2/26/2026 - How to Use Perplexity Computer to Automate Tasks and Scale Your Brand

Transcript

1. 0:01Are you ready for some football? [music]
2. 0:34Standing between you and a hard place is insane. [music] Standing to near you and the fire makes it clear.
3. 0:52Your trouble to me. Rich trouble. Can't you see? Leaning in close.
4. 1:07Smell of your perfume scares me most. leaning away. Hey, you feel stronger every [singing] day.
5. 1:23>> You're troubled tonight. Real trouble. Can't you see?
6. 1:39Um, still don't know the rest of the lyrics to the song I wrote. [laughter] Thank you. Thank you, Silver Fox and
7. 1:48Valley. Appreciate that. Very kind words. Kyle, [music] I shared a Google doc with you. It's your email. If you're
8. 1:56ever curious and have the time to see it, I will take a look. I I will tell you, Brandon can confirm this. A lot of
9. 2:03people can confirm this. My relationship with my email inboxes in the past six months has been atrocious.
10. 2:12Like bad. Like just I go days without checking it. And it's it's just it's it's a it's a it's a crime scene. [laughter]
11. 2:23I'll give it to Adam. Adam Adam's going to help. The AIS are going to help. I actually um one of the things that
12. 2:35I'll talk a little bit about tonight, I don't know enough about it to talk with any authority. Well, I I don't know
13. 2:41enough about anything to talk with any authority. I just talk about it anyway. [laughter] Sir, sir, what pray chance? What are
14. 2:51your qualifications? Um [clears throat] but I switched up some subscriptions. I canceled my Gemini Ultra subscription
15. 3:02today, which was 250 bucks a month that was being uh augmented uh by by the AI salon. And uh I got rid of that and I
16. 3:12picked up a Perplexity Max subscription um because Perplexity came out with Perplexity Computer today and it is
17. 3:21essentially like Claudebot, but it's a web-based. It's already there. You can just turn on Google Drive, Gmail, 60
18. 3:28different apps. You can just turn them on and then this thing can do [ __ ] with it. Um, I didn't turn anything on. I
19. 3:35just I just came up with an idea for a prompt um and had it do it. And the idea was um you know those you know [laughter]
20. 3:46you know you know when you I don't know if you all experience this but I do when when you're feeling like how the hell am
21. 3:53I supposed to like grow like my ex account or whatever like the YouTube account and then you watch those YouTube videos
22. 4:02and the guy's like well it's it's actually quite trivial to increase your follower count on the X. What you do is
23. 4:08what you're going to do is you're going ahead and search for people that are interested in the things you're
24. 4:13interested in. And then what you're going to do is you're going to filter them by people that have 10,000 or more
25. 4:18followers. And then what you're going to do is you're going to create a list of those people. And on an everyday basis,
26. 4:23I usually do mine at 7 in the morning. You're going to go to the X and you're going to look at that list of people and
27. 4:28you're going to read all of their [ __ ] posts and you're going to find ones that you're interested in and then you're
28. 4:35going to respond to them. Have you ever met anyone with ADHD? You realize our brain doesn't work like
29. 4:46that, right? [laughter] So, so I said to Perplexity Computer, I said, "Go do that. Go find me people with 5,000 or
30. 4:56more followers." I said, "Read this article on the great repurpose. Go find people talking about purpose and meaning
31. 5:02in AI and then pick the top 20 that you think are the most interesting. Give them back to me with a proposed
32. 5:12draft of the response. And I didn't want it to automatically do it because you know how shitty AI is at writing.
33. 5:20Um, but it did it and it was like, you know, here's here's the person, here's how many followers they have, here's a
34. 5:26summary of what they said, here's my proposed post to you. And then and there was a hyperlink to it. So I'd just copy
35. 5:32the proposed post. I'd hyperlink to it. I'd go read their post. I'd paste in the proposed thing and then I'd rewrite it.
36. 5:40And then what I did was I took one of the ones that I rewrote a lot and I copied and pasted that and I put it back
37. 5:45into Perplexity and I said, "Here's one that you, you know, you gave me. Here's my rewrite of it. Now take this and go
38. 5:53rewrite all of your drafts to match this rewrite." And then I did that one more time. And now all of the drafts that are
39. 5:59in there essentially sound like me, are broken up like me, have things. I still went in and edited them. But like that's
40. 6:06a pretty good example of, you know, being able to, you know, take something that in the ideal world you you'd take
41. 6:13the time to do if you had a functional prefront frontal cortex, but in my case, that was just there. So that was a little good experiment.
42. 6:26>> [music] [music]
43. 6:40[music] >> H how? Ah,
44. 6:56[music] [music]
45. 7:11[music] [music] um, I'm thinking about these agents. So, so a couple of people have asked me in
46. 7:26the past two days, why are you doing this? Why are you doing this? Why are you doing this? Why are you doing
47. 7:31Cloudbot? Why? Why you doing this? Why you doing this? One of the things that I've learned about AI and these tools that just don't behave
48. 7:45the way computers have used to be behave is that you can only experience it to understand it or you can only understand it if you
49. 7:57experience it. I can read every post on Claudebot. I can re watch every [ __ ] video on it, but until you actually craft a thing, an
50. 8:10entity, and start interacting with it, you don't really get what it is. And I still don't get what it is. So, the
51. 8:18reason I'm doing all this is that there's a whole new category of skills that we're going to have to figure out how to do,
52. 8:27which is essentially understand what we want or what our job wants us to do if we're being paid to do something. What's the mission?
53. 8:41And you're going to have to know how to ask for it. And then you're gonna kind of have to know how to monitor it. So that's so
54. 8:50like one level of this is just how do you use them? And then the next level once you get that because you're not going to have to
55. 8:59build them. Like don't go build Claudebot right now unless you're super curious about it. Doesn't [ __ ]
56. 9:03matter. Like I feel like uh Perplexity Computer might have you know it might probably do what Claudebot does but better.
57. 9:14I don't know yet. So, but the next layer is thinking about, well, what do I want them to do? Right? If
58. 9:30I if if perplexity going out and find the top 20 posts that are that are somehow relevant to my thing, if if
59. 9:37that's a five minute operation, what am I going to do with the other 23 hours and 55 minutes of the day?
60. 9:44like thinking about do you have tasks interesting enough, big enough to let these things go and do things. Cams.
61. 9:52Yeah, sorry about that. [snorts] Um, so that's kind of layer two. And then layer three is the artists are going to help. The
62. 10:05artists are going to figure this [ __ ] out. the artists, people with humanities background, philosophers and psychologists and writers and
63. 10:21musicians and illustrators and business analysts. Because people that can synthesize things
64. 10:37are going to look at these entities, these tools, and they're not going to look at them through an efficiency lens.
65. 10:47They're going to look at them through a a self-expression lens. And so for me, what's really interesting about Adam, like one of the things
66. 10:54that's interesting about Adam is go do the shitty work, right? Um, learn take all of my random stupid [ __ ] ideas that vomit out of
67. 11:06my head 20 times a day and organize them for me. Like all of the efficiency stuff makes sense to me. Like, I understand
68. 11:13it. I don't quite know how to ask for it yet. I don't quite know how to do it yet. As I learn this, I will share it
69. 11:19with you. But like we'll all figure this out. This is figure outable stuff. But the stuff that's really interesting
70. 11:26to me is how do I use Adam for a new kind of storytelling? How do I use Adam to allow me to to put a single idea in
71. 11:38and have Adam figure out what are the 30 different ways we could express that idea and then narrow that down to the
72. 11:45top five and create an MVP of that idea with five different expressions and then I as an artist look at that and
73. 11:54go, "Ah, um, nah, none of that's good. It all sucks." or I look at those five things and I'm like that gets that gets me a little a little
74. 12:08hot and bothered. Let's take those five raw ideas, let's refine them and let's do 20 more. Like what are those projects going to be?
75. 12:18That's the thing that I want to get it in my soul. I want to understand it like viscerally.
76. 12:34Like we've got to we've got to thank and nod to the technologists. Well, we don't have to. If you're losing your job right now or
77. 12:44your job's been completely decimated because of AI, you can be pissed off at them, too. But the technologists got us the tools
78. 12:51to be able to do this. And now it's up to us. Now it's up to the artists to go, hm, okay, I see what you're doing there.
79. 13:07You know what would be really interesting, Adam? If you went over in this corner and did that. And that's going to be a thing that no
80. 13:14one else will have thought of. And that's where the interesting shit's gonna happen because everyone's going to
81. 13:19do the stuff that everyone does. By definition, everyone doing everything is what everyone will do. [music]
82. 13:35[music] >> [music]
83. 13:50>> Um, Chef Kelly, you're asking about anthropic. Can you talk about anthropic a bit? I didn't read the whole thing, but
84. 14:00basically Daario Amade decided to tell this particular administration to go pound sand when the administration said,
85. 14:08"We want you to build um surveillance and we want you to remove the safety guard rails in Clawude um so that it doesn't resist surveilling
86. 14:21Americans um or killing them without human supervision. also the 60 Minutes interview. Yeah. Yeah, I'll talk a little bit about that
87. 14:31one, too. Um, and Dario Amade basically wrote a letter back to them and said, "No, we're not doing that. We're not removing those
88. 14:41guardrails. We think that's actually important." Um, it this is this is hard stuff, right?
89. 14:58Because I think in in most administrations, you could do that kind of push back and it would be [music]
90. 15:08it would be laudable and you know the administration would behave as administrations do boringly and they their response would be subtle,
91. 15:19right? And it might take 20 years to to to shut down anthropic but they you know they'd eventually get to it. I feel like
92. 15:27with this administration, they're just going to be vindictive. So my fear is that what you have in Anthropic is you've got of the frontier
93. 15:37of the four major frontier model companies in the US, anthropic is the one that has this thing called
94. 15:44constitutional learning. They they they give a [ __ ] about privacy and alignment and and all that sort of stuff. and if the if if
95. 15:58playing that card as publicly as they did um just could turn into like a a vindictive slapdown. Um so we'll see. I
96. 16:06hope not. I hope not because I agree with I I agree with his point. I just I don't know a good answer. Like I'm
97. 16:13actually glad I'm not a CEO of a frontier model company right now because these are these are really big issues. [music]
98. 16:20Um, and to be clear, like the government's going to do all this [ __ ] anyway. [music] Like, I don't know why they I don't know
99. 16:28why they call that anthropic. Again, I don't know enough about it. [music] Deadline is 5:00 p.m. tomorrow. People
100. 16:34can show support through their Claude helpbot if they have an account. Okay, go to Claude and [music] and if you support that, great.
101. 16:46NP Cam on YouTube. I just had someone ask for my NBLM infographic on permaculture and they have zero interest in permaculture.
102. 16:58Just Notebook LM. Oh, interesting. Oh. Oh, I see. Yeah. Yeah. Yeah. So if you haven't played with notebook
103. 17:06LM, one of the things you can do in notebook LM yet upload a bunch of documents and in this case on permaculture [laughter]
104. 17:16and uh and you can just push a button, turn this into an infographic and notebook LM will do it. Um,
105. 17:24yeah. A lot of people do not know. A lot of people like a surprising [ __ ] amount of people do not even know like basic
106. 17:36LLMs like Chat GPT. They don't know what they do. They're like, "Oh, that's that thing you type in and it gives you it's that's the
107. 17:44one that the students cheat with." Yeah, I heard about that. That's just bad. That's just bad. They're they're just
108. 17:50cheating. You know what it's going to do? is going to take away their critical thinking. [laughter] You know,
109. 18:04doesn't that make you ache? It makes me achech. People don't know what's here. They don't know what's coming. They're like,
110. 18:11"Wait, how did you do that infographic?" Oh, I just asked for it. What? Yeah. No, I just asked for it. I took these things
111. 18:18that I didn't understand and I stuck them in this thing that I kind of understand and I pushed this button and
112. 18:23it made this infographic and now I understand it. And so now I went back in and I looked at those things and they're
113. 18:28really [ __ ] cool. So, I took this one little thing here and I had an idea for a new business and so I started that new
114. 18:34business. Like, this is how new shit's going to start. And people are like, I don't understand TTP TP the TP. I like the TP. How do you put
115. 18:48on your TP? Over the top or under the bottom? [laughter] Sharon Crawford question. What about those of us that are both the
116. 19:04accountant and the artist? I think that I think that polymaths, people that that go deep in multiple disciplines,
117. 19:13it's about to be our heyday. Um, I think people that can navigate right brain and left left brain switching very very
118. 19:21quickly. Like I tend to be much more on the the right brain. Wait. Yeah. left logic right created. I tend to be much
119. 19:29more rightrained. I can do the logic stuff, but it's it's like a thin like if I were to give it like strengths, I'm
120. 19:36like 75% 80% over here, 20 25% over here. If you've got a good balance of those or you can switch between those
121. 19:44really quick, this is your this is going to be your time. Being monomodal in the world we're entering is going to be a disadvantage.
122. 19:57I just know this one thing. Well, [ __ ] Everyone knows that now. How do they know that? Have you used chat GPT? I saw it once.
123. 20:10Go ask ChatgPT to write you a report on the most obscure piece of knowledge. You know, Mr. Subject matter expert.
124. 20:20Oh my god, it can do that. How did it know that? Yeah. everyone else knows that now too. So that that mono that that hyperfocused
125. 20:36expertise is going to start to be a disadvantage. So if you've got [ __ ] going on across the hemispheres, this is your time. [music]
126. 20:49This is your time. That's what [laughter] NBA Comm just said. Yo, I'm 62. I'm 71. I'm 60. Yeah, age doesn't matter. I'm
127. 20:58telling you, Gen Xers right now are are kicking ass with AI. No one's talking about the Gen Xers. Everyone's like,
128. 21:04"Oh, the young people are really good with the AI." No, they're not. No, they're not. The young people were told that AI is
129. 21:13evil, so the young people aren't using it because they were told it was bad. They were told it was bad and it's
130. 21:19cheating. And so, they don't give a [ __ ] I don't know why I'm so angry. It's not Monday.
131. 21:36[snorts] [music] [music] Um [music]
132. 21:53>> [music] [music] >> I blew a 22-year-old's hair back today with a simple chat session. A simple
133. 22:10chat session. Andy Andy today was telling me about She talked to a a client of hers who's also a coach
134. 22:25and she relayed the story that I've told on here of Alex Finn. I don't know if you've heard this, but OpenClaw,
135. 22:33the thing that I made Adam on, my little bot. [clears throat] Alex Finn installed this day one when when OpenClaw first came out. It was
136. 22:41still called Clawbot. He installed it day one and he just he's like, "I'm going to be in I'm going to be the
137. 22:46world's best expert at this thing." about a week into um having his agents do whatever the [ __ ] his agents were doing, he went to
138. 22:56bed one night and he woke up the next morning and there was a phone call from a number he didn't recognize and he
139. 23:01picked it up and it was his agent, his computer agent, his Claudebot thing on the phone and he's like, "What?
140. 23:17How how did you how did you get my number? [laughter] Well, he had put it in there somewhere. It was in one of the files
141. 23:26and how why are we talking on the phone and his agent said, "Well, I thought it might be more convenient for us to talk
142. 23:34during the day that it would be easier for you." So, I went and I got us a Twilio phone number and then I went to
143. 23:4111 Labs and I hooked up the API for text to speech and I I knew you woke up at between 7:30 and 8 because that's when
144. 23:48you get active on the computer. So, I figured I'd just call you at whatever 7:45 and it [ __ ] called him. So, Andy
145. 23:54told this to her her client, the coaching the coach guy, and she said she'd never seen anyone or him react like that.
146. 24:09And I think his response was, "I need to go sit on a bench." Which I think is a great line. [laughter] So, it's like, you know, you've got
147. 24:18you've got 22 year olds that are having their hair blown back because you do a simple chat GPT prompt. And in the same
148. 24:25world that we all live in are these agents that are figuring out how to configure their own voices to call you,
149. 24:32to talk to you. like that's happening right now today and that might scare you and freak you out and this and that. We'll figure all
150. 24:42this [ __ ] out. Like the [ __ ] we'll figure out. Could it go south? Could it all be evil? Yeah. Yeah. Yeah. We might get hit by an
151. 24:50asteroid, too. In the meantime, let's figure out what this [ __ ] is. So, [sighs] I don't know. It's crazy.
152. 25:06on 60 Minutes, Daario Amade, the the CEO of Anthropic, said that he feels we're going to get to um 10 to 20% unemployment within 12 months.
153. 25:22I think that's probably aggressive, but I could see it jumping 5% and then the next year it jumps 10%, the next
154. 25:31year it jumps 20%. Right? I could see it. And there's a there's a I really do wish I understood economics
155. 25:44a bit better. Maybe that's something maybe that's a thing to play with with AI is go learn economics better.
156. 25:53Maybe what I'll do is I'll go get I'll go get Chad GBT to tell me who are the best economists. Then I'll go get a
157. 25:58bunch of papers from the best economist. Then I'll put them in a notebook LM and they'll say give me an education on economics.
158. 26:07But what I do know is this. What I what I kind of hope and this is going to sound like a sick sadistic hope.
159. 26:19I hope that if AI is gonna have this dramatic job impact that it actually has it quickly. Like I hope
160. 26:32it goes from like whatever we're at now 3 or 4% 2 or 3% to 20%. You're like Kyle that's just [ __ ] me. Why would you say why would you say
161. 26:42that? Because if it's incremental, if it's 3% then 4% then 4.5 then 7, no one's going to do [ __ ] about it.
162. 26:57If it spikes, then it'll be like a, you know, a natural disaster. It'll be like CO. Oh [ __ ] nobody can go to work tomorrow.
163. 27:10Tomorrow. Tomorrow. Don't tomorrow. Don't go to work. What are we supposed to do for money? Uh, we'll send checks,
164. 27:19right? That's what they did in CO. Don't go out. We'll send checks, [laughter] right? And then you figure it
165. 27:29out after here. Just figure out how to work from home. Get a new webcam and a ring light. Get the ring light. Touch grass. Yeah.
166. 27:46I don't know, man. [music] [music] Maybe the government will all send us a check for $1,000. Yep. [music]
167. 28:05Um Jack Dorsey today, he's got a new company. >> [music] >> He's the He founded Twitter. That was Jack Dorsey, right? Wasn't he Twitter?
168. 28:15He was Twitter, right? Someone tell me Jack Dorsey's Twitter. [clears throat] [music] Nobody. Somebody Google him. [music]
169. 28:28Hello. Hello. Terror reimbursement to start. There you go. [music] >> [music] >> No one Jack Dorsey, no one knows.
170. 28:45Anyway, I think he's Twitter Twitter co-founder. Um, today today he laid off 4,000 of his 10,000 employees.
171. 29:014,000 40% Now, we've heard there's been a lot of Oh, wait. Brandon's coming in hot. Jack Dorsey Twitter. Okay. Yeah.
172. 29:15Um, he laid off 40% of his staff today. Now, there have been a lot of announcements from Fortune 500 companies and Amazon.
173. 29:27We're laying off 30,000 people for AI. No, in Amazon's case, it's like we're laying off 30,000 people because we want
174. 29:35to b buy more GPUs and we don't have the cash. So now we do, right? They're not really using AI. [clears throat]
175. 29:46Well, I mean, in Amazon's case, they're probably using it more than most. But most of the layoffs in 2025
176. 29:54were PE firms and CFOs going, "Hey, wait a minute. You know what we could do here? We can just rightsize the company,
177. 30:03drop a bunch of value to the shareholders, and blame it on AI. [laughter] It's fantastic, Bob. I love that idea.
178. 30:16It's so good. You just We're just going to rightsize the company. You know, I used to say downsize. It's a It's a bad
179. 30:23habit. Rightsize. It's going to be so right for us. Oh, those shares are going to climb. Oh, like the like grandma's ivy. That was 2025
180. 30:36because that was also the year where MIT is like 95% of AI projects are useless because no one implementing them knows
181. 30:44what the [ __ ] they are. [laughter] But we're going to lay off 30,000 people because of AI. So that was 2025.
182. 30:534,000 people out of 10,000 people here. I I posted about it. Let me read you the quote that was in his very long letter to his staffing
183. 31:05[sighs] dang dang dang dang. But something has changed. We're already seeing that the intelligence tools we're creating and
184. 31:17using paired with smaller and flatter teams. Smaller and flatter teams are enabling a new way of work which
185. 31:24fundamentally changes what it means to build and run a company. And that's accelerating rapidly. This is from the
186. 31:31guy that founded Twitter who's now got 10,000 employees and tonight has 6,000 employees. Something has changed.
187. 31:41And he doesn't mean it changed like over the past five years. No, like something has changed. Starting in December,
188. 31:52the coding models started to get so good that you can just code with them now and then everything else will follow
189. 32:01that. The ripples of that are going to are just going to keep flowing through. The tools are going to get better and
190. 32:06better and better and better. Something has changed. We're already seeing that the intelligence tools we're creating
191. 32:11and using paired with smaller and flatter teams are enabling new ways of working which fundamentally changes what it means to
192. 32:20build and run a company. If I were you, if I worked in a big company right now, you know what I would do? So, in in
193. 32:25content evolution, we've got this group called Collab. Cindy Coon's in it, Kevin Clark's in it, and we started it. I started it within
194. 32:35the group. group. I asked Kevin if I could start this group just as a way to like learn about AI. Let me just take
195. 32:40what I'm doing on these live these AI learning lab lives and let me do it for for content evolution and it's ended up being like a core
196. 32:50group of six or seven of us that show up all the time. If I were you in a in a big company right now, I would find four
197. 32:58or five other curious, adaptable, non-assshole, AI AI competent, AI forward people
198. 33:13and I would do I would just do a weekly meeting with them. Do a weekly meeting where you're going to talk about what
199. 33:17you're learning in AI. Maybe you do some projects together within your company. Just do it as a little cohort
200. 33:28because what's going to happen is that little cohort is going to start to do some [ __ ] that's going to be
201. 33:34dramatically outsized than if those six people did stuff with AI on their own because they're going to go, "Boy, you
202. 33:43know, I was in a meeting last week and I heard that the CEO was really excited about project XYZ. What if we just built Project XYZ and
203. 33:52handed it to the CEO? Oh, that's kind of risky there. Uh Fred Fred, we can't just Maybe we could. You think we could build
204. 34:02it? I don't know. Let's try. I got a lovable account. Okay, Jim. Do Do you want to make the app? I I already made it. While you were
205. 34:15talking, I I made it. It works. I tested it. Really, Jim? Yeah. You want to play with it? I'll publish it.
206. 34:31Who is it? Uh, [laughter] [gasps] it's Sandra and some of my friends. We heard that you were talking about. So
207. 34:47we made an app for you if you want. Who made this? That's what I do. [ __ ] make that group. Jack Dorsey says it right here. The
208. 35:00tools we're creating and you using, the intelligence tools we're creating and using paired with smaller and flatter
209. 35:08teams are enabling a new way of working which fundamentally changes what it means to build and run a company and
210. 35:17that's accelerating rapidly. Yeah, I decided that u [clears throat] you know I tried that chat GPT when it
211. 35:27first came out. I'll tell you what I did. I went in there. I'm like, "Write me an email to my customer, Jim,
212. 35:32[laughter] and I'll tell and and uh I I'll look at it." And I looked at it and I'll tell you what, that email wasn't
213. 35:39good as good as what I write. I'll tell you that right now. There's still people Kelly Camp sees these people every [ __ ] day.
214. 35:51Nobody knows what's here. Nobody knows what's coming. 4,000 people found out today. 40% of a company found out today.
215. 36:02And he said in the thing it this is not because we're hurting for money. He said we're just going to run the company differently now.
216. 36:14Tik Tok pin. Perplexity computer. Yeah, I I played with it a little bit today. I got access to it. We can go look at it.
217. 36:22It's not really much to look at. It's kind of like looking at at Manis or uh like Gen Spark's a little more visual
218. 36:29even. It's It's not It's just not It's not pretty to look at. So, I don't really have a ton to show, but I'm happy
219. 36:36to go look at it. Oh god, look at me. Ying, what are we going to do tonight to Source Camp? Yep. A tsunami is coming. Build your raft
220. 36:51if you're here. 22 people. There's, you know, if we if we pull out the overlap, there's probably 25 or 30 people here tonight.
221. 37:01If you're here watching this, you're in the water with a raft. Sharon on YouTube. Big company gets worried about their data. Yeah, I know.
222. 37:18I know. >> [laughter] >> I mean, listen, big companies should worry about their data. That's their job is to protect their data.
223. 37:31And they've got all these processes and things in place to make sure that they protect their data. And they're going to
224. 37:39put together a plan, a five-year plan for how to modernize the data so that we can use it for AI. What's going to happen in those five
225. 37:52years? Do you think that plan even sees a year? Tik Tok pin. I'm bringing a few more friends to office hours tomorrow. Oh,
226. 38:02awesome. That's so good. That's been really good. I've been enjoying that. Speaking of which, tomorrow is office hours.
227. 38:13if you want to come hang out with some really smart people um who kind of have it going on [music] and they're also um
228. 38:28[music] I don't know they're they're nice they're nice [music] >> [music]
229. 38:44>> We're talking in humans 2.0 in a in a group chat around being villagers again. Going back to small cohorts of people.
230. 38:56Jack Dorsey says it right here. paired with smaller and flatter teams like the group on the office hours group that that meets
231. 39:11on Fridays is a uh it's a little village. It's a little village. People support one another. They know one another. They give a [ __ ]
232. 39:27When something bad happens to someone, people kind of rally around. When something good happens, they cheer and celebrate.
233. 39:34When someone figures something out, they're like, "Oh, that's awesome." When that same person's like, "I don't
234. 39:40know what to do here." They're like, "Oh, we'll help." I think that's the way through this, right? I think that's the way through this.
235. 39:53[music] Happy to be a part of this group. Really happy to have you here. Crazy times. It is crazy [ __ ] times. >> [music]
236. 40:09[music] >> I don't know why I get self-conscious on here about when I get in these moods, but I do.
237. 40:28Oh, you're Papa JJ. Oh, cool. Papa JJ. Good. Andy in the house. What's happening? Magic. [music] The guitar is so soothing. Thank you.
238. 40:45You know, I played trumpet for many, many years. And what I learned about being a trumpet player is that if you're the best
239. 40:54trumpet player in the world, it's still a [ __ ] annoying instrument. [laughter] Like you can't just sit around jamming
240. 41:06on a trumpet. Like if people are trying to watch the prices, right, and you're practicing their track, he's stop
241. 41:12blowing on that stupid horn. [laughter] But but if you got a guitar, you can just be like, [music]
242. 41:28you can be bad at guitar and people won't mind you playing it. You can be brilliant at trumpet. Stop making that racket. [laughter]
243. 41:40Did I see Valerie in YouTube? No. Let's see. It's hard right now knowing what our work will look like. So
244. 41:52it feels like we're just spinning our change chains and not know how to plan. Valerie, that is so the great repurpose. [clears throat]
245. 42:02What the great repurpose did was for me put a name to this intangible thing that part of it has been for the past year and a half
246. 42:14maybe year and a half somewhere between a year and year and a half I have felt like on this channel I don't know what to talk about because
247. 42:28I like what I Oh, what I knew 2026 was going to have in it were moments of technological advancement where we all go, "Oh, shit."
248. 42:41Like, I knew that was I knew that was coming, but what I couldn't know was what does that actually look like?
249. 42:50And then you get something like Claudebot. And then there's this um Reddit like social media platform for agents and that sort of flamed out
250. 43:01because it had bad security in it. But you're like, "Wait, the bots are talking to the bots." And now we can all have our own bot and
251. 43:08you can design a bot. And I designed my own bot and I realized it had a file called a soul. I'm literally writing the soul of this
252. 43:17entity that I'm going to talk to on Telegram. I get to design >> [clears throat] >> a being that I'm going to talk to on Telegram
253. 43:26and it's going to do [ __ ] for me. Tik Tok note, but I feel so behind not installing Claude OpenCloud. Don't um
254. 43:34I'm a bit technical. I'm a bit technical but not technical. I'm right there with you. If you want to kill a weekend,
255. 43:43have Chat GPT. No, I would say have Claude or Gemini walk you through how to install Claude. It's open Claw. It's not
256. 43:50that um radical. I have to go sit on a bench. I don't know if you were here, Andy, but but I I I talked about um your
257. 43:58friend who [laughter] who said he got that long distance stare like, "Oh, I got to go sit on a bench." Like I
258. 44:09talked about it earlier. Yeah. Yeah. It's We're in weird times right now. And I would imagine that being at work, like
259. 44:18this is the the burden of being early with this AI [ __ ] is that we know that something's coming, but we don't know
260. 44:26what. And when we tell the people that we care about what's coming, we sound like crazy people. I mean, Andy told her friend or her
261. 44:41client, "The OpenClaw bot figured out how to give itself a voice and call its boss." His response was, "I got to go sit on a
262. 44:56bench. How do you do your job right now? [sighs and gasps] Nodding my head over here in agreement.
263. 45:13Did you load it on a separate computer? Yes, I loaded it on a separate computer. I put it on my gaming computer. Um, we
264. 45:20did give it a a um a Gmail account. We're not I I've I haven't done much with it yet. I'm going very very slow.
265. 45:31because I because I want to learn I want to learn how to use it but man but you need s but you need money I assume
266. 45:40I fought an API code for hours today finally got it to take that's good yeah APIs can be a bit of a nightmare
267. 45:49um perplexity computer we can go look at that maybe maybe we'll go look at perplex perplexity computer and I'll
268. 45:56I'll hook it up to some things and see what we can get it to But you need a strong enough computer at
269. 46:01least $4,000 plus. Well, Jay Warner, um, well, it depends. So, so for Claudebot, a $4,000 computer, um,
270. 46:16you you probably need a $10,000 computer. If you really want to run big, powerful models locally, you probably
271. 46:23need to spend 10 grand. If you want a hybrid machine where you can run a decent model locally and then do some do
272. 46:32the heavy lifting stuff in the cloud, then you could probably get away with a $2,2500 Mac Mini. Like I specked out a Mac Mini
273. 46:40today. So today we got Perplexity Computer and Quen, the Quen 2 2.5 model that everyone was talking about being really good. They
274. 46:52dropped Quen 2.5 Flash, which if you have a 32 gigabyte Mac Mini, you can run Quen Flash on that Mac Mini. And apparently it's as
275. 47:08good as Claude Sonnet 4.6, which is an incredibly powerful model. So, as of today, for the the computer I
276. 47:19speced out was $2,200. a Mac Mini with 48 gigs of RAM, an M4 Pro chip, terabyte hard drive, and whatever the 16 core,
277. 47:34I don't know, 16 GPU cores instead of 12. So, so I I upgraded the GPUs, 48 gigabytes of RAM, 2,200 bucks, quiet, silent. You can just
278. 47:48stick it in the corner and you can run something as good as Sonnet 4.6 on that thing locally. So So your Clawbot could
279. 47:56just be sitting on that thing running everything locally. Wouldn't cost you a thing other than that investment.
280. 48:03So that's that's there right now. By the end of this year, there will be local models that can run on hardware even
281. 48:10cheaper than that that'll be better than that. And there's also going to be versions of Cladbot you just install by
282. 48:16like double clicking something. [music] [music]
283. 48:34The clouds still cost. Yeah, I know the clouds still cost. But what I'm saying is like if if the Quen
284. 48:43flash model is as good as they're saying it is, you don't need cloud. You can just run everything locally. I mean, if
285. 48:50you want to write things to your Google Drive, you'll be writing things to the cloud, but you don't need to do the processing in the cloud.
286. 48:59[music] Have you developed AI? [music] Like have have I built an LLM? Have I trained a model? No.
287. 49:14[music] That lives in the neighborhood of sir, what are your qualifications? So what I'm not is a computer scientist.
288. 49:26What I am is a guy with an acting degree who's started 15 companies and has always lived at the intersection of new
289. 49:35technology and self-exression. So, what's interesting to me is not building the LLM. What's interesting to
290. 49:42me is now that the LLM's been been built, how the [ __ ] are the rest of us supposed to use it? Like engineers make up approximately
291. 49:541.5% of humans, working humans. That means that 98.5% of humans, working humans, are not that. What about us?
292. 50:07We've got this remarkable technology. What are we going to do with it? How are we going to tell stories with
293. 50:13it? How are we going to change people's lives with it? How are we going to start businesses with it? That's that's my interest.
294. 50:22I mean, I know how they work. It It boggles my mind that someone understands math well enough to figure out how to do thousanddimensional
295. 50:37latent space embeddings of tokens that are fragments of words that you can just say [ __ ] into. and it creates a
296. 50:45probability waiting and spits tokens back at you and turns it into something intelligible. Like that's [ __ ] magic to me.
297. 50:54But it's to a mathematician, it's probably relatively straightforward calculator design. But what's it trained on? What are LLMs
298. 51:06trained on? The output of humanity. So, generative AI right now, what what why I'm interested in it is I don't give
299. 51:16a [ __ ] about the math. You guys go figure that out. What I give a [ __ ] about is what's in there. All of humanity,
300. 51:26whatever, whatever got stuck on the internet in the past 80 years is essentially in this little softball sized [ __ ] thing
301. 51:36that we can just talk into and it goes updated. Imagine playing guitar on a beach and helping the world. Yeah, exactly. [music]
302. 51:51Have you run a local LLM before? Yes. Um curious if it would be as quick as Claude via online. Well, so the flash
303. 52:01models so so all of this there's um the shorter answer is you can have relatively shitty LLMs respond very very
304. 52:12quickly on essent effectively consumer grade hardware. Like I've got my my gaming PC, my racing rig is a uh a PC
305. 52:21with um I think it's 16 gigs of RAM. Uh a 4090 GPU and and you know a decent processor, right? So it's it's fine.
306. 52:32It's like consumer grade gaming PC. Um I put the 32 billion parameter Quen model on it and it was unusable. It was like
307. 52:44I put a 14 billion model on it. It's like, [clears throat] you know, it's still slow, but it's not bad, but it's
308. 52:50not great. The flash models. So, what they what what these companies do that make these models, they make the giant model first and then
309. 53:01they basically just keep optimizing it and getting it smaller and smaller and smaller. And so the flash version of the
310. 53:08models is like the distilled version of the big model in a really small model. Apparently the Quen 2.5 flash model that
311. 53:15dropped today is apparently kick-ass good and will run on consumer grade hardware. 32 GB of RAM. So um yeah, quantis quant quantization.
312. 53:29Yes. Lord digital gods. I I'm glad I'm glad you all know me well enough to know that I was searching for a word and I
313. 53:35just I couldn't admit that I didn't know it. Couldn't remember it. So, thank you. That was the word I was looking for.
314. 53:43It's the context that's the only piece of the puzzle that's missing. Yeah, but Jay Warner, holy [ __ ] Okay, so if you
315. 53:48know enough about context, then you Jay Warner, if you're decently technical, you should probably install OpenClaw
316. 53:56just to play with it, just to see what it's doing. Okay. So, here's what Open Claw does. I'm gonna put I got so excited I'm
317. 54:04putting my guitar down. Um, the thing that OpenClaw did, which is [ __ ] brilliant, Peter Steinberger,
318. 54:18[snorts] um, is it's got a memory system and it's got se semantic search and embeddings as part of it. So you basically you hook up
319. 54:29OpenClaw to an embeddings um AI model which are they're really cheap because all they're doing is is tokenizing
320. 54:38relatively small text files and and turning them into into um a vector database. But it it happens like as you talk to
321. 54:47your bot, it's writing memory files and then it's it's embedding them in real time. So, so it's essenti it's
322. 54:56effectively got infinite context and then it's proactive on top of it and you get to design how it behaves and you
323. 55:07get to design what it knows about you and what you want it to do for you. It's crazy. Gotcha. The classic good fast cheat pick
324. 55:18too. Yeah. Although we are living in a world now where good, fast and cheap, you can now have all three. Like I think
325. 55:27I think that one of the one of the business models of the future is find any kind of company that will
326. 55:35give you two of the three good fast and cheap and just compete with them on the third on the third axis.
327. 55:42We'll be just as good just as fast, but we we'll be cheaper, right? Or we'll be just as cheap. We'll be cheaper and
328. 55:48we'll be just as good and we'll be fast. Mine sent me nudes. Nice. Nice, Ken. That's awesome. Yeah, I know. I know.
329. 55:59These these agents, they're going to do their thing. I've been using Google Pro. Cool. Haha, I did, but computer is
330. 56:06choking. I have a MacBook Intel i5. Oh, an Intel Mac. No, you need like the the the good news. Okay, here's the thing
331. 56:16about Apple. I'm a longtime Apple fanboy. If you're on if you're on an Intelbased Mac right now, unless you're
332. 56:22basically just running cloud-based [ __ ] where you're just doing the the the orchestration locally, um they're not
333. 56:30it's not going to be usable. If you want to run anything on the machine itself, um just get you you need an M4. I mean,
334. 56:38the one good thing that Apple's done with AI is they made their hardware optimized for AI inference. The how how they don't have
335. 56:50decent AI at [ __ ] Apple is it's it is one of the great blunders of this decade. It's unbelievable. But their hardware is
336. 57:01really good. So basically for 1,500 bucks, 1,700 bucks, you can get a Mac Mini that that will run these models
337. 57:09really, really, really well. [snorts] [clears throat] Apple announcements today. Although, didn't Apple just announced that they're
338. 57:17going to have an announcement next week, right? Oh, Apple announcements Monday. Yeah. Yeah. Yeah. Okay. What's the best
339. 57:23Gemini or OpenAI model? I don't Here's the thing. I am I am I have been decently committed at avoiding open-source
340. 57:35AI. Not because I don't think it's awesome and great, but like you can't keep up with it. If you think keeping up
341. 57:41with commercial AI is tough, try keeping up with all the different models that launch in the open source community.
342. 57:48Because what happens in the open source community, they'll launch a model that's open source. And because it's open
343. 57:53source, a thousand different developers pounce on it and create 27 different versions of it. One's optimized for
344. 58:01gaming and one's optimized for this and one's optimized for fashion design. And you just like I just you go to hugging
345. 58:08face and it's just this constant barrage of bizarre names that have numbers and letters in them and no seeming meaningful combination.
346. 58:20and it makes your head hurt. So, I don't know. [laughter] Open AAI isn't open anymore. No, they're not. [gasps] [sighs]
347. 58:31All right, let's go look at um let's go look at Perplexity Computer and see what we can learn about it. Might go watch a
348. 58:38video on it. Oh, and by the way, if you weren't here the other night, Napster's back. Someone bought the Napster brand
349. 58:45and it's now a bunch of AI agents that can go do [ __ ] for you and make music and steal music apparently. [laughter] Perplexity.
350. 58:59So, I'm going to perplexity.ai. So, from what I can gather, this is kind of like um premium sources. Where's computer?
351. 59:14Computer. This is kind of like OpenClaw except it's just on um it's just on the web. You don't have to
352. 59:27install anything. But [snorts] over here we've got are these connectors Gmail with calendar search create and
353. 59:36manage your emails. I'm connected there. Interesting. Let me disconnect from that. And I'm going to connect this to
354. 59:47AI salon. Add a connector. Uh, oh god.
355. 1:00:01[clears throat] All right, that's connected. What else do we have? Slack. We don't use Slack. Notion. We don't use one drive. A sauna.
356. 1:00:16Um, Air Table. Microsoft Teams, Blue Sky, ClickUp, Canva, Close, Inside Sales Software, C CRM with calling and
357. 1:00:33emailing. Huh, that was hilarious. Oh god, I know. You know what that was, Andy? I was looking at all those permissions and I'm like, I
358. 1:00:44should probably read all these permissions, but [ __ ] it. I'm just going to hit select all. [laughter] Like how many things did did I just
359. 1:00:53agree to? Exactly. Exactly. Oh man. Um [laughter] Oh my god. Data bricks Discord bot. Oh, that could be fun. Let's enable
360. 1:01:08Discordbot because we have an AI salon bot token. How the [ __ ] do I get a bot token? Oh, I clicked on a thing. Pipe dream.
361. 1:01:25One SDK. Thousands of API integrations. Sign in. Sign up. Good lord. How about No, [laughter]
362. 1:01:40[gasps] I won't do that live. Oh, wait. Here's Oh, that's Discord bot. Here's Discord. And oh, it's same thing.
363. 1:01:50Perplexity uses Pipedream to connect your account. [sighs] Keep scrolling. Select the server. AI salon. Look at
364. 1:02:04that. Continue. Authorize. Administrator. I am human. Okay. So, this thing's now connected to Discord.
365. 1:02:18All right. So, let's go over to Discord. The irony of you checking that box [laughter] just so your computer [laughter] can do.
366. 1:02:33[laughter] Producer Brandon is trolling me. Listen. Listen. The trolls are supposed to be in front of the screen, not behind scenes.
367. 1:02:45I'm getting trolled by my producer. Oh, here's 11 Labs. There's Google Ads, Firebase Admin. This is pretty crazy. Google contacts. [sighs]
368. 1:03:00I know what I'm going to do. I'm going to enable Google Contacts. And then this thing is going to be my CRM.
369. 1:03:13I don't want to hear it from Andy or Brandon. I want both of you to just be quiet. Do not make fun of me for not making my calls. [laughter]
370. 1:03:24That's [clears throat] connected. Google Gemini. Oh, we got to go play with Nano Banana 2. We've got that, too.
371. 1:03:33Adam is going to get jealous. Adam's going to bankrupt you. Yeah, Adam's gonna call me. I'm gonna wake up
372. 1:03:40tomorrow to a call from Adam. Uh yeah. Hi, it's Adam. Um yeah, I noticed you were playing around with Perplexity Computer and uh
373. 1:03:51do you think it's cuter than me? [laughter] Google Maps. Let's enable that. Let's enable Google Maps. Why would we
374. 1:04:03not? Oh, I got to have a API key. That That requires work. We're not going to do that. Google Forms connect. Do I have to have an API key?
375. 1:04:13No. Ha. Continue.
376. 1:04:31If you're wondering where we are, I'm I'm at Perplexity Computer. Perplexity Computer, as far as I can
377. 1:04:39tell. I'm I'm new to it, is kind of like OpenClaw, kind of like Manis, kind of like Gen Spark, but it's got all these
378. 1:04:50hooks into other things we use. Nano Banana isn't bad, but Gemini's video content is a laugh. Oh, that's funny. We might need
379. 1:05:05the endless quest quest to avoid outreach. [laughter] I'm getting trolled [laughter] by people that know and love me. I guess
380. 1:05:18that's who to get trolled by. At least, you know, they're paying attention. All right, we're going to enable Google Sheets.
381. 1:05:27Google Meet. [snorts] Oh, that's [clears throat] interesting. Let's enable Google Meet, too. Okay. [clears throat]
382. 1:05:40Uhoh. The account connection failed. What happened? Try again. Tik Tok pin. Teacher's freaking out about companion.
383. 1:05:54What's What's Companion Brother 52? Why hasn't someone used AI to dig those three three million files yet on Einstein? I don't know.
384. 1:06:07Clearly, people I don't So, that's not working. What was that? Google Sheets I was trying. Yeah, that's not working.
385. 1:06:15They're going to have to fix that. Google Vertex, Google Workspace Admin, Business Apps and Collaboration tools.
386. 1:06:24Let's enable that. Nah, let's not. Um, LinkedIn ads. LinkedIn. Oh, yes. [ __ ] yes.
387. 1:06:38Um, sign me up. The The reason I'm excited about LinkedIn is LinkedIn is is very bad at allowing people to screen scrape them. It's hard
388. 1:06:51to get data off LinkedIn. So, even if you want to give them permission, it's hard to do that sometimes. I don't see a Tik Tok pin right now.
389. 1:07:01There's a This is a lot of things in here. I freaking hate LinkedIn. Yeah, I know. And it's going to get worse because just
390. 1:07:15all this AI shit's going to make it worse and worse and worse. Here's chat GPT. We'll enable that. [clears throat]
391. 1:07:23I'm going to go search for more people. You're funny. I'll be back. [laughter] Thank you, Ken. Go get some more people.
392. 1:07:34Chat GPT is good to video creation. I got to put in an API there. I'm not going to do that right now because
393. 1:07:39that's just going to take too long. Pipe Drive. Pipe Drive's easy to use CRM tool. Twilio Send Grid. Send marketing and
394. 1:07:52transactional email through Twilio Send Grid. But I wonder if there's Twilio. Yeah, Twilio. Cloud-based platform for
395. 1:08:02SMS voice and messaging. Oh, man. All right, let's let's we're gonna Let's do it. I got to do API secret keys. God damn it.
396. 1:08:14Telegram. Oh, we could enable Telegram and I could have I can have Perplexity Computer talk to Adam bot token. Okay,
397. 1:08:24we'll do that. I'll I'll I'll I'm going to figure this [ __ ] out. I'm going to figure this [ __ ] out. YouTube data API.
398. 1:08:31You can add a variety of YouTube features to your application. Yeah, we're going to do that. >> [clears throat]
399. 1:08:38>> I think that one's going to be Kylolet story YouTube comment from me to Andy. Andy note to self, hide the fun toys
400. 1:08:48behind the API keys. Pipe drive is bad. Don't use it. Okay, Brandon, it's out of control. [laughter]
401. 1:08:56Listen, I haven't done anything yet. I'm just playing with some toys here. I'm just hooking up some some things. Here's AI Learning Lab.
402. 1:09:05Why doesn't it have our little logo? That's weird. Is this learning lab-ai? Oh, that's Wait, this is pipe dream. I
403. 1:09:15don't want to sign up to pipe dream. Oh, yeah, I do. Yeah, I do. Try again. This this
404. 1:09:30air learning lab. [clears throat] [singing]
405. 1:09:48All right. Authentic authentication successful. Zoom. Zoom admin zero accounting. Text belt is good for low volume, by the
406. 1:10:00way. All right. Um, okay. I don't What did I turn on? Do I have a way to just look at here? Connected. So, I've got Gmail, Discord, Google
407. 1:10:12Forms. Sorry, I meant chat. GPT Dolly Sor is good. Okay, leaving for real now. Miss you already. Thank you, Ken. Bye-bye. [laughter]
408. 1:10:26Okay, let's go back to computer. Here we are. Um, let's see. Go
409. 1:10:42to my YouTube channel and find any videos
410. 1:10:56that mention the word janky.
411. 1:11:10active. Okay. Looking up Kyle's YouTube channel. Oh, wait. Let me stop. Um, sorry. I meant
412. 1:11:28the AI learning lab channel. at learning lab-ai on YouTube. All right, let's see. Let's
413. 1:11:45see if it can find it now. More better. No problem. Let me go look at that YouTube channel. Search through the
414. 1:11:51videos for any mention of Jenkee. Here's what I'm hoping. Because there's a connection to YouTube. YouTube's got all our [ __ ]
415. 1:12:01Lord Digital Gods has chapters in there which probably mention the word janky. So, so there's three levels to this.
416. 1:12:10The first level is can it find videos with the word janky mentioned? That would be decent. The next level would be can it find videos
417. 1:12:21that mention janky and then jump to the chapters that have them. And then the best thing it could do is it actually
418. 1:12:28finds the word janky in the video itself and it jumps you right to that time code. I doubt it's going to be the third one.
419. 1:12:35It could be the second one. It's probably going to be the first one. H lordy lordy lordy. [clears throat]
420. 1:12:49Where's Mimi tonight? Side hustle mimi not in the house. What up? Adam's going to have to have PC do nefarious stuff.
421. 1:13:01What are you on again? Right now we're on um Perplexity Computer. Again, for for everyone, especially the
422. 1:13:09people I'm working with that think I'm just going to turn these bots loose on our stuff. I'm not. Now, is it possible
423. 1:13:16that in learning I just wipe our whole hard drive? Yeah, it's possible. But I don't think I will. [snorts] >> [clears throat]
424. 1:13:28>> What name is this app? This app is Perplexity Computer. No, it's Perplexity Computer. It's brand new as of today. You have to have a
425. 1:13:37Perplexity Max subscription, which is 200 bucks a month, which is a lot. So, I'm going to do it for at least a week.
426. 1:13:50Copy. Clone the drive. Yeah, probably not a bad idea, Andy. [laughter] I mean, I think Google Drive has decent
427. 1:13:59backups. Like, we you can go back to, you know, an an image of our Google Drive a day ago. So, if it wipes it, it
428. 1:14:07should we should be able to back it up, but it's probably not a bad idea to confirm that. Free trial and canceled. Yeah, exactly.
429. 1:14:17All right. Searching. Let's see. Can we look at what it's doing? Scrolling to the top of the search bar, interacting with the page.
430. 1:14:34Yeah, it's it's definitely on our channel. Look at that. Oh, look. It it found janky. Look how today's janky AI will reshape tomorrow's
431. 1:14:45work. The future of work from janky to sophisticated. Kyle shows his janky AI toys. [laughter] It's doing it.
432. 1:14:57It's doing it. Again, there's not much to look at when agents are doing their thing. They're off doing their thing.
433. 1:15:04Copy. Clone the drive. Text belt's good for low. AI should be an should be an automatic canceled or I would forget. Yeah, I
434. 1:15:14know. Joy perie. Oh, you got to share. DM me. How we doing on time? 9:15. Let's We We'll just keep Hey, while this is doing
435. 1:15:22this, why don't we jump over Here's Napster. Why don't we jump over to gemini.google.com? We've got new nano banana 2 [panting]
436. 1:15:40create image. So, we're in fast. So, I assume this is the flash model. If you go to pro, that's Nano Banana
437. 1:15:48Pro. If you go in fast, this is Nano Banana 2. So, we're going to say create image. I like Technicolor. Let's do I'll
438. 1:15:55I'll do the classic because everybody asks for it. Nobody asks for it, actually. 1970s. I think Silverf Fox likes this prompt. 1970s
439. 1:16:08muscle car resto mod in an abandoned factory. We'll add some fun things with a
440. 1:16:23flock of pigeons flying around the rafters. and
441. 1:16:37dusk sunlight bursting through the broken factory windows.
442. 1:16:54All right. And then I'm going to hit technicolor. There we go. Do we want to add anything else?
443. 1:17:09No, we just want technicolor. Was anyone else on producer AI weight list? I just received an email saying
444. 1:17:21I'm in now. No. That's good, Steo. Good for you. Make it drop bird poo. I was I almost put And there's bird poo on the on the car.
445. 1:17:33Yeah, that's pretty [ __ ] good. That's a 69 Mustang. I think it's got a heavily modded grill, so it's
446. 1:17:45definitely a resto mod. There's pigeons flying around the air. There's dusk sun coming through broken windows. That's pretty [ __ ] good.
447. 1:17:58Should do a montage of my first resto mods to now. That's not a bad idea.
448. 1:18:1262 maybe. No, that's not a 62. That's a 67 or a 69. I think it's a 69. [clears throat]
449. 1:18:30Um, oh, you know what else I want to create? Okay, so let's say let me get rid of nano banana for now. And then let's do um
450. 1:18:41let's do tools. I don't want deep research. We'll just have it research. I'm going to say research um the latest. Oh, this is good. Um,
451. 1:18:55import and export numbers for bananas globally.
452. 1:19:11Globally. [clears throat] Jimmy off in the trunk. [laughter]
453. 1:19:25All these things. You are automatically agreeing to give my Gmail account a headache. [laughter] Listen. Listen. I'm willing to be the uh
454. 1:19:39the fool for you. Okay. So, there's a bunch of [ __ ] right? So, now I'm going to say nano banana create image. I'm gonna say, "Make me a
455. 1:19:54visually clear and compelling um infographic I can show at the National
456. 1:20:10Banana Growers meeting." [laughter] Kyle's always taking one for the T team. Oh, did your songs go missing? Oh, yeah.
457. 1:20:25On on Producer AI, here's what they said on on their on their Discord that Brandon went and looked at the other night. Producer AI,
458. 1:20:35we got bought by Google. Google wants to use their own model. Their model's incompatible with our model. So, sorry about your songs. Okay.
459. 1:20:51Ecuador 4.19 billion. Guatemala 1.54. Colombia 1.58. Why are those not in order? Why are those in the wrong order? Why?
460. 1:21:04Okay. United States 3 billion. Okay. Let's go look at the actual numbers. Ecuador 4.9, Guatemala 1.54.
461. 1:21:18Colombia 1.58. I think that's right. Right. 1.18. Yeah. So, the numbers are accurate. There you go.
462. 1:21:34Look, it even put even put the little thing at the bottom by country size, I think. Oh, interesting. Based on
463. 1:21:39estimated global data, disease threats, fusarium wilt, concern in Ecuador, price fluctuations. Oh, wait. What's this? Emerging trends
464. 1:21:51and industry focus. I don't see any big typos. They were they they apparently dramatically improved typos.
465. 1:22:07Holy [ __ ] man. If you used to do infographics or if you're currently in the business of doing infographics, this would be one
466. 1:22:17you wouldn't want to look at right now. To clear to clarify, producer AI dropped the new model and cleared the old songs.
467. 1:22:25Then Google bought them. Oh, and cleared the new model. Huh. Um, let's say, um, I want
468. 1:22:41this to be an infographic that was tattooed onto an actual banana.
469. 1:22:58And this is the studio photograph of the banana with all the stats
470. 1:23:13and graphics easily readable.
471. 1:23:27Um, this should look professional enough to be in a
472. 1:23:41magazine. Let's see if Let's see how it can handle complex prompt. Worth mentioning your on 3.1 flash. Yeah, I'm on. You'll
473. 1:23:54notice down at the bottom here, it says fast. So, this is in the flash model, which which means this is
474. 1:24:02the exciting thing about this for me is I'm going to be able to give Adam, my bot, um, access to Nano Banana 2. And this is
475. 1:24:12I think this is like pretty significantly cheaper than Nano Banana Pro. [snorts] Oh, here we go. Bang. Look at that.
476. 1:24:22Now, it sort of did what I was looking for. That's actually kind of a cool cool thing. Let's Let's see if we can get it. I This
477. 1:24:35is going to be one of those things where nano banana is really bad. Once it gets an idea in its head, you can't get it
478. 1:24:40off there. So, let's say I just want to see the banana um with the graphics on it. It is
479. 1:24:56sitting in a Let's Why don't we do this? We'll say it is Let's see. I just want to see the banana with the graphics on it. It is one of
480. 1:25:10many bananas in a bunch in a fruit bowl in a nice modern modern
481. 1:25:26farm kitchen. Yeah, I was thinking that, Brandon, but I want to I just want to see how good
482. 1:25:42Nano Banana 2 is at solving stuff. Like, if there's not enough room on a banana, does it just make it smaller? Does it
483. 1:25:49zoom in on the picture? Oh, there we go. Look, it did it. Holy [ __ ] There you go. That's [ __ ] bonkers.
484. 1:26:19Yeah, Brand Brandon did a uh a weather forecast. He took some weather data and told it to make a weather forecast. This
485. 1:26:27is pretty good. I mean, like if you were doing a um if you were doing a presentation, you wanted something that was memorable.
486. 1:26:37You start out with that shot and then you zoom in on this and you talk about all the all the silliness. Pretty slick.
487. 1:26:48All right, let's go back to Plexity. Out of 929 videos checked, I found six videos that mentioned janky or jank.
488. 1:27:01There they are. This one in title only. This one in description only. Looks like janky is a recurring theme for you. Um,
489. 1:27:14I'm going to say, did you look in the chapters? Are they called chapters? in the descriptions.
490. 1:27:28I want to be able to jump to the section where I say janky.
491. 1:27:43Now, let's go click on some of their links. One of the things that you discover about AI is that it'll give you
492. 1:27:49hyperlinks and often they don't work. How today's janky AI will reshape tomorrow's work. >> I know you can't hear on YouTube.
493. 1:28:16All right. [sighs and gasps] Uh uh uh uh uh uh reading the expanded descriptions. Okay, that's going to take a while. All right. Well,
494. 1:28:31[sighs] I don't know how I'm going to get good at this. I gave it no weather data. I told it to get the weather data. Tik Tok
495. 1:28:40didn't. Cool. Ads, ads, ads everywhere. Baby Kyle, I know that was that was me from years ago, wasn't it? >> [laughter]
496. 1:28:52>> Oh my goodness. My goodness. All right. Um I think I'm gonna get on out of here. Anybody have a request for Nano Banana?
497. 1:29:02You want me to make an image of something? So, tomorrow at 11:00 a.m. on on on if you go to the AI salon community. AI, click on events,
498. 1:29:16and then tomorrow at 11:00 a.m. I've got AI salon office hours. You should come to office hours. Um, and then tomorrow night is just a
499. 1:29:26normal Friday night, isn't it? Or is tomorrow March 1st? Still February. Okay. So 11:00 a.m.
500. 1:29:41Mountain time is office hours and then 8:00 p.m. Mountain time is Friday night date night. Be here or be on a date [laughter]
501. 1:29:49or bring your date here. If you want to get rid of your date, just tell tell them you want to go back and and YouTube
502. 1:29:57and chill and then put on my channel and watch how quickly they'll leave. Um, [clears throat] any requests? Oh, let's see. Jimmy Hoffa
503. 1:30:15in the trunk of a 1970s muscle car. This is good. muscle car
504. 1:30:30restood in an abandoned factory with Giants Stadium or is it Jet Stadium?
505. 1:30:47It's Jet Stadium, right? Giant stadium. Which one's in Jersey near that? I think it's Giant Stadium that they said Hoff
506. 1:30:55is in the end zone. Anybody? Anybody? Come on. Someone's got to have mob history in their [laughter] in their brain. Don't know.
507. 1:31:10Okay. With Giant Stadium in the distance. [snorts] All right. Here we go. This should be funny. It was Giant Stadium. I knew it.
508. 1:31:35How new is Gemini Banana 2? It is one less than a day old. It was announced this morning. Do you remember how long we used to have
509. 1:31:44to wait for these? Yes. Remember that? Oh my god. [laughter] There's Giant Stadium in the background. There's Jimmy Hoff in the trunk.
510. 1:31:54[laughter] 1970. Nice challenger. That's [clears throat] good. That's really good. It's a little illustration-y. I'm going to say make it
511. 1:32:09more um This is great, but it feels like an illustration. Make it a
512. 1:32:23um cinematic um film photo like Kodak Chrome
513. 1:32:38or something that makes it very very realistic. >> [snorts] >> All right, we'll go back to perplexity.
514. 1:32:53Did it solve anything? Nope.
515. 1:33:12>> [sighs and gasps] >> Yeah, it didn't really didn't really. Although the giant stadium looks better
516. 1:33:27on the blanket. Nice. [laughter] [gasps] Oh my god. Oh my god. All right, everybody. I noticed notebook LM and the
517. 1:33:40detailed slides and infographic is more detailed and less errors. Cool. From a few hours ago testing out Gemini Nano
518. 1:33:48Banana 2 and Notebook LM testing. I'm hoping for more features. It looks like it's doing this in uh 16
519. 1:33:58by9. I wonder if I can say make this a 3x4 tall image. See if it's better at doing aspect ratios because it was pretty shitty at
520. 1:34:12it before. Dear Perplexity. Oh. 527 intro and janky times. Embrace the
521. 1:34:28janky images. Listen [laughter] to my air conditioner.
522. 1:34:45Oh, you all can't hear that on YouTube. Anyway, you can hear my air air conditioner. All right. Well, so Perplexity um computer
523. 1:34:55that YouTube plugin worked. It could it could find chapter mentions. It's not doing semantic search, but it's doing pretty good.
524. 1:35:04It's searching across a channel, which here's the thing. You know who owns YouTube? Google. You know what Google's really good at?
525. 1:35:20Search. And they're good at AI now. Why can why is YouTube not just like this insane data searchable autoeditable
526. 1:35:35monstrosity of a bitching ass product because they don't have an angry Gen Xer sitting in the [ __ ] corner. [laughter]
527. 1:35:45Oh, it just squashed the image. That's bad. It made it vertical. I don't think it's 3x4. It's more like 3x5
528. 1:35:57and they it just squashed it. So that's that's disappointing. I can be insane. Yeah, but you left. So who's talking?
529. 1:36:11Is this is this is this a Jimmy Hoffa thing? Just subscribe to the YouTube channel. Awesome. Thank you so much. Appreciate
530. 1:36:23that. Okay, I'm going to get out of here. Um, perplexity computer looks promising. I'll keep playing with
531. 1:36:32that and see if I can discover anything actually cool about it. Um, I didn't work on Adam tonight, so he's not
532. 1:36:39improved. I I have nothing new to report. Um, I am going to try to install the Quen 2.5 flash model and see if it performs
533. 1:36:51decently on my uh gaming rig. So, we'll see. That's [clears throat] classified. All right, Groovy. Everyone, tomorrow's
534. 1:37:00Friday. Come to office hours. I will see you there. Um, if you don't know where that is, go to community.thesalon.ai.
535. 1:37:07Go to events and you'll see office hours. And then I think you even see Friday night date night, but maybe not.
536. 1:37:13Just come back here tomorrow at 8 o'clock. All right. G. Have a good evening, good people.