1:27:03Friday, we've got office hours uh at 11:00 a.m. Mountain. And then we've got Friday night date night where it'll be
498. 1:27:10an ask me anything night. Depending how big a deal the Open AI announcement is, it might be a combo of AMA and let's go
499. 1:27:22look at at Open AI. Fair for Scorsese, but what about the kids? What about the kids? You mean young filmmakers?
500. 1:27:36they have access to the same tools he does. I I mean just like you know video cameras democratized film making and then
501. 1:27:47YouTube democratized distribution, right? You could certainly argue that the digital video camera and YouTube
502. 1:27:57are a bigger disruption to Hollywood than you know than anything that came before it. Right? So AI is just another in the path.
503. 1:28:06Um, let's see. What's this? Chef Kelly. This man was making cinema on film. Canisters, lenses, heavy lights, and
504. 1:28:14booms. Editing by hand. Locations. Locations. If the man wants to use AI, he's earned it. [laughter] Um, anyway, thank you for hanging out
505. 1:28:27tonight. I hope this was valuable and useful. Um, pay attention tomorrow. It It looks like we're going to
506. 1:28:36get something. I think they usually go noon, 11:00 a.m. uh Pacific is when they generally make their announcements. So,
507. 1:28:44we'll see if we get something tomorrow. If they ran that commercial on the finals tonight, uh that's probably uh probably not
508. 1:28:54coincidental timing. Beautiful. Thank you, producer Brandon. Thank you all for being here. and I will see you
509. 1:29:02tomorrow if you're in the mastermind Friday if not. Peace out.