6/5/2026 - The Power of Personal AI Memory and Why You Need to Own Your Data

Transcript

1. 0:00Hey, hey, hey. [music] [music]
2. 0:16[music] [music] >> [music]
3. 0:32[music] >> Friday night, date night. Happy Friday, everybody. It is Friday, June 5th. The year is 2026.
4. 0:47Trying to think back of when I imagined the future to be when I was a kid. I mean, it certainly felt like Back to
5. 0:57the Future. What was that? That what? That That was 2020 something, right? 2024. That's the prompt of the week. What is
6. 1:16Oh, future self. Write a letter from Oh, write a letter from your future self. That's a great idea. Okay. So, welcome
7. 1:25welcome to the AI learning lab. Um, we'll be here for about 90 minutes tonight. So, we got solid hour and a
8. 1:32half. It's an ask me anything night. So, um, if you're here, welcome. If you want to pop open the comments so you can
9. 1:40comment and ask questions, that's the link is right up here underneath the AI learning lab logo. If you want to pop
10. 1:46out this window, so you can create another tab and and look and explore the AI salon in a separate tab, you can
11. 1:53actually do that. So you create one tab and you pop this window out and that's right down in the corner down here. And
12. 1:59then that'll keep playing. And then in another tab, you can have the AI salon if you want to go post in in the in the
13. 2:06AI AI learning lab channel or anything like that. We're all in here, but but we aren't all there. [laughter]
14. 2:15We're not weird, we are irregular. Um, so this is an AMA night. It's ask me anything. So if you have questions, just
15. 2:22pop them in the comments below or you can DM uh producer Brandon over in the AI salon. You can do anything you want to do. Um,
16. 2:31it was a decently exciting week. We got some announcements from OpenAI. Um, Mr. Strawberry, the guy that's been trying
17. 2:40to predict this AI stuff for a while, he was saying that chat GPT 5.6 six was coming out this week. It didn't. He sort
18. 2:48of said that was disappointing. So, whatever. Um, but it it pro it looks like maybe next week, maybe the
19. 2:54following week, sometime soon, we're going to have the Mythos model from Anthropic and we're going to have Chat
20. 3:00GBT 5.6 from OpenAI. Um, OpenAI, they announced some things this week, um, including they they are updating the way
21. 3:12memory is handled inside Chat GPT. So, we can go look at that. Um, and that's it. If you have any questions, pop them
22. 3:20in the comments below and I will dig in. All right, let me see. I'll also just make stuff up.
23. 3:36Charlie bit my finger. [laughter] What's the first thing you'd actually want chat GPT to remember to make your life easier? Um
24. 3:53well I think the thing about the thing about memory like why I think
25. 4:09memory so there's a okay so there's a couple of things about memory let's talk about memory memories. You know, people are
26. 4:15talking about AGI and they're talking about agentic systems and and they're talking about these powerful models. Memory is the thing that
27. 4:29is going to make these AGI tools personally valuable. Valuable for you as a human, valuable for you as a worker.
28. 4:43Um, and and the reason for that is right now if you just start fresh with a brand new LLM, you've never used it before. If
29. 4:51you create something like OpenClaw, you know, you get an OpenClaw instance up and running and you create an agent
30. 4:59and you just leave the defaults and you talk to that agent and that agent goes to sleep and then you wake it
31. 5:06up. The next time you wake it up, it remembers nothing about what you talked about before. If you've been using something like chat
32. 5:14GPT for a long time, it's remembered all of your chats, you know, from the beginning of when you started using it. And so you can just
33. 5:24say to chat GPT, hey, what was that project I was working on that involved the Air Force or whatever it was? And
34. 5:31then it will go, oh, you know, let me look. And it'll it'll understand that. The way the the modern memory systems
35. 5:39are working is that every time you interact with it, it's actually pulling um the important information about your
36. 5:48conversation and what it's done for you and it's putting that into a memory file. Um, and so these things are going
37. 5:56to get smarter and smarter and smarter about who you are, how you write, what you like, what you don't like, um, the
38. 6:03projects that you're working on, the people that you care about, the how you respond to emails, how you, um, you
39. 6:10know, invite people to parties, all that sort of stuff. So over time, imagine having a personal assistant that has flawless memory,
40. 6:25knows everything about you, right? I wanted to reme remember I don't like long answers. Yeah. Like that's a
41. 6:33perfect one, Lori. like and and so and then here's here's the thing where I think I am just starting to grock, you know,
42. 6:44no no uh pun intended or no relation to to Elon Musk's gro. I'm just starting to understand that even though I think I don't do
43. 7:00particularly a lot of prompting and use of AI just by the nature of the fact I do stuff on this channel.
44. 7:09I've created a fairly massive data set in the past four four years of how I use AI, but some of that's in
45. 7:18chat GPT, some of that's in Claude, some of that's in Perplexity, some of that's scattered around a bunch of other
46. 7:24things. There's a bunch of image generation tools in MidJourney. I've got mood boards and I've got personalization, which I trained by
47. 7:33selecting which image I liked best, at least 500 of them. All of that is data. That's my data that I can leverage,
48. 7:46but I can't leverage it from a single central place because it's in all these other places. But we now live in a world where taking
49. 7:58what we consider a fairly vast amount of data like my entire chat history of you know every large language model I've
50. 8:05worked with for five years and my entire image generation history and my entire song generation history all of that is
51. 8:12data about how I do what I do. the value of centralizing that and pulling that in and then being able to deploy that under my terms
52. 8:26and say, "Oh, I want to go use this agentic system and I can just point that aentic system at my data, my memory.
53. 8:35I want to own my memory. I want to own my data. I've never It's only recently that I'm getting Oh, that's got value to me.
54. 8:50Sure, it's got value to open AI. It's anthropic. They're going to make new models out of it. And I don't really
55. 8:55give too much of a [ __ ] about that. But now I'm getting we have AI too, right? It used to be that if you had too much
56. 9:05data, you had to be a data scientist to get value out of it. That's not true anymore. So anyway, um you're going to start to
57. 9:15see systems that try to centralize it. There's one called Nessie right now, N SSIE, like Loch Ness. Um that's a Y Combinator
58. 9:24startup that will allow you to go pull your prompt histories from multiple channels and feed them into an agentic
59. 9:31platform like Codeex. I think that's interesting. I'm experimenting with that. I ultimately want to own that
60. 9:36myself. I think there's probably a huge business in um selfdata management. Uh there's probably a lot of huge businesses in
61. 9:47that and I think there's a lot of value in that. All right, let's see. Fun fact, new clients and students in
62. 9:57classes I'm teaching have never heard of GPTs. That's how ahead we are. Yeah, I know. It's It's crazy. It's crazy.
63. 10:11Did you ever have an opinion on smart glasses? I um I have always had an opinion on smart glasses is that I just didn't give a
64. 10:19[ __ ] about them. Um and now I do. You can teach an old dog new tricks as [snorts] long as you create a new thing
65. 10:31to connect the smart glasses to. You did when Google first came out with them. Well, I bought So, I bought one of those
66. 10:41Google Glass things, the little prism that was over one eye. Remember they called those people glass holes? I wore mine in public once for
67. 10:52about 15 minutes. Um, and then I was like, that's just dumb. It was just bad software and it was just dumb, but it was too early. But,
68. 11:03um, Sam today from from office hours. We had a really good office hours today and uh and and Sam who's he's he's in the sales game and
69. 11:15he's he was a sales guy and then he got into prompting and then he got really good at prompting and he became a prompt
70. 11:21engineer and then he became he parlayed that from that to being a product manager and now he's in a senior
71. 11:28position and he's doing innovation. Um, and so this is a guy who deeply understands prompting and the and the the value of creating long
72. 11:41structured prompts that do very sophisticated things. And in the past two months, he's started using codeex, which is OpenAI's
73. 11:53former coding tool, codeex. If you're a human being, you hear something like Kodex and you're like, who cares? Same thing with claude code,
74. 12:04right? It started for developers, for engineers, and then they said, oh, no, there's also Claude co-work. We have same [ __ ] You
75. 12:16just tell it different things. So Sam's been using codeex and he said it's it's absolutely remarkable because
76. 12:24instead of him having to think about the prompt and here's all the data sets and him having to be the bottleneck of
77. 12:30figuring out what to do, his job now is just to understand what are they trying to accomplish and then tell Codeex to go figure it out.
78. 12:43and he says, you know, it's it's efficient enough that rather than just having codeex go do one thing for him
79. 12:51and then launch that thing, he can say to it, I want you to give me three options. And so Codeex will run off and
80. 12:59it'll spin up sub agents. It there'll be like a main agent that he talks to and you can now do it on your phone, right? If you go into
81. 13:10chat GPT, this is one of the reasons that I'm leaning toward codeex over cloud code. I'm just in chat GPT. And if
82. 13:18I go to my little pulld down menu, there's a codeex option. And so it just looks kind of like chat GPT where I can say start a new chat,
83. 13:30but I'm now chatting with this autonomous agent. And so Sam can be at work and just at lunch go, "Hey, I realized we need a new sales report
84. 13:42that focuses on southern Ohio and and it's crossover um with neighboring states um for for people that are selling pink couches,
85. 13:57whatever." And I need it to be an interactive dynamic thing. and then submit and then that goes away and he
86. 14:06finishes his lunch and then he goes out and plays paddle ball or pickle ball, whatever. I don't know what Sam does.
87. 14:13I'm not I'm not Sam's keeper. And then midway through his pickle ball tournament, [clears throat] your interactive report generator's
88. 14:25done. And he goes, "Huh?" Looks at it. He's like, "Yeah, that's pretty good. All right, send that to Bob. Sends it to Bob. Then Bob's got a
89. 14:36report. So that's a very different way of working. So this is a long meandering answer to the question,
90. 14:45why am I all of a sudden a fan of smart glasses? Well, in that example that I just gave you, if Sam's at lunch, he can
91. 14:52easily just pick his phone up off the table and, you know, say something to his his agent. But if he's on the pickle
92. 14:59ball court and his little glasses goz, they say, "Hey, you know, your report's done. You want to look at it? Sure.
93. 15:09Whatever the, you know, yes, you know, talk to the camera and he can just give feedback to it without having to switch context. He can
94. 15:21just be in what he's in and give it instructions and have it go off and do So all of a sudden I see those glasses
95. 15:27as an input device for so here's the here's the value of these agentic platforms in concept right now they're
96. 15:35janky and they require a lot of setup so it's not super simple but the promise of them is very simple inputs yield agents working for hours
97. 15:52to build you the thing you And then when it's done, they let you know and you give it a little bit of feedback. So not dissimilar to how we
98. 16:01work with an LLM right now, except instead of the LLM producing one answer and returning it, it's potentially got
99. 16:10four, five, six, seven of these agents writing code and doing designs and doing research and assembling that research
100. 16:20and turning it into pretty graphs to put into the application. that's being built and there's a coordinator coordinating.
101. 16:26All that stuff is happening for you. So, yes, I think they're a pretty big deal and I think the glass has just
102. 16:33become a new new kind of vector. Um, Chat GBT has memory now, but how do I know what what it actually remembers
103. 16:42about me? So, you can go look. I haven't looked at Chat GPT's memory in a long time. So, I'm going to do some danger demoing here,
104. 16:52which is the the risk of me looking like a boneheaded idiot or high, but it hasn't stopped me before.
105. 17:01I'm going to put your comments over here so I can see them. I can see clearly now. I model switch so much that I have to take notes
106. 17:15when I leave Claude and begin Gemini or Chat GPT or Grock or Manis or so I can see a flow to find the last part of the
107. 17:24puzzle then who has that knowledge ex so Lori Blair what you're describing is I I think what you're describing this
108. 17:34whoever can solve what you just described and Nessie I think Nessie is a company that's trying to solve this.
109. 17:44Whoever can solve and maybe smart glasses are this, right? Maybe whatever is the interface that's how we talk to these things
110. 17:56that becomes the repository. I I don't know. But but I think again if all my glasses or if all my inputs are doing is
111. 18:04interfacing with codecs and OpenAI owns codecs, then they're going to own all of my memory and all of my context. And I
112. 18:11don't like that. I think we should all own our own. Um, okay. Let's go look at [laughter] Oh, I'm gonna have screen sharing issues
113. 18:23again tonight. It is so weird. I just don't like it.
114. 18:38Okay, so here we are at chat GBT. You can still see this, right, Brandon? See? Okay. Um, if you want to find out
115. 18:50where your memory is, you go to settings or you go to personalization. You can go to you can go to probably I think
116. 18:57there's multiple entry points. Um but if you go to personalization base tone and style you can change the personality of chat GPT
117. 19:12characteristics warm more or less. enthusiastic more or less emoji less headers and lists um fast answers. Okay, so custom
118. 19:26instructions I don't have any custom instructions right now about you. Here's my occupation. So this is what I put.
119. 19:33Okay, enable memory and then memory summary. Chat GPT automatically remembers important information about
120. 19:40you and keeps it up to date. You can still view and manage your old saved memories. So this is what we used to have, right?
121. 19:51This is a legacy version of memory which is no longer used. Chat GPT's improved memory automatically stays up to date
122. 19:58leading to more relevant useful replies. So that's actually good to know. And what can I do here with this automatically manage
123. 20:08view history? Um, it's probably not a bad idea if you want to be a data hoarder. You could go grab all the [ __ ] that it used to have about
124. 20:17you. But this isn't being used anymore, but it's still in there. So, if you want that, if you're curious what's in there,
125. 20:22that's probably worth going and looking at. But if I now say manage chat GBT may use memory to personalize queries and search
126. 20:31providers such as Bing. That's interesting. That's a legacy from the Microsoft deal. Record record history.
127. 20:44Oh, let chat GPT reference all previous recording transcripts and notes when responding. Oh, that's kind of cool. So,
128. 20:51if you if you're a big user of voice, um you can discreetly turn that on or off. But anyway, in this memory section, if you click manage,
129. 21:00it's now this this kind of more concise thing. So overview, you're a Kyle Shannon, a Denver-based entrepreneur,
130. 21:07storyteller, educator, and community builder. Much of your current work sits at the intersection of AI creativity,
131. 21:14business transformation, and humor, human- centered learning. You lead and contribute to several o overlapping
132. 21:21initiatives including Storybine, AI salon, AI learning lab, AI readiness programs, and a range of creative
133. 21:26projects. You tend to favor practical human first explanations, strong visual communication, hands-on experimentation,
134. 21:33and work that helps people confidently engage with emerging technology. That's like that's not a bad description of who
135. 21:40I am and what I do. Here's AI salon and AI readiness. Here's Story Vine. Here's my creative work. It's got Sydney and
136. 21:46Artificial Love Story. One of your most significant ongoing projects is the musical Sydney and Artificial Love Story. Um,
137. 21:57now what I don't know about memory, add or update. Oh, add or update. Okay. So, so I wonder if there's anything I want to add here. Um,
138. 22:10I like fly fishing and haven't done it in a while, but am getting back
139. 22:27into it. [clears throat] All right, let's see where it throws that new section. Fly fishing. You enjoy fly fishing, especially for
140. 22:38trout. I don't know where it got that. And it's a hobby you value even though you stepped away for a period of time.
141. 22:44You're getting back into the sport and rekindling your interest on the time on the water gear techniques and fishing
142. 22:51destinations. So, that's fairly full of [ __ ] Um, but it's possible I've talked about stuff about fly fishing before. Anyway,
143. 23:03so that's memory. That's memory. What else is in here? Now that we're in here, I might as well go look at some stuff.
144. 23:10Codeex, you get notified about Codeex tasks, apps. Fun question. All right, let me let me get out of here. Let me go back to the
145. 23:22go back to here. Yes. Community feed in the AI salon. Okay.
146. 23:39So, if you're in the AI salon, the community feed is right right below the welcome section is events and and
147. 23:47connection. The community feeds right in that that section. Mind your tabs? Yes, sir. Sir, yes, sir. You mean community chat? There's a
148. 24:02question. [laughter] Oh, nice. Okay, good question. Okay,
149. 24:18so if you want to go have some fun, jump into community feed and then producer Brandon put this put this question up
150. 24:28there. Wrong answers only. What does GPT stand for? And I won't tell you right now. Wrong answers only. Jump in. Jump in. And uh
151. 24:40we've got uh get people talking. Giant panda tricks. These are good. These are good. Gibberish production terminal.
152. 24:47Goofy people tinkering. Go for programming therapy. Solid. Solid from Dennis. All right. Cool. Yeah. So, go do that.
153. 24:55Go jump in there. Engage with the uh engage with the folks. Engage with your peers. I love it. Um
154. 25:11let me got to grab something here.
155. 25:34Okay. Um, one of the things if you're new to the AI salon, one of the things that you should um,
156. 25:49consider checking out is the AI salon mastermind. So you you can see down the lefth hand side here, you'll be able to
157. 25:57to log in uh to the mastermind and learn about it by just clicking on that section um and you know learn what the practice
158. 26:07lab's all about, what the great repurpose is all about, some of the prototyping workshops, but the mastermind is really if you're
159. 26:15um if you feel like you're you're overwhelmed with AI a lot or if you feel like like you know I don't know what to
160. 26:23do with it? I don't know what to how do I make money with it? Um, everything keeps changing. How am I supposed to
161. 26:30keep up with it? I thought I had these skills mastered and now I feel like I'm starting over. It can get very
162. 26:35overwhelming. Like I've had some personal experience with this in the past week where I just my body just let
163. 26:41me know, hey, you're completely overwhelmed. You're completely oversaturated and you've got to find a way to make
164. 26:48some space. So, the mastermind um has things like the the the weekly mastermind practice lab where we get
165. 26:54together and talk about creating a daily practice for using AI. We've got workshops in the great repurpose about
166. 27:01doing things like decoupling your identity from your job. Because as AI is about to transform every single job,
167. 27:08lots of us are going to be left with, well, wait a minute. If I don't if I'm not getting paid to do the thing that
168. 27:14I've spent the last 30 years of my life perfecting, what what is my value? How do I get paid? Who am I? Where's my
169. 27:20value? What do I want to do? Those are the kind of questions that we're exploring in the mastermind. And so,
170. 27:26it's the kind of thing where, you know, people who've been in it are like, you know, they're they're like, this is the
171. 27:32first space I found where I I feel like I can I don't have to have the answers. um or or um this is the first time in a
172. 27:41long time I feel like I've been able to breathe is one of the quotes we've heard from someone about about the mastermind.
173. 27:47So um it's it's 49 bucks a month. It's 500 bucks a year. You can you can buy it monthly or or annual and it is um
174. 27:59incredibly value for me personally. Incredibly valuable for me personally. Uh if you haven't joined, uh join. Um
175. 28:06and we we do have scholarships coming up. So if you're interested in the scholarship, reach out to producer
176. 28:12Brandon or to Andy in the AI salon. Okay, let's get back to some AMA. What questions do you have? All right, Henry
177. 28:21Hikes. What skill should regular office workers learn right now so they don't get left behind? Um great question.
178. 28:39I think one of the most important skills if you work in an office is let it be known that you are curious
179. 28:51about AI and you're playing with it. But I work in an office where they don't allow us to use AI. Let it be known that you're curious
180. 29:02about AI and you're playing with it. Even if you work in a place where you can't use AI, talk about the projects
181. 29:08you're doing with your kids. Talk about what you're doing on the weekend. Talk about the fact that you spent Friday
182. 29:12night with this dude on in the AI salon who's just talking about AI. Um, there are now I'm starting to see in the
183. 29:25past six months or so, hiring managers are no longer hiring people who don't have AI experience. One of the things that we've experienced
184. 29:36in the AI salon pretty regular regularly because we have these values of play first, create excellence, generously
185. 29:46lead. And when we talk about generously lead, what that looks like is teach what you're learning. Learn out loud, right?
186. 29:52I go play with AI tools to to understand what's possible. I then build some stuff with them or attempt to.
187. 30:00And then take what you're learning in both of those things and tell other people about it and share that knowledge
188. 30:06in the AI salon, share that knowledge on X, on LinkedIn, in the lunchroom, at your office, at dinner parties.
189. 30:15Well, no one wants to talk about AI. Why would I do that? Because at the point at which your company or
190. 30:24your boss says, "Someone's got to figure this AI stuff out." If you're the obvious choice because you're always talking about AI,
191. 30:33you will potentially survive a layoff. Yeah, that's a really good point. Learn something beside Microsoft Copilot. If
192. 30:41you work in an office and they just have Microsoft Copilot, nobody uses Copilot. Well, no, people use Copilot that were told to use
193. 30:52Copilot because they're a Microsoft shop, right? They they they use Microsoft for for Teams and for, you
194. 31:00know, Office and all the all the stuff. And so Microsoft said, "We're going to turn on this AI thing." And then the IT department's been
195. 31:10turning on the AI thing and then the CEO's like, "We're not making any money with AI. What's going on? We turn it on
196. 31:21because it it's not good." And it's it's just it's just a the way Copilot is kind of integrated is it's just an efficiency
197. 31:32amplifier. is just taking all of the processes you currently have and adding more complexity to them to make them
198. 31:39more efficient, which makes you just keep using Microsoft, right? I hate to be so cynical about it, but
199. 31:47it's like, you know, they've got it sort of woven in there in all sorts of places. So, the first thing I would do is I would speak
200. 31:56up, be vocal that, you know, this stuff. And then the second thing I would do is just start playing. Play first.
201. 32:02Go explore. If you haven't made images in chat GPT, go make images. If you haven't told ChatGpt to build you a website or clawed or lovable,
202. 32:16just go do that. Well, I'm not very technical, right? Lovable.dev. If you go to lovable.com, I think it's
203. 32:27lingerie. So, have fun. lovable.dev. dev. No e l-le.dev. I would like to make a website about the World Cup being in the United States and
204. 32:44the ignorance of the American fan versus the sophistication of the global fan. Watch what it builds you. It will melt your face.
205. 32:59and then go to work Monday and say, "Hey, guess what I did over the weekend?" And show them the website that
206. 33:07you built. I didn't know you were into World Cup. I'm not. I didn't know you could build websites. I can now. [laughter]
207. 33:17Are AI agents actually useful that yet, or are we still just giving fancy names to chat bots? That's actually a really good question. Um,
208. 33:28AI agents will never be valuable just because they're there. They also can be really expensive. So, when I when I was talking earlier about
209. 33:40these agentic platforms, you give them a goal and they go off and they spin up these sub agents and they go do all this
210. 33:46[ __ ] They're using tokens. They're they're they're using compute. So, if you're paying 20 bucks a month,
211. 33:54you can very quickly run out of computer use, the ability to do [ __ ] because your 20 bucks a month only buys
212. 34:02you so much. If you're paying 200 bucks a month, you can also blow through those really quick. If you don't have a really good idea of
213. 34:12what you want those agents to do, it's literally just flushing money down the toilet. So I would say that probably 95%
214. 34:22of people playing with agents right now are just playing with them to play with them. So I would say for the most part
215. 34:28those agentic systems are not valuable right now. Um what was the second half of that question? Are we are we still just
216. 34:38giving fancy names to chat bots? These aren't chat bots. So here's the big distinction. We're actually I think we're exiting the era of chat
217. 34:50bots. I think we're going to look back on the first four years of the AI the generative AI revolution quaintly like with with nostalgia.
218. 35:00Remember when you used to go to chat GPT and you would prompt it and it would give you an answer and then you would think
219. 35:08is that any good? Did that hallucinate? Hm. Because that era is is at best going to become optional. The new era is you've
220. 35:23got something like codeex that it's going to do all of its own prompting. You're going to tell it you want to
221. 35:28accomplish something. You're going to give it access to tools like it can go surf the web. It can go look at all your
222. 35:36emails. It can look at your calendar. Um, it can create spreadsheets, it can create word documents, it can create
223. 35:46presentations, it can write code, it's got access to your GitHub repository, right? It's it it's this incredibly
224. 35:55sophisticated suite of activities that it has access to. You get to control that. You have to set that up,
225. 36:06but then it's got access to all that. And then you can just say, "Oh, run through my emails. Find the things that
226. 36:13you think are most important for me to pay attention to. Let me know what those are with recommended responses
227. 36:22and then do that for me three times a day and just it will and it'll just go. It'll just go and it's it'll just go."
228. 36:30So it the agents are the ones using the LLMs, not you. You're just talking to this orchestrator. This think of it like a chief of staff.
229. 36:43So in in in chat GPT land, you're the chief of staff. You know, I'm the captain. Now [laughter]
230. 36:57in in agent world, it's doing all that work. You're just you're just guiding it. You are the captain, right?
231. 37:10Kyle, can you show the basics of midjourney? I still think that's one of the few platforms that's considered
232. 37:16acceptable as a creative use of AI. Well, it's fascinating, Chef Kelly, that So, there's a couple of things about
233. 37:26MidJourney. So, so yeah, I'll hop over to MidJourney and and do some do some demos. I'm show show you how to navigate it.
234. 37:34It's it's it's gotten pretty sophisticated. There's a lot of piece parts to it, but it's relatively simple to do. Um
235. 37:43the CEO of Midjourney from day one has basically said, "I don't give a [ __ ] about any of your concerns about
236. 37:51copyright this that like if there is a platform that is the most egregious, if you have an issue with the ethics of how
237. 37:58these models were trained, Midjourney is probably the worst one to use." Now that said, because of how it was trained, it's
238. 38:07probably the best one in terms of its quality. Um, and there's also, and I'm not just going to just say it was just
239. 38:15about unethical training. It just basically it's been trained on a lot, but there's something about how their
240. 38:20algorithm, how their what they've how they've trained their model is it just produces more artistic results than other platforms.
241. 38:33The image generation within Gemini, the nano banana within Gemini and the image 2 generator within chat GPT are both
242. 38:42remarkable image generators. Um, they can do text, they can do sophisticated infographics. They're remarkable, but
243. 38:50they don't have the artistic oomph of MidJourney. So, anyone that's doing art for the most part is using MidJourney.
244. 38:58Um, but just know if you want to use it commercially, that's the one that, you know, you you potentially put yourself
245. 39:06or your client more on the hook just because of how it was trained and because of they don't really give a [ __ ]
246. 39:11about those issues. So, I just want to give that caveat because that's that's, you know, you say you still think it's
247. 39:17one of the few platforms that's considered acceptable as a creative use of AI. It's acceptable. Um, I think it's an acknowledged higher
248. 39:28quality creative product, but a lot of the creative community really doesn't like MidJourney because of the training
249. 39:36aspects. But let's go play. I thought I had a window open. I guess I don't. [clears throat]
250. 39:50Let me share this tab. All right, fantastic. So, let's see where does this. [clears throat] So, when you first log
251. 39:59into MidJourney, I think you see this, which is just a big long scroll of [ __ ] that's being generated on MidJourney. [cough]
252. 40:14Okay, so lots of things you can do down here on the left. Um, any one of these images you can click on and you can see what
253. 40:23the prompt is and you can see if they here's the here's a profile. That's their personalization profile. Here's
254. 40:29the aspect ratio. So, there's all sorts of information. You can click on this prompt and now that's in now you can
255. 40:36make an image out of that same prompt. So, so if you don't even know where to start, you can just scroll and just find
256. 40:44images you like and use that as a starting point. Like I wonder if that one has like that's got a profile. So
257. 40:50yeah. So if I click on that profile and then it says rose gold metallic look, delicate stripes set against blue background
258. 41:00with a few delicate gold stars. And then I'm going to say um there are let's say stars and then I'm going to say it's raining cats and dogs.
259. 41:16So, I can change that up and just hit submit. And then if you if you want to make stuff, you go down to the second tab, the create tab.
260. 41:27And so there it's Oh, I didn't get any cats and dogs. Weird. I wonder if I have my settings set weird. Oh, I do. Let me go to Let me
261. 41:41turn down weirdness. [laughter] So, I'm going to click on that prompt. That puts it up there. Then I'm going to
262. 41:53do uh that profile and I'll try again. And so the thing that I just did there. Oh, this is really fast actually.
263. 42:03Midjourney is being very fast. I don't know why I'm using 8.1 version 8.1. Oh, I'm in fast mode. Okay, that's why.
264. 42:14Okay, so is MidJourney still paid? Yeah, MidJourney still paid. I think I think they've got a $10 a month subscription.
265. 42:22Um, they might give you a couple of uh generations a day, but I don't even think they do anymore. But anyway, the
266. 42:29way you do this is um you can put in big long complicated prompts. You can also just put in a prompt like um number sign and then I can
267. 42:44make an image and like you're like what do those images have to do with number sign? I don't know. I don't know. And then you
268. 42:52can also go let me turn stylization really high. Let me turn weirdness sorted to here. And then go number sign and see what it does.
269. 43:10Those are cool. And they're like, but I don't I don't like any of those styles. So, one of the things you can do is you
270. 43:17can click go back to this explore tab at the top and at the top of this where all these pictures are, you see this styles
271. 43:25button. You click on the styles button and [clears throat] like here's hot ones. Here's random ones. Here's the top ones of the day.
272. 43:38Here's ones that I've liked. I haven't liked any yet. So, let's just go to some random ones. And for every one of these
273. 43:44styles that I see, I can make if if I say try this style, this says submit your current or latest prompt in this
274. 43:53style. So I'm going to I'm going to go try style for that sort of surreal looking one. And I like this one with
275. 44:00the lines. I'm going to try that. Um Oh, this one's kind of cool. I'm going to try that. I'm going to try this one. Um, I can also
276. 44:16if I say, oh, this is really cool. Rather than say try style, which uses my old prompt, I can click on this F sref in the lower
277. 44:25lefthand corner. That now puts that style in my prompt. And I can say um birthday card with a cake and seven candles
278. 44:45and a puppy. A a lab puppy. All right. And now I'm going to go back to my create tab and I'll have a whole bunch of new
279. 44:59new art. These were the ones I tried. So here was the this was the pound the pound prompt with no styles. Now here's
280. 45:08the pound prompt with this style. >> All right, Kyle, I can't I can't take this anymore. It's a hashtag. >> It's not. >> Yeah.
281. 45:18>> You've called it a number sign. You're now calling it a pound sign. >> Yeah, [laughter] it's both. It's that too.
282. 45:27The hashtag sign. Listen, >> I I mean, if we were broadcasting this show in the 90s, >> welcome to Boomer Nation. [laughter]
283. 45:37>> Also, I'm wondering if uh Meta still has free majour knee hack. >> Oh, yeah. That's We can go look at that.
284. 45:44I'll go check that in a second. Oh, that's pretty cool. Look at this whale. It's got a little eye here. That's cool.
285. 45:55So there's that. This is Sorry. Hashtag hashtag gave us these. >> If you call it a tic-tac-toe board, I'm
286. 46:06logging off for the night. [laughter] >> So, let's see. Did we get seven candles? No, we didn't get seven candles anywhere.
287. 46:15Wait. One, two, three, four, five, six, seven. No, that one we got seven candles. That one's not seven. That's not seven. That's not seven.
288. 46:25But what we could do, so here's a fun thing you can do. I'm going to copy this image. I'm going to flip over to chat GPT.
289. 46:36I'm going to go new chat. I'm going to paste in that image. So, one of the fun ways to work with with these image
290. 46:43models is flip them between tools. So, I'm going to take that style. I'm going to say um recreate this image as is, but make sure there
291. 46:58are seven candles on the cake. So now I'm going to use chat GPT just as an image editor, but it's going to maintain the style.
292. 47:16And and both chatbt and gro nano banana are pretty good at this. So while that's doing that, I'm going to head over to meta.com.
293. 47:36Is it meta meta.ai create log in?
294. 48:04I'm going to go create and I'm going to go back to I'm going to go grab from midjourney. [laughter]
295. 48:31Um, let's see. puppy with birthday cake. And then I'm going to put in the code.
296. 48:45This is the style reference of that one we just did. So this is if you have a Meta account and you go to meta.ai, AI. It uses
297. 49:02midjourney behind the scenes. Although it doesn't look like it's making anything right now, does it [cough]
298. 49:32H dog with cake. I'm going to put in that code.
299. 49:54No, it didn't use it. >> [clears throat] >> Huh. I don't know if they're still using MidJourney, but anyway, this is another
300. 50:01place you can make images. Um, let's go back to chat GPT. See if it fixed our image with seven candles. Looks like it hung.
301. 50:15One of the things I'm seeing chat GPT lately is it hangs on the image generation a fair amount. There we go.
302. 50:241, two, three, four, five, six, and seven is the seventh candle, I guess. Is that the same? Yeah, same exact image, but edited,
303. 50:35right? That had too many candles. This has just enough. So, there you go. But, what other questions do I have?
304. 50:49Followup. Can you create a mood board or image projects in chatbt that you can keep returning to? Um,
305. 51:05not really. You can [clears throat] we can experiment with something. Let's experiment with something. Chat, what's
306. 51:17your name? Chef Kelly, what's your name? What's your name? So, if we go, we're going to create a new project.
307. 51:27Project name is going to be um image consistency [clears throat] uh test.
308. 51:43Am I sharing this? I think I am. >> [clears throat]
309. 51:59>> Okay. Um, so when you create a project in chat GPT, so let's here here's what we're going to do. We're going to hop
310. 52:13over to to MidJourney and I'm going to go find a style. This is me experimenting with lots of
311. 52:29styles. [clears throat] Okay, here's here's a whole bunch. Let's see. Go download. Download.
312. 52:44Download. Download. Download. Like that one. Download. Download.
313. 53:10>> Okay. So, I just went and I grabbed a pile of pictures. Right [clears throat] now, I'm back in chat GPT. I created a project
314. 53:20and I'm in sources and so I'm going to say add sources. And so, we're going to upload. Yeah. So, here's all my images.
315. 53:49Okay. So there I got a bunch of images in there. So now I'm going to do a new chat. I'm going to click create images.
316. 54:00Aspect ratio. You can do real aspect ratios now. I'm going to say um use the images in this use the images uploaded
317. 54:15as sources for style. and create an image of a zebra
318. 54:32on an African plane with a 70s muscle car in the distance doing donuts.
319. 54:49Okay, so None of those images had cars or zebras in them. So this should in theory this should give us what's
320. 55:02effectively a mood board in chat GPT because then anytime you go to this project you can you can do this
321. 55:13and then you can change it you know but you can do that in in midJourney as well [clears throat] but if you've only got one subscription
322. 55:20if you don't want to add a midjourney subscription you can just go find images you like and use those sources in a
323. 55:27project or in a in an individual chat. Let's see how this does. I'll check. There are 13 images listed. That's about right.
324. 56:12It's going slow. I'm going to reload it. The rule says I should include every image from the user's thing. I'll check.
325. 56:24There are 13 images listed.
326. 56:54A watched pot never boils. Pound sign. It is the pound sign. So we called it on push button telephones, weirdos. It's a hashtag.
327. 57:14[laughter] That's bugging me how long that's taken.
328. 57:29All right, we'll [clears throat] come back to that. Other questions? We'll come back to that. I think the answer is yes. Yes, Chef Kelly.
329. 57:38I need coffee. Is AI actually going to replace people at work or mostly replace the tasks nobody wanted to do anyway? Um
330. 57:57I think that there's going to be a lot of there's going to there's going to be a lot of people laid off
331. 58:08assuming that AI could do their job and then realized that there were parts of their job that AI can't do, but it's
332. 58:16going to be the non-obvious [ __ ] Yeah, I think the tasks get I think the tasks get automated. I don't know. It's going to be all over
333. 58:31the place. Some companies are going to keep their employees. They're going to say, "Hey, we're going to get more
334. 58:35efficient and we're going to deploy our employees to do more valuable stuff for our clients." So, I can see there being bosses that
335. 58:47that value the the people that got them there. I think it'll be a minority, but I could see that happening. Um,
336. 58:59in order to survive in a company like that, you're going to have to be AI lit. Like I think AI literacy and literacy is
337. 59:06not even the right word. Adaptability and curiosity and and exploring AI is going to be a prerequisite to work for
338. 59:14for some amount of time certainly in knowledge work. Um and I think it can help you in other things like like even
339. 59:22if you like work for an appliance company and they're like no one's into AI, no one's into AI and all of a sudden the boss
340. 59:30realizes oh we got to get AI going. Who knows about AI, right? So, I don't I don't even think it's, you know, relegated to knowledge work.
341. 59:40[laughter] I don't know. I don't know. I think there's going to be a lot of mistakes made. Um I think certain categories of
342. 59:49work, I think in certain sectors go away entirely, you know. Um, but then there's going to be entire new sectors created that we can't even
343. 1:00:03imagine right now. So, it's just going to be [ __ ] chaos for five years. [laughter] I'm good for now. I'm good for now. Um,
344. 1:00:20let's see. You had a question. You had a question, Brandon, about vibe coding. I thought that was actually a pretty good
345. 1:00:25question. Pop that up there. Let me see if the uh this thing showed up. What's going on with this chat GPT not working?
346. 1:00:45Chef Kelly, uh chat GPT seems to be failing miserably at that image generation. I don't know what's going on
347. 1:00:52with that. Is vibe coding a real skill or is it just chaos with better marketing? Um, vibe coding like prompting is
348. 1:01:07so yes, it's a real thing. The term vibe coding was coined by Andre Karpathy about a year and a half ago and
349. 1:01:17he's a coder and he just said I I've started doing this thing I'm calling vibe coding which is when I used AI historically to write
350. 1:01:30code for me. I would write the code and then I would go look at the code as a human. I would look at all the code and
351. 1:01:39I would understand the code and then I would put it in my codebase and then I would go back and I would get code
352. 1:01:45and then he said he just started using these tools like lovable or I don't know if he was using lovable I think he was
353. 1:01:51probably using cursor at the time or something like that they got good enough that he could just say I want a function I want an app that
354. 1:02:00does this and it would make that and he just decided to stop looking at the code. Um, if you don't know how to code,
355. 1:02:11then vibe coding is very real. Now, is it just chaos if you don't know what you're doing? Yeah. But also, there's people in the AI salon
356. 1:02:25community who are not coders, no Emmy is one of them, who have launched apps to the app store. So, so yes, it's real. Now, why is it real?
357. 1:02:37How How could she do that? Was she just going like, "Give me an app." No, she had an idea for um stickers, like a
358. 1:02:46sticker app for stickers for kitty cats and Puerto Rican celebration and things like that, right? I did. That was a crazy process. So, so,
359. 1:02:58so first she had this idea, right? Oh, I know what I I want I want an app that makes joy and celebrates my culture and celebrates kitties
360. 1:03:12and it's stickers. It's stickers and fun, right? So the thing about vibe coding is and the thing that I think becomes our job moving forward
361. 1:03:24is have the flipping idea and then know what it's froggies. Oh, it's froggies. That's right. It's froggies, not kitties. Sorry. Damn it. I
362. 1:03:37knew it was a cute animal. I forgot it was froggies. Froggy stickers. [clears throat] But have the idea [cough and clears throat]
363. 1:03:50because the value there is not the app. The value there is her idea. Right? She can express that idea in lots of
364. 1:03:59different ways. She chose an app as one of them that becomes it's called kokey cuties. Beautiful. All right. Groovy. Groovy. Groovy.
365. 1:04:10Groovy. All right. Let's see. Yeah, that's that's probably an interesting idea. Let me check one last
366. 1:04:19time in chat GPT and see if it made this image. Why is that not working? I'm going to say try again. Try again.
367. 1:04:36All right, we'll come back. Let me go.
368. 1:05:06See if I got that right. Okay, we're going to do one final thing here. If you have questions, feel free to pop
369. 1:05:20them in there. No, I used Gemini to generate the images and the rest was done with Xcode. Very cool. Okay. Um,
370. 1:05:29when did this come out? This come out today. Why is there no date on this? Yesterday. [snorts] So, what you're looking at
371. 1:05:39here, this is a this is an article that um, Anthropic put out. So, I would consider this article a precursor to AGI. Um, and they're kind
372. 1:05:51of preparing us. Am I not sharing the right tab or am I just not sharing? You're good. Okay. Um, they're preparing us for
373. 1:06:06No, they're trying to prepare us. I don't know if they're trying to prepare us. They're they're educating us that that
374. 1:06:16if you think things are moving fast now, you ain't seen nothing yet. So, they're kind of preparing us for AGI. Cloud is
375. 1:06:23my new friend. Yeah, Cloud's Cloud's really cool. So, the article, when AI builds itself, our progress toward
376. 1:06:29recursive self-improvement and its implications. For most of AI's history, humans drove every step of its
377. 1:06:35development cycle. but anthropic. We're delegating a growing share of AI developments to AI systems themselves,
378. 1:06:40which is speeding up our work. Taken far enough and given enough compute, that trend points to an AI system capable of
379. 1:06:49fully autonomously designing and deploying its own successor. This is called recursive self-improvement. We are not there yet and recursive
380. 1:07:00self-improvement is not inevitable, but it could come sooner than most institutions are prepared for. That sentence
381. 1:07:08is something that Daario Amade has been saying for the past two months and he didn't really say that before.
382. 1:07:16And Sam Waltman's saying similar sort of things and the the guys over at Google are saying similar sort of things.
383. 1:07:24Elon Musk is certainly saying similar sort of things. Although it it seems clear to me that Elon Musk is bailing on
384. 1:07:32on the AI race for now. Um because they just decided to rent their Colossus 2 supercomput to Google or part of it to
385. 1:07:42Google for almost a billion dollars a month. Um, using public benchmarks and previously unreported data from within
386. 1:07:51anthropic, the Anthropic Institute is showing that AI is already accelerating the development of AI systems. To take
387. 1:07:58just one example, today anthropic engineers on average ship eight times as much code per quarter as they did from 2021 to 2025.
388. 1:08:10This technical trend, the technical trends discussed in this piece suggest that AI systems are going to become much
389. 1:08:17more capable in coming years. These trends have huge implications. AI that can build itself would be a major development in the history of
390. 1:08:25technology. One that could bring enormous good for the world in science, healthcare. So when they say enormous
391. 1:08:33good for the world, oh that it links to Dario Amed Amade's machines of love and grace, which if if you haven't read this
392. 1:08:41piece, this is an important piece which shows you Amade's um vision of of what they're trying to build. Oh, you can't. Yeah. Okay. Yeah.
393. 1:08:52Perfect. Okay. Um, but recursive self-improvement also might increase the risks of humans losing control over AI systems. If say
394. 1:09:03systems are capable of fully building their own successors, the ways we secure them, monitor them, and shape their
395. 1:09:09behavior all grow much more important. Okay, so right now you've got a person using a computer to generate code.
396. 1:09:26And then there's a chatbot. I'm trying to understand. I don't I don't know if I if I understand their graphic. Oh,
397. 1:09:32they're just showing increasing sophistication. Oh, I see. Okay. So, so first it was just a computer wrote
398. 1:09:44code to make the first one. Then they used a chatbot to improve their code. Then they used an agent which kind of
399. 1:09:52did it autonomously. Then they did sub agents which did it faster. And then basically where we're going is
400. 1:10:00the agents make the new version of the model and it just keeps it just keeps going in this cycle um and getting better and better and
401. 1:10:09better and like improving you know daily if not hourly. So it's okay here's the all right evidence from the outside
402. 1:10:19world. The rate at which AI models are improving, models improve, is accelerating. The length of tasks they
403. 1:10:25can reliably complete on their own has been doubling roughly every four months from an early trend of doubling every seven months. In
404. 1:10:35March 2024, claopus 3 could complete software tasks that take humans about four minutes to complete. A year later, sonnet 3.7
405. 1:10:46took about an hour and a half. A year after that, claude opus 4.6. This was the model that most engineers said,
406. 1:10:53"Okay, things have changed. It managed 12-hour tasks. If this trend holds, tasks that take a skilled person days could come into
407. 1:11:04range this year." So anyway, so this is this is an article worth going to read. All right, let me come back here. Come
408. 1:11:17back here to you good people. All right, thoughts, questions. I might let you out early. I'm going to go to bed early tonight because I'm
409. 1:11:27getting up at 400 a.m. to go fishing. [laughter] This is going to be hard for those of us to use with control issues, I bet. I
410. 1:11:40don't know, Lori. I think that In a lot of ways, I think if you've if you're if you've got control issues, the agents are going to
411. 1:11:54work to please you, right? So, you could argue that that you're coming in like people with control issues are
412. 1:12:03coming into their own, right? Because they can just say to the agents, "No, I want it this way. I want it this way. I
413. 1:12:08want it this way." You can be as controlling as you want. They'll just keep bending it into your vision. So, um,
414. 1:12:17can I build a website in OpenClaw? You can have you can certainly have OpenClaw, yeah, write code and build a
415. 1:12:25website. Will it be any good? You have to give it all sorts of skills and things like that. Um, if you're playing
416. 1:12:31with these agentic tools and [clears throat] haven't learned about skills, go learn about Claude skills. Volume down. I should talk lower or
417. 1:12:42it's down. Was I too close to the mic? Might have been too close to the mic. Um, okay. Weekend homework. Well, I'm going fishing. Um,
418. 1:12:58I've said this for the past couple of weekends, but I'm just going to keep saying it. I I would take some time this
419. 1:13:03weekend to think about if you had a system that you could just tell give it goals to go do something. What would you
420. 1:13:10do? I would also start, you know, here's some good homework. You know, Lori, to your point earlier, go a good exercise this weekend,
421. 1:13:26start to catalog all of the AI tools that you've used over the years. Like, are there tools like Idog where
422. 1:13:36you have used them a lot, but you haven't used it in forever? Or or did you [clears throat] used to use
423. 1:13:41Perplexity and you don't anymore? like think about all the tools that you've used and start to catalog them because I think
424. 1:13:52this idea of doing a little bit of self-discovery and and sort of pulling your data from those systems so that
425. 1:14:01you've got it locally um seems like an exercise worthwhile. Kelly Camp, what are the pros and cons of learning AI later in the game rather
426. 1:14:10than earlier? The pros are certainly so. So, Kelly, Sam on the I don't know if you were on the on Oh, yeah. You were on the call.
427. 1:14:20the at the end, you know, Sam Sam is a really good example on the office hours call of someone who who invested deeply
428. 1:14:31in learning prompt engineering, got a job because of it, got promoted because of it, like changed his career because
429. 1:14:38of it, and has essentially thrown all of that out. Like he doesn't need to do that anymore. Now, is that wasted time and wasted knowledge
430. 1:14:51that he has? No. Because it informs everything he does moving forward. And he understands if he needs to dig into
431. 1:14:58one of these agentic systems and do a a manual prompt or build a skill, he's got the ability to do that, right? So, so I
432. 1:15:08think the the the knowledge that he gained during that those times carries forward, but the specifics of the tasks are wasted.
433. 1:15:20So, the advantage of coming in later is that, you know, someone that comes into AI, they're like, I should figure this AI
434. 1:15:28stuff out. If they start a year from today, these agentic platforms are going to just be like chat GPT. They're going
435. 1:15:34to be really easy. You'll probably just talk to them. Hey, I want to start a business. Okay, we're going to go start a
436. 1:15:42business. It sets up all the [ __ ] on the back end for you. So, the advantage of coming in late is you're going to come into more
437. 1:15:50sophisticated tools. The cons of that are that all of the opportunity goes in the early in the early times. The the
438. 1:16:01other one that I think is the more important thing is that the um
439. 1:16:15the emotional like the great repurpose stuff and the emotional toll and the anxiety that AI produces is intensifying.
440. 1:16:29So, even if I come in and it's just really easy to use AI and I just talk to a thing on the phone and it goes and
441. 1:16:35spins up a business for me, [clears throat] if I've never been an entrepreneur before and I actually liked doing data entry,
442. 1:16:44but data entry jobs are now gone, but I I enjoyed that, but now I'm being asked to be an entrepreneur and they're like,
443. 1:16:50there's this AI thing that'll just make you businesses. Okay. like the emotional toll of that for someone coming in new, I think is going
444. 1:17:01to be really intense. It's going to be really shocking. And I think that for people who have been in it for a while, like we've been
445. 1:17:09experiencing the highs and the lows and the overwhelm and and the did I learn all this for nothing and oh [ __ ] now
446. 1:17:16here's the customers, what do I do now? all of that emotional processing, someone coming in new, it's just going
447. 1:17:25to be like a shock to the system. So, you know, I think it's going to be easy for easier for them technically, harder for them emotionally.
448. 1:17:36All right, kids. [clears throat] So, go catalog where you've been. [laughter] Go pretend pretend like you've gone to
449. 1:17:46the doctor and they're like, "Have you used AI in the past four years?" Uh, yeah. Yeah, I have. Have you used things
450. 1:17:54other than chat GPT? Oh, god. Yeah. Yeah, I have. [laughter] [clears throat] So, emotional whiplash is inevitable. It's either happens
451. 1:18:04earlier or later. Yeah, I think it does. And I think for those of us who who are going through the emotional whiplash as
452. 1:18:12as new people come in and are shocked by it, I think part of our role, part is an education role, but part of
453. 1:18:19it is just, hey, I've been there. I know what you're going through. Right. [clears throat] That's what I lead with at the doctor. [laughter]
454. 1:18:30Yeah, I've used a little more than chat GPG. Um, yeah, I would just start to be aware of where all of your data is. You know, if
455. 1:18:41you're a good data hygiene person and you have really nice organized folders, great. But if you've got stuff all over
456. 1:18:47the place, like I historically have, and you've got accounts all over the place, just start to gather them. It's going to
457. 1:18:53be valuable over time to be able to pull that pull that context in for these agentic systems. It's going to allow you
458. 1:18:59to have some superpowers. All right, I'm going to get out of here. My voice is getting scratchy. I got to get ready to
459. 1:19:05go fishing, go to bed, and uh I'm out. Thank you very much for being here. Thank you, producer Brandon. I
460. 1:19:16know he's tired, too. Had a long week. We both [clears throat] had long weeks. So, uh have a beautiful weekend,
461. 1:19:22everybody. Um I hope you're doing well, and I really appreciate you being here. Peace out. Good night.