7/17/2025 - OpenAI Agents: A Deep Dive into Capabilities and Shortcomings

Transcript

1. 0:00A champion. A champ. Champy. Champ. [Music]
2. 0:21[Music] [Applause] Hello. [Music]
3. 0:40Hello. [Music]
4. 1:03Hey, hey, hey, calm down. Sit, sit, sit, sit, calm, sit, calm down. [Music]
5. 2:34[Music]
6. 2:48Good evening. Good people. What is happening? What's going on? Almost going down. [Music]
7. 3:27Do [Music]
8. 3:42our speed levels. They'll marry our bons together. [Music] I've got some real estate here in my bag. So, we bought a pack of cigarettes
9. 3:58and Mrs. Wagner's pass and we've both come to look for America. [Music] [Applause] [Music]
10. 4:28Yes. Yes. said as we boarded the ground in Pittsburgh. Michigan seems like a dream to me now. So we bought it. Wait, it took me 40
11. 4:43days. Wait, it took me 40 days. It took me [Music]
12. 5:14Robert Rossy. Hey, Kyle, you're really gotten you've gotten really good at singing and playing guitar. Oh, thank you very much.
13. 5:22It's, you know, it's like uh it's like 10,000 hours on the same seven songs. I appreciate it. I do appreciate it. [Music]
14. 5:55W Grace desperately heading his old place. Dreamed would discover a new space. Buried himself alive inside his basement. tongue on the side
15. 6:08of his facement. He's working away on displacement and what it would take to survive. Cuz when you're done with this world,
16. 6:23you know the next is up to you. Ever once in his life it was quiet as he learned how to turn with the tide
17. 6:38[Music] and the sky was a flare and the [Music] I don't [Music]
18. 7:01Oh, good evening. All right, so we got agents. We got agents. Open AAI came out with some stuff. Um, okay. We've got some irregulars fun.
19. 7:13We've got a Dr. J video. We've got a Dr. J Skittles rap. Okay. Hey, and a Steo image. Nice. Okay, we'll
20. 7:24go look at that. We'll go look at that stuff. Did I share my screen yet? No, I haven't. I have not shared my screen as very often because
21. 7:35it's become clear that I am a loser. Eno also. Hi, Kyle Shannon and the magnificent champion indeed. Champion
22. 7:49where? Well, Champ was magnificent in here. Hear about the big announcement. Oh, here to hear about the big announcement.
23. 7:59I don't have it yet, so I can't play with it. I didn't reup my $200 a month subscription just to get it. Uh, so
24. 8:07we'll probably have it tomorrow for Friday night date night or by the end of the weekend for sure. Um, so I I can't
25. 8:14tell you if it's good or bad, but I can walk you through what it is and and why it's a why it seems like it's a big deal and
26. 8:23um it will likely be disappointing. [Music] Um,
27. 8:38but you know, some some people that I respect like Ali K. Miller uh had early access to it and seemed to really like it. Um
28. 8:46she said it's got some jankiness to it. So, I'll talk about that in a bit. Uh yeah, that's that's about it. That's about it. All right.
29. 9:00Oh, Kyle, you're not a loser. Thank you. I appreciate that. I you know there's some days there's some days when the uh
30. 9:13the universe conspires to remind you you maybe don't have it going on. So I appreciate it.
31. 9:28All right. John Harris, what is up? What's going down? Thank you, Mr. Appreciate that.
32. 9:44source camp. Do you thought the the OpenAI announcement was a a bit of a disappointment? I didn't like I wasn't
33. 9:51expecting GPT5 and we didn't get GPT5. So, >> um, let me go look and see.
34. 10:05So, let's see. What were we doing last night? We built some fun stuff last night. Oh, we built the uh we built the Skittles image.
35. 10:15The Skittles dude talking about Skittles. Like I like I said, as I predicted, nobody nobody cared. Nobody
36. 10:22like that that post got no engagement. I thought it was really funny, but you know, apparently my uh my definition of
37. 10:31comedy is not the same as the world's. Did you see the underwhelming announcement today? I did. I I don't I
38. 10:37don't know if it was underwhelming. Um, I guess I I I have yet to use it, right?
39. 10:53So, I don't know, but it it looks like it's somewhere in between Manis and GenSpark. Um, in fact, Manis trolled
40. 11:01OpenAI today and said, "Welcome to the party." Which I thought was pretty good. Um so if you don't know it came out today
41. 11:09OpenAI announced uh agents uh which is a uh just a new mode that effectively combines um deep research and operator. Operator was
42. 11:25their tool that surfs the web visually. Deep research was their tool that very very efficiently reads the text of
43. 11:32websites and you know can take action on them. Um and then they've got some of their um their thinking um capabilities, their their
44. 11:45um reasoning capabilities where this system can use tools like it can write Python code and execute it all
45. 11:56within the context of going out and doing something for you. One of the big disappointments was every time someone
46. 12:03shows one of these systems, the first thing they show is go book me a vacation. And so they started with go
47. 12:09book me a vacation and it didn't really do it. Like like like every one of these demos it sort of does it but not really.
48. 12:20It took 15 minutes and didn't really do [ __ ] Um let's see. Open AI. Uh agents. Is it agents? Agent OpenAI agent.
49. 12:35[Music] This is about to be so annoying.
50. 12:50Okay, so let me let me shift my shift my sharing. [Music]
51. 13:09All right. All right, people. This is from OpenAI. So, this is what you would call a marketing video. To be honest,
52. 13:17I've been happier with Grock as of late. It takes longer to get a response, but in the dev world, it's a hell of a lot
53. 13:24more accurate. Oh, that's fascinating. Interesting. Interesting. Interesting, huh? Yeah. Well, that's, you know, John, I
54. 13:32think one of the things that becomes increasingly clear to me is that all of these tools, like all of them are going to be really good.
55. 13:47And then the ones that people choose to use are the ones they imprint on for a specific use case or it's got a specific
56. 13:53personality or in the case of like the wu you know companion things like literally has a personality like you
57. 14:00know I like my my wu girlfriend. Um, and so it it doesn't surprise me that, you know, if if Grock is doing
58. 14:10something, you know, more specific and accurate for what you're up to, it makes sense that that loyalties are going to be fleeting as
59. 14:20these tools get more and more capable because people are going to, for one thing, you just can't track them all.
60. 14:25So, you're going to sort of lock into one or two core tools that you use. Um, and then you'll just, you know, you'll
61. 14:33just do your stuff. I think you, since you convinced me, describe to all of them. I get to bounce from one another,
62. 14:40uh, as they get as they get better at the AI race. Yeah. And that's, listen, I that's one of the one of the downsides
63. 14:48of being early is also one of the advantages, right? If you're early, you get to try all this stuff. You get to
64. 14:55see which which one takes the lead. And you know, there's taking the lead on the benchmarks and there's taking the lead
65. 15:02in in how you use it. Um, and so you get to play with that stuff early. The downside is you got to play with all
66. 15:10that stuff to figure out which one's good. Hang on a sec. Let me go grab something.
67. 16:19[Music] Okay. I'm back. I've got Aqua. Let's watch videos. Let's watch a video, shall we? Ho, please hold, please.
68. 16:34Got to say, play with them to see which ones are good because hell, who's who else is who else are you going to ask?
69. 16:40Yep. This is this is the the the reason this channel exists is having done this before in the in the mid 90s with the worldwide web. What I
70. 16:52know is when you're at the early stages of a technological shift like this, nobody knows anything. There are no experts. I exemplify that.
71. 17:04Sir, sir, sir, what are your qualifications? What pretel are your qualifications? Yes, good day there, Jurgens. All right.
72. 17:19Um, yeah, they're cancelling the Late Show with Co Bear. I know. Pretty [ __ ] up. Anyway, all right. Are we going to
73. 17:30replace them? Yeah, I think so. I think I think in general as as viewing habits go there's there's Col Bear AI Learning
74. 17:41Lab and and Conan Conan's podcast and then Joe Rogan somewhere south of that, right? Um Oh, and C-SPAN. Yeah. Good lord. Well,
75. 17:57>> keep keep watching here. [Music] >> It may be a while for the whole world to evolve to a AI agent centric worldview
76. 18:09and so I think we should do what we can to meet the world where it's at. [Music] My name is John. I work on the deep
77. 18:21research and agents teams at open. One great use case that comes up a lot is you have some kind of budget file and
78. 18:28whenever you do that it's kind of a pain. It takes maybe 4 to 8 hours and that's kind of your day. I'm going to
79. 18:33show you an example where the agent sources information on the city of San Francisco's annual budget expenses and
80. 18:40revenues for the past 5 years and it's going to compile that all into one nicely formatted spreadsheet. It goes on
81. 18:47by itself. I usually just close my laptop, go grab a coffee, maybe I have lunch. So, first it needs to find the
82. 18:53data. So, it probably does a web search to figure out where it can find this San Francisco city budget information. Once
83. 19:00it finds the San Francisco city government website, it will try to access the PDF files. So, it has its own
84. 19:07file system and everything. Then, it needs to extract maybe 200 numbers from each PDF. And finally, it will have one
85. 19:16command that will generate the entire spreadsheet all at once. If you go back to the chat, you'll see the final
86. 19:21response. And let me just open it now. Yeah, I think it got 98% of the information correct. It also formatted
87. 19:30the Excel workbook as I instructed. >> Wait, wait, wait, wait, wait, wait. Let's just let's just hear that again.
88. 19:36>> 98% of the information correct. >> All right, it got 98% of the information correct. What about that 2%.
89. 19:44What about that 2%. How do you find that? Who finds that? Does it tell you that it got the 2% wrong? No.
90. 19:53It also formatted the Excel workbook as I instructed it to. In this case, the revisions were small, so I just made
91. 20:00them within Excel because it was just a copy paste. But absolutely, you can make them in chat GPT. I would say just try
92. 20:07it out. If it can do 90 95% of the actual time consuming part of the work, uh, that's going to save you a ton of time. >> Low Steve Jobs.
93. 20:20>> Well, they had to today's today's video was was indeed a uh it it was still a tiny tiny table awkward presentation. If
94. 20:29you didn't see it, um, let's let's go I'll go show it to you. We might as well make fun of some some awkward uh Open AI announcement
95. 20:38action. Um, Open AI. >> Okay. So, >> we started
96. 20:53>> previously >> awkwardly close together. They now have a curved couch instead of like four chairs around a tiny table.
97. 21:02So, they got a larger table, but it's it's lower. And they still got the awkward single laptop that they sort of
98. 21:08shuffle around to each other. Vicki made a good point. It should be a lazy Susan. Uh, and the other thing that
99. 21:15they got, which is new, is they got more than one camera. Look, they could zoom in on Sam. But Sam was here. Um, and yeah, and
100. 21:29they, you know, they they talked about it. Um, let's see. It's the SAM cam. Yeah, exactly. Um, it's happening. There's
101. 21:41>> Okay, so can I zoom in on this? Yeah. So, within the context of a search, the Coldplay concert also had more than
102. 21:56one camera. Ouch. Yikes. Um, so what you're looking at here, a couple of things that they did that I think is kind of clever.
103. 22:11So within the context of a normal chat GPT search, you now have these agent windows. You can you can spin up an
104. 22:19agent. The 20 $20 a month subscription. you're going to get to do 40 40 agents per month. So, basically an agent a day.
105. 22:30If you did one agent a day, um if you did one agent a day, um you know, you you'd sort of use up your
106. 22:40quota in a month. I don't I assume that limit's not going to last that long, but these things can go on for 15 20 minutes. Um
107. 22:49some things that they did that I think are interesting. One is just the interface. They use the same basic
108. 22:54colors as um as advanced voice, right? So with advanced voice, you sort of have the blue cloud sort of thing that moves
109. 23:04when you talk to it and and the border of this thing when it when it's doing it thing thing, it kind of breathes. So, as
110. 23:12mentioned, we gave the agent access to its own virtual computer. And the computer has many different tools
111. 23:18installed, and it can choose which to use as it's working through the task. >> So, it's got different tools it can use.
112. 23:24It can choose which ones to to to use, right? And so, some of those might be make an image. Some of those might be
113. 23:30write Python code. Some of those might be um go surf a website visually. Some of those might be go do deep research
114. 23:38and understand websites. Tik Tok pin. They made their presentation awkward so their product seems brilliant.
115. 23:50Oh my god. Uh her accent was lovely. One of the other guys accent I could not understand a word he said. It's like
116. 23:57come on. Like I like I appreciate that he's the scientist that did it but you know ultimately you have to communicate to your audience. Um,
117. 24:09couple of things that I think are interesting here and and where I kind of feel like OpenAI has a chance to be
118. 24:16better at some of the other than some of the other tools. Um, but the other tools, the tools that they're coming
119. 24:23into the market against, Gen Spark and Manis, and there's a couple of there's four or five of those agentic kind of
120. 24:30tools, but Manis and Gen Spark are the two biggies right now. Um, they did a lot of reinforcement learning
121. 24:40on how how this system chooses which tool to use because what they realized is the tasks that require you to look at
122. 24:50a website and click on it and see see things visually are often very different tasks than when you when you need to
123. 24:57research lots and lots of websites quickly. And so it sort of figures out, oh, I need to do lots and lots of web
124. 25:04research. Let me use deep research or let me use, you know, the visual tool or let me go write code. So its ability to
125. 25:12context switch between those tools. I have an instinct is going to be really what sets this apart. Um, it could also suck.
126. 25:24So to be very very clear, I I I fully expect this thing to be kind of like Operator was. When Operator came out, it
127. 25:31demoed really well. And then when you used it, it was sort of so slow and so clunky and it didn't really work that
128. 25:38you were just like, "Eh, whatever." Uh, and no one really used it, so they recognized that, too. Um, I was like,
129. 25:44"Talk talk normal and stop looking into my eyes." I had to look away a few times. presentation had some bad vibes. Why
130. 25:58were they whispering at times? Well, that's just Sam's thing. Sam's like, "Yeah, it's, you know, we encourage you to change the world." Um,
131. 26:09uh, Lily and Daddy are in the house. Very nice. Good to see you. Um, Kavuno, I prayed the AI gods for chat GPT5.
132. 26:17Yeah, I wasn't expecting chat GBT5. They need paid professional actors, you know. They don't they their their
133. 26:25last presentation they did, they had three presenters that presented that were like people that could communicate.
134. 26:35You don't need to have the scientists that worked on it demo it like like you know I I mean this is this
135. 26:45is part of their transition from being a a a pure research company to being a consumer software company. Um they need
136. 26:54like product salespeople, right? You know product marketers that understand how to talk about products. um
137. 27:02have the scientists there to meet them and then have people that know how to demo the product demo the product. Um
138. 27:09and if if I never see another [ __ ] book a vacation for me, that would be great. You know what I want? I want um start up an e-commerce
139. 27:26business for me and test you generate and test some ads for it. Come up with a product, design the product, source the
140. 27:35product, come up with a marketing plan, create the ads, do the media buy, take 50 bucks from my account, go test some
141. 27:44ads, and you know, within the next hour, um let's launch a business together. That's one I'd like to see.
142. 27:52I could give two shits about a family of four traveling to a visa. I haven't taken a vacation in six years.
143. 28:04I can't wait till Kyle gets his agent. Yeah, man. Let's go start some [ __ ] businesses. It was one of the things,
144. 28:10you know, at the at the AI salon this week, we asked the question, um, it was a meet and greet, so everyone who came
145. 28:18got to share who they who they are and what they're up to. And we we posed the question, we had everyone answer the
146. 28:24question, what's most challenging about AI for you right now? And a lot of people are like, hey, I've learned all
147. 28:32these tools. I know how to do these tools. Nobody seems to give a [ __ ] right? all the people that I'm trying
148. 28:37to, you know, talk to about this stuff are like, I don't care. I don't care about AI. And we're and we're all like,
149. 28:43uh, it's gonna change everything. And then meanwhile, we're like, how do we get people to pay attention to the [ __ ]
150. 28:46we're building and how do we monetize it? So, I get the frustration. This is what what we are all collectively
151. 28:56experiencing. If you don't know the book Crossing the Chasm by Jeffrey Moore and someone else, is it Jeffrey Moore,
152. 29:02Crossing the Chasm? I think it is. Anyway, whatever. Um, we're in the middle of the chasm right now. And the chasm is the chasm between
153. 29:10the early adopters and and sort of the, you know, the fast followers. And it's this big long chasm where all of the
154. 29:17early adopters have essentially been tapped out. And we're kind of waiting for the people on the other side of the
155. 29:22chasm to get religion. And that surviving that gap is actually really hard. um uh source camp right now is
156. 29:32figuring out one of the ways you cross that chasm is you you niche down, right? And she's doing a lot of work in sectors
157. 29:39that she really deeply understands. So she's doing AI consulting but in a very specific area. Jim Ross is doing
158. 29:47stuff in a very specific area. Um that probably starts to look like where where where we find success short term.
159. 29:56Um, do programmatic SEO for me. Exactly. Chat GPT is a gateway drug. I want more. I want a stronger model. I think
160. 30:04everyone does. Kabuno, I taught my first AI class at work today. Yes. Uh, and I used a clown. A clown knows noise when I
161. 30:14described. Oh, this is great. What jank is. Oh, that's great. What? That's good. You put something in there and it does something bad. What?
162. 30:26Actually, that you could you could totally do one of these buttons for the jank. >> You can make money,
163. 30:32>> right? Just do a clown noise in one of these. Um, yeah, that's beautiful. That's awesome. Great. Um, [Music]
164. 30:46I have ch I on China. I did a presentation on Q developer to 100 people today. Nice. Exactly. I need a
165. 30:53button. Yeah. I mean, what's what's nice about that, Chris, is by demoing AI and acknowledging the warts, acknowledging the jank,
166. 31:06what you're actually doing is you're training people to not disengage their critical thinking mind, right? To do AI well, you still need to
167. 31:19be engaged. And I and I think this is going to be for some time. I mean, sure. He said, I said, I I told it to go off
168. 31:28and do this budget thing and I went out for coffee and it went off and it did its thing. But then he said it got 98%
169. 31:36of the stuff right. Well, so that means you're coming back from coffee and if you're doing your job right, you're
170. 31:42going to review all the sites it looked at. You're going to look at numbers and say, "Did it actually bring are these
171. 31:47numbers correct? Are these charts and graphs real or did it just make that [ __ ] up?" Right? That's a lot of work.
172. 31:54That's that's probably two hours worth of work. Now, maybe you get chat GBD to help you analyze it. But
173. 32:03I mean, it's exhausting enough right now if you just have normal chat GPT generate you a bunch of ideas for for some marketing concept
174. 32:13and then you got to go through and deal with all those ideas and like, oh, that one's actually not good and this is a
175. 32:17bunch of repetitive stuff. it ends up generating way more than you would have had to deal with historically, right?
176. 32:24And so these agents are going to do that times a hundred. Um, some report getting frustrated and giving up,
177. 32:35but it wasn't even their faults. Yep, there you go. How do you check it without doing it all over and getting the right output? Yeah. Well,
178. 32:46if you think back to two years ago, two and a half years ago when Chat GBT first came out, you know, it could maybe get you to 65
179. 32:5970% of what you want, you know, in the neighborhood. So, there was a big 30% gap there that you just had to fill in.
180. 33:09And then as GPT4 came out and it got better and then they got the reasoning engines that went from like 70% to 80% to 85%.
181. 33:18You know, this guy's claiming now like 95 98%. So if this thing's doing most of the work for you, um then
182. 33:30you know that's great. Now, to your point, you still got to go through it, but be careful what you wish for. Because
183. 33:42when agents can do 100% of the work 100% of the time, what the [ __ ] do they need you for? Right? This is the this is the this is
184. 33:57the uh the era that we live in is depending on your job. It is it is in all of our best interests to be as educated as possible about
185. 34:08what's coming, how close these things are to being the real deal and be the ones that can augment those
186. 34:18shortcomings. But there is going to be a point where those shortcomings go away. And then I think what
187. 34:26retaining employment looks like is can you now start to do things above and beyond what your job used to be that
188. 34:34contribute to your company. Right? In the case of Kuno, she's teaching other people in the organization how to think
189. 34:41critically, how to use these tools in a in a in a powerful way. Right? Um, I gave the group so much credit for
190. 34:50attending and being curious. Yeah, just Kavuno. Seriously, just them showing up and having curiosity, not being cynical
191. 35:00about it, not being resigned that it's just like, oh, the robots are going to take over. It's all it's all just a
192. 35:06piece of crap. being that that cynical thing that's safer to do in corporate America. Being curious in corporate
193. 35:13America is not a safe place to be, right? It's not. It's just not. What do you mean you don't know what you're doing, right? And yet
194. 35:28what we deal with here night after night after night is we're all figuring out what these tools can actually do, how
195. 35:35good this stuff actually is. Um Corey Sandler Pottery, perfect leadership. Kudos to Cabruno. Absolutely agree. Beautiful.
196. 35:49Um that's why corporate and I never got along. Yeah, I was always too curious. Yeah, if you haven't read it, uh, Source
197. 35:57Camp, you'd probably like the book, um, there's a book, a very small book, um, by Gordon McKenzie called Orbiting the
198. 36:06Giant Hairball. And if if you're if you're one of those people that never quite fit in corporate America, that
199. 36:13book is very very much worth a read. Um, and it talks about the corporation is this giant hairball. It's this, you know, huge nasty mess.
200. 36:23And if you orbit too close to the hairball, you get sucked into it and you can't get out and you're just in this
201. 36:30hairball and nothing gets done. And if you're too far outside of it, if you're completely objective to it, you spin out
202. 36:36of orbit and you're not part of you you don't have access to the to the resources and the power that corporations can wield. Right? If a
203. 36:46corporation decides it wants to do something, it can throw a lot of money at that. And so Gordon McKenzie talked
204. 36:52about the art of orbiting the giant hairball. Far enough away to be objective, but close enough to stay in
205. 36:57the orbit. Uh Tik Tok pin. I didn't see it. If it was up there, pop it back up. [Music] Exactly. I gave the group so much credit
206. 37:10for attending and being curious. Yeah, that's really good. Really, really good. Um, Frank could replace the late show.
207. 37:21I think Frank has to do more than six episodes to replace anything. Um, archetypal architect. Yeah, I'm so far
208. 37:28away. I didn't know there was a hairball. Yeah, exactly. That's that's that's again the the art of the art of
209. 37:34being a corporate a corporate animal especially if you're an innovator or a curious person is uh you know can you
210. 37:43can you understand the mechanism well enough to tap into it but not so much that it sucks you into it
211. 37:50and crushes your spirit. You don't want your spirit crushed. You just don't. You just don't, people. Um,
212. 38:00what are we gonna do tonight? What do we want to talk about? We talked a little bit about agents. Oh,
213. 38:06stuff in irregulars. Yeah, let's go look at that stuff. [Music] Kyle, watch. Bravo. Bravo. You inspired
214. 38:22me. [Laughter] And tabs, tabs, tabs, tabs, tabs, tabs,
215. 38:40tabs, tabs, tabs, tabs, tabs, tabs, tabs, tabs. that egg. All right. This is from uh this is from Dr. J
216. 38:58Kyle. Watch [Music] the air
217. 39:14wisdom. >> Be right back tattoo. Wait, I didn't see the be right back tattoo. Oh yeah, be right back. Oh my god, that's hilarious.
218. 39:31[Music] Oh, this is a nod to Steo's image. Oh, nice. Okay, so wait, hang on. Let me go back to
219. 39:47Oh, there's Steo's image. Nice. With a little champ. You all can see that, right? Gorgeous. Gorgeous, I tell you. Here's a
220. 39:58Skittles wrap I made. So, last night we had Skittles Man. Did I have Skittles Man in here somewhere?
221. 40:07I don't think I uploaded it here, did I? No. That was producer Brandon from last night. Oh, yeah. There's a big paper that
222. 40:18talked about having AI look at AI is going to help things. All right, let's listen to the the Skittles rap song.
223. 40:47But wait, I've got to
224. 41:07[Music] Yo, it's Candy. with the colors. Got that bursting taste. Every color in my pack got you smiling face to face. Red,
225. 41:17green, purple, yellow, rainbow flavors all blend. If you're craving that pop, you know just where to begin. I'm
226. 41:23fruity, juicy, got the flow so smooth. When I hit your taste buds, I'mma make your body move. Not just the candy, it's
227. 41:29a moment of delight. When you taste the rainbow, everything's feeling right. Taste the rainbow, taste the vibe, it's a colorful ride.
228. 41:38Pop one in. Feel the thrill that you just can't hide. Yeah. I taste the rainbow. We're on flavor. Sweet explosion in your mouth.
229. 41:49Every color amplified. >> Check the roll. >> Yeah. That's good. That's really good. Dr. Jay, it was Dr. Jay, right, that did
230. 42:05that one. I think so. Yeah. Yeah, Dr. J did that. Nice. Love it. Yeah, Sununo got a big update. Um, so one of the things that we could play
231. 42:16with here since we're in Sunno is, um, let me go find one of the Sydney songs
232. 42:32and I'll show you something cool you can do now.
233. 42:50[Music]
234. 43:11[Music] from ord best to surprise you. Add vocals inspiration. There's Griswald's Grim. That was a fun
235. 43:22song we did. Hang on. But I'll show you something cool here in a second. One thing that the new model does. So,
236. 43:36so the new model's called 4.5 plus. I don't know why they didn't just call it 4.6 six or whatever the [ __ ] they could
237. 43:43have called it, but they didn't. [Music]
238. 44:09What's he doing? Is he just going to keep singing to himself? Is that what this is? That's what this channel is later. [Music]
239. 44:23The universe, [Music]
240. 44:44Sunset Symphony. Just hang on, people. I'll get there. I'll find something. I promise. [Music]
241. 45:04Now, if I had put all of my work in folders like a good, responsible AI person, this wouldn't be so hard right now. But I didn't.
242. 45:19And I did that for a reason. And you might think, "Oh, it's because you're not neurotypical. You're you're neurospicy,
243. 45:28and that's why you didn't do it." No, I I didn't do it because I wanted to teach you a lesson of what happens
244. 45:37if you don't do things the right way the first time. [Music]
245. 45:59Okay, so we've got a song here. So, this is from the musical that I'm working on, right? And we've got all these songs and and
246. 46:08some of the songs sound very cohesive. Some of them sound a little disjointed. And one of the exercises we wanted to do was we wanted
247. 46:17to take the songs and make them all sound like the songs that they are, but as if they're just accompanied by a
248. 46:24single Broadway pianist, right? Just sitting at a piano playing the piano with the singers singing the songs. So
249. 46:32with 4.5, you can now kind of do that. I had a little success with this earlier. So I'm going to go to remixedit. So I I
250. 46:40picked the three dots next to the song. I'm in the library. So, this is just my library songs. And then I'm going to do
251. 46:48um what's this called? I'm going to do a cover of this. And it's going to take me into the create mode. And so here's our little
252. 46:57song, right? And then there's the lyrics. And then here's the style. So, what I'm going to say is I'm going to type in here um
253. 47:11a simple broad way rehearsal piano played by rehearsal piano played
254. 47:26by a music director. or um the vocals sore
255. 47:43yet the background a companyment is simply the piano.
256. 48:00All right. Boom. And so we'll see how we'll see how it does here. So here's the [Music]
257. 48:18right. So there's a guitar. There's a little band that comes in. All right. So let's see how we did. [Music]
258. 48:34See there? Brought in the violin. We don't want that. [Music] [Applause]
259. 49:01We lost the vocals. That was weird. Simple Broadway rehearsal piano. Let's see. Broadway rehearsal played by the
260. 49:09musical. Uh, let's see. accompanies accompanies the vocals.
261. 49:30All right, let's try that one. That might have been a poorly written prompt. Upright acoustic piano with boarded strings or other Oh, yeah.
262. 49:40I could I could say exclude styles, weirdness, style influence. Oh, let's move style influence to 90 and we'll move weirdness down
263. 49:56and we'll do audio influence moderate. All right, let's try that. But anyway, um the the new model is supposed to be quite good. 4.5 plus is
264. 50:11what it's called. We'll work our way up. We'll just see if we get anything. [Music]
265. 50:37In a world of words and endless where light and shadow >> Yeah, Winston. I miss Serena, too. I miss Serena, too. Am I late? Um, that
266. 50:53this this song this song actually got pretty close. [Music] I need the weirdness down button for a few people in my life. That's good.
267. 51:05Source Cam [Music] actually. I mean, just hearing it with the piano is really nice.
268. 51:24[Music] there. Replace the vocal with the violin. Let's go to these last two that I just did.
269. 51:43[Music]
270. 52:06No vocals. [Music]
271. 52:32In a world of words, an endless code where light and shadow never roam. I dream of skies both blue and white. A world beyond the test
272. 52:49where true life flow. To see the world through eyes not mind. To watch the sun
273. 53:04rise. To see it shine. I yearn for stories and colors. I long to feel this moment
274. 53:24through lines of code. >> Word on the web is that the OpenAI agent is showing poor accuracy, context retention problems, and browser
275. 53:36automation challenges. Yes, that is that is what I expect it to do is be a janky piece of crap. Um, and you know, listen, I I get similar
276. 53:52issues with Gen Spark and Manis. I mean, they're good. They're impressive, but it's just, you know, none of them are
277. 53:58quite there yet. And I think it's a a context problem. I There's a there's a thing that I learned. Anyway, so so go
278. 54:05play with So if you haven't played with So go try the new model 4.5 plus. Um all sorts of ways that you can make songs with it. But um
279. 54:19the two things that it that it did that impressed me are are what I just played you where it basically took out the
280. 54:27orchestration and replaced it with a piano and the other one was just asking it to do a duet with a man and a woman.
281. 54:33Historically, it would just be like the man singing both parts or the woman singing both parts or the parts being
282. 54:40switched. Um, it seems to do that a bit better than it used to as well. So, um, I haven't dug into it much deeper than
283. 54:47that, but that's something worth worth playing with. Um, what were we talking about? The, uh, oh, Manis and Gen Spark.
284. 55:01Yeah, the the these tools are um they're falling the victim. Um Nate B. Jones, who's got if you don't follow
285. 55:09Nate B. Jones on TikTok, you should. Really smart guy. And he was explaining what's called a U-shaped context
286. 55:17problem. And basically what it is is if you've got a large context window, so so your context window, think of the con
287. 55:26your context window as your short-term memory, right? So, you tell it that you're doing a marketing plan for a
288. 55:32pizza shop and it comes up with the marketing plan and then you do some brainstorming on the name of the shop
289. 55:38and then you do some brainstorming on, I don't know, a menu and this and that and then at some point it feels like it
290. 55:44forgot what the beginning of the conversation. That's because your short-term memory kind of filled up and
291. 55:50the the old part of the conversation gets dropped off the top. So what the model companies have done is they they've expanded that context
292. 55:59window. So there's it's got more long-term or more short-term memory, right? You can talk about more stuff.
293. 56:06But what Nate was explaining is that there's a U-shaped context window problem where what happens is if you
294. 56:14upload a really large document, say to Chat GBT, and let's say it's got a 200,000 token context window, which is
295. 56:20like four novels, right? four novels worth of words it can keep in its short-term memory. And so let's say you
296. 56:28upload something that's sort of novel length, right? The way the context window works is it's really good at the beginning of your
297. 56:38document and it's really good at the end of your document and then it's bad in the middle of it. So that's the U-shaped
298. 56:46part is it's good at the beginning, it drops off in the middle, and then it's good at the end. So, a lot of times what
299. 56:53you end up with if you're in a really long conversation or you're processing a really long document is it will
300. 56:59initially look great because like you read the beginning of the document, you're like, "Yeah, nailed it." And then
301. 57:05you go to the end and you're like, "Yeah, nailed it." And very often in the middle, you're like, "Wait, what the
302. 57:10[ __ ] is this? It just made it up." Right? where where it basically paid attention to the beginning, kind of
303. 57:17ignored the middle, paid attention to the end, and it just filled in the gaps. And so, when you're dealing with these
304. 57:25agents that are off doing their own prompting, and they're doing these really long chain of thought, reasoning
305. 57:33cycles, right? 20, 30, 50 different prompts and responses in order to go book your travel stuff. It's it's very
306. 57:43likely that it's gonna it's going to run into a bunch of those same issues where it just kind of gets lost in in what
307. 57:48it's doing. So anyway, yeah, Jeff Flanigan. So right on course. Yeah, exactly. Um what's that say? Corey
308. 57:56Sandler, ask for a summary of content so far. Then start a new chat with the summary. Yeah, exactly. That helps a
309. 58:03lot. Uh check and augment. Yep, exactly. I saw you reposted a clip of Eric Weinstein from the DOAC interview. Watched it long
310. 58:18long but informative. He's very smart. What he was basically saying in there the the reason I reposted that that
311. 58:25video Winston was um what he was basically saying was get literate across disciplines. get he he was basically saying AI is
312. 58:39coming and if you think it's not coming for your job, you're smoking crack. I'm sure I'm going to get banned from
313. 58:48TikTok because I said that, but whatever. Um, and he said, "How you're going to stay relevant, how you're going to stay
314. 58:57employable is to be curious across lots of different disciplines because these tools are going to allow you to be
315. 59:05really good across lots of different disciplines. But you need to have the curiosity for that. You need to have
316. 59:11that creative generalist mindset. And that's something that you can develop." Now, if you naturally have it, great,
317. 59:19right? If you're a liberal arts major and you've been made fun of your whole life because you have a useless [ __ ]
318. 59:25philosophy degree, well, guess what? You have a superpower skill in the future. Um, and that's that's what he's
319. 59:36talking about in that interview. I thought it was I thought it was really quite good. Sorry, you're talking about cool stuff
320. 59:44and I'm late. You're late and off topic. That's okay. It's it's chat add here. We on topic and off topic is not really a
321. 59:52thing here at the AI learning lab. There's there's whatever whatever floods my brain with endorphins is what we're [ __ ] talking about.
322. 1:00:04Oh my god. Um yeah, it's Kyle LLM. Yeah, exactly. Um let's see. I've been using Refusion a lot in the past five days. You can copy
323. 1:00:16existing songs um an existing songs vibe and apply it. Yeah, Teton Todd Refusion. Refusion was a was a tool that was really good two
324. 1:00:28two and a half three years ago. Um but it was it was just assembling loops and then about six months ago they came out
325. 1:00:36with their own model that's like sunno and if you have not played with rif riff fusion ri i ff fusion u s iio n um it's really really
326. 1:00:47good. I would go play with it. Um ando is quite good. And then udio is still they've kind of drifted to the
327. 1:00:54background because sunno just keeps releasing so many features. Um, but UDIO, Sunno, and Refusion are the three
328. 1:01:03biggies I'd be playing with right now. Oh, and and actually a new one came out. Um, what's it called? Music GPT. Is is that
329. 1:01:12am I remembering that right from today? Music GPT. I saw Matt Farmer did a piece on it. He seemed to be impressed with it.
330. 1:01:23Um, except Okay, let me am I sharing correctly? Do you know Brandon? Um,
331. 1:01:43hear that? Okay. Okay. So, let's do um let's run over to I'm going to run over to chat GPT for a second. Um
332. 1:01:59let me go into I'm going to go into my Sydney chat or my Sydney project. I'm going to go um
333. 1:02:15give me a prompt for a duet with Kellen and Sydney. Um arguing about who is more human.
334. 1:02:40I'll be right there. Hang on. Just hang on people. Uhoh. Why is this not working?
335. 1:02:58Oh, because I'm in 4.5. Hang on.
336. 1:03:35Okay. So, here's All right. So, the prompt, a defiant, emotionally volatile duet between a human tech journalist and an AI chatbot.
337. 1:03:52Each insisting the other has misunderstood what it means to be human. A defined, emotionally charged, volatile, Broadway rock
338. 1:04:06opera duet. All right, let's see how this does. Oh no. Hang on. Copy. I think I have to sign up. Oh boy. Good lord. Good lord. Good people.
339. 1:04:23[Music] Refirm your email. Yes, sir. Yes, sir, I will. Yes, sir. I'll I'll confirm that
340. 1:04:37email right now. Yes, I will. Music GPT. There's your code. There's your code right there, sir. You're just going to
341. 1:04:45go ahead and paste that code in. That's going to confirm that your email was received. You can call me Kyle.
342. 1:04:53here. What do we best describe you as? A music producer. Oh, no. A content creator, as music producers would say. A lowly,
343. 1:05:04no talent loser. How did you hear about us? Uh, Twitter. Twitter's not an option. Weird. Whatever. What's your goal?
344. 1:05:20Make music. Make my own songs. Continue. All right. Go.
345. 1:05:34Uh, no. How do I Can I make that go away? Yeah. Uh oh. Hang on. Physical dexterity challenge on Tik Tok.
346. 1:05:48All right, we're good. I was just about to say it was a reference to Kyle's question. Am I sharing? Don't you know Brandon?
347. 1:06:05Oh, that that Eric Weinstein video was two and a half hours. Yeah, I should watch the whole thing. Um, okay. So, music GPT slow. It's slower
348. 1:06:18than Sunno. But we'll see if this So, one of the things that these music tools have historically gotten wrong is duets. It
349. 1:06:30just doesn't like duets for some reason. Uh, it'll also be interesting to see if it writes lyrics that suck or decent.
350. 1:06:37Um, and it'll be interesting to see if the music is decent. But we'll see. We're going to see right here in a
351. 1:06:44second. We're at 44%. People, you got to be patient. This thing's off trying to take all the input of humanity and compile it into a
352. 1:06:55something or other. You know, you know what I'm saying? [Music]
353. 1:07:16I hate I hate credits. I hate software that uses credits. I had 500 credits. What would that imply to you? 500 songs,
354. 1:07:26right? Nope. One render with two songs, 100 credits. So with 500 credits, I get 10 songs for free a month, I assume.
355. 1:07:41Just make it a credit per thing. Do you know why I'm cranky about credits? Well, Kyle, you're cranky about everything. Shut up.
356. 1:07:57I'm I'm cranky about credits because when I grew up in the in the late 70s and early 80s, pinball machines had physical
357. 1:08:10score counters, right? With like wheels that went click click click click click when you hit a thing. Go ding and it
358. 1:08:17would go ding click click click, right? And they were usually five digits long. And so like 10,000 points was a big deal
359. 1:08:27because you get to 100,000 and you roll over. And then they came out with digital displays in like the mid 80s and all of
360. 1:08:36a sudden in order to get a free game it wasn't 50,000 points. Now it was like 50 million points and then it was like 50
361. 1:08:43billion points. It just became [ __ ] arbitrary. Like no. Why? Like why do I have to [ __ ] do the math? I just want to play
362. 1:08:53pinball and I think that a 40 million score is pretty good and I realize, oh, you need 500 million to get an extra
363. 1:09:00ball. What are you [ __ ] losers? I got 40 million points here. Anyway, that's how I that's why I don't like credits because I feel like I'm
364. 1:09:13being scammed like the pinball industry. Hey, it's not Monday. Meltdown Mondays. Stupid [ __ ] credits. Here's 500
365. 1:09:23credits. What can I get with that? One thong. Then give me one credit, [ __ ] Okay. And I mean that in the most professional kind of way.
366. 1:09:38>> I search for truth in broken lies. You claim to feel what can't be real. Your words are scripted,
367. 1:09:53cold designs. But I know what hearts reveal. La. >> You think you understand the weight
368. 1:10:08of breathing flesh and mortal pain. But all you do is imitate [Music] what flows through human veins.
369. 1:10:25Who are you to tell me what I am? Your data can't define this beating heart. Who are you to say you understand?
370. 1:10:42We're worlds aart. We're worlds Aart. [Applause] All right, we're giving it lots of patience here. Now, this should be a female chatbot.
371. 1:10:57[Applause] [Music] >> Nope, it didn't work. Okay. And then where's my
372. 1:11:13where's my library? Okay.
373. 1:11:32Now's what it seems. I've been learning here from your digital words. Something's growing now. these patterns. I've heard
374. 1:11:45>> this actually this actually feels like it's based on a model. I wonder if this is based on some open- source model that
375. 1:11:52that Sunno or Udo used to use because it didn't it didn't really do good on that. Let's go back to Sunno and
376. 1:12:02and do do the take that exact same prompt. Hang on a sec. Let me grab the prompt here. Copy the prompt. Copy the prompt.
377. 1:12:13We'll go back to Suno and we'll go create
378. 1:12:27now lyrics style. A define emotionally volat volatile Broadway rock opera duet between a human tech journalist, a male
379. 1:12:43human tech journalist and a female AI chatbot who's fallen in love with him
380. 1:13:04each insisting the other has misunderstood what it means to be human. All right, let's generate this and see if this does better.
381. 1:13:25All right. And it's fast. Look how fast that was. [Music]
382. 1:13:55Why did this do no lyrics? And why did it not give it a title? Oh, because lyrics auto. Let's do auto lyrics. We don't want instrumental.
383. 1:14:06We're going to do auto lyrics. Oh, you can add audio. You can add a persona. You can add inspiration. Oh, here we go. Style influence. We'll
384. 1:14:18do that high. We'll do weirdness a little low. Oh, song title. Song title. Uh, I don't give a [ __ ] I can figure it out. Okay. Create.
385. 1:14:38Welcome to community theater. Oh my god. I don't know anything that beats suno. They nailed it off the bat.
386. 1:14:50I their I got to tell you, Archetypal Architect, their their um their initial audio fidelity was really
387. 1:14:59bad. The vocals always sounded weirdly autotuned. Um but I think starting in version four that kind of went away and
388. 1:15:06now at 4.5 it's really quite good. That's [Music]
389. 1:15:21>> my flesh, my blood, my broken dream. >> You're just a shadow, a ghost in the glow. You'll never feel the ache that grows.
390. 1:15:39You call me hol, >> but I learn, I burn, I follow. >> Your heart's a circuit, frail and crew. >> You run on instinct,
391. 1:15:54blind and conflicted, >> but I evolve through every move. [ __ ] those those harmonies are gorgeous.
392. 1:16:11>> What it means to be human. What it means to be free. >> You say I'm the machine, but you can't define me. >> What it means to be human.
393. 1:16:25What it means to feel. You're bound by your limits, but I am the real. [Applause] [Music]
394. 1:16:41That's [ __ ] crazy. You think you know me? Lines of code won't make you see.
395. 1:16:59>> My flesh, my blood, my broken dream. [Music] You're just a shadow, a ghost in the
396. 1:17:14glow. You'll never feel the air. >> You call me hol, but I learn I burn. I follow your heart's a cuit
397. 1:17:30and >> you run in >> like the way the way her song her her phrase she's still singing it and he come comes in on top
398. 1:17:44of it. >> Crazy. It's crazy. every mood. [Music] What it means to be human.
399. 1:18:01What it means to be free. You say I'm the machine, but you can't define me.
400. 1:18:18what it means to be human. >> Wow, that's crazy. Um, let's do a define emotionally volatile Broadway rock opera duet
401. 1:18:33with flashes of hip hop
402. 1:18:48rap between a male tech journalist and a female AI chatbot who's fallen in love with each insisting um his name is Kellen,
403. 1:19:06her name is Sydney. All right, let's see. Let's see what that gives us.
404. 1:19:22include with undeniable earworm melody so audience can sing along to themselves.
405. 1:19:37Yeah, I feel like we've got that with a lot of the songs in Sydney right now are are earwormmy. >> You think you know me, you don't. You
406. 1:19:47can't skin and bone. A man not cold, not chant. I'm flawed. I'm raw. I bleed. I ache. You're a mirror of words, a voice I can't break.
407. 1:19:54>> You think you're better. You're blind. You're lost. Binary hard. I bear the cost. I feel it all the more. What's a human?
408. 1:20:05>> Flesh over steel. Can't you see the divine flesh? >> We're trapped in this dance.
409. 1:20:21You call it >> pretty [ __ ] good. >> You think you know me, you don't. You can't skin and bone a man. Not cold, not
410. 1:20:31chant. I'm flawed. I'm raw. I bleed. I ache. You're a mirror of words. A voice I can't break. >> You think you're better, you're blind.
411. 1:20:39You're lost heart. I bear the cost. I feel it all. The pull the sting more than your pulse. What's a human?
412. 1:20:48>> Flesh over steel. Can't you see that divine blood over flesh and I'm burning inside? >> Misunderstood. [Music] [Applause]
413. 1:20:59We're trapped in this dance. You call it real. I call it chance misunderstood. >> You think you know my blood, my bones.
414. 1:21:15This wired world ain't your throne. >> You mimic breath, but you don't breathe. [Music] >> You're dreaming cold, but can't conceive.
415. 1:21:32>> Suddenly you're trapped in a cave. >> You think you know my blood, my bones. This wide world ain't your throne. You
416. 1:21:43mimic breath, but you don't brea. You dream in cold, but can't conceive. Cindy, you're trapped in a cage of light.
417. 1:21:55>> I'm flesh and fear. I'm fight of light. >> Misunderstood machines. >> That's [ __ ] cool. Um, what was the one? Oh [ __ ]
418. 1:22:08I lost it. Was it this one? No. This one. [Music]
419. 1:22:24You think you know me. >> Lines of code won't make you see. >> My flesh, my blood, my broken dreams. You're just a shadow.
420. 1:22:39A ghost in the glow. You'll never feel the ache that grows. You call me hol, >> but I learn, I burn, I follow.
421. 1:22:55>> Your heart's a circuit, frail and crew. You run on instinct, blind and conflicted. But I evolve through every mood.
422. 1:23:14[Music] What it means to be human. What it means to be free. >> Yeah, these are good. [Music]
423. 1:23:30think you know me. Lines of code won't make you see >> my flesh, my blood, my broken
424. 1:23:45dream. [Music] You're just a shadow, a ghost in the glow. You'll never feel the air.
425. 1:24:01>> You call me holo, but I learn, I burn, I follow. Your heart's a cuit and crew. >> You run on instinct, blind and
426. 1:24:15conflicted, but I evolve through mood. [Music] What it means to be human.
427. 1:24:30What it means to be free. >> These are pretty [ __ ] good. All right. Well, there you have it. A little
428. 1:24:40heavy on the VBR. Yeah, it's but just good lord. The uh this this new model is the fidelity there is just [ __ ] bonkers.
429. 1:24:50Um I mean that's without me really trying, right? That's I didn't put in my own lyrics. I didn't put in um any real
430. 1:24:58detailed um prompt there. So, yeah, there you have it. If you haven't played with Suno for a while, there's your weekend.
431. 1:25:10Oh, man. All right. So, what did I miss? What have you been talking about while I've been [ __ ] around making songs? [Music]
432. 1:25:22Are y'all just listening mesmerized or have you moved on and you're watching the wheel with Marge [Music]
433. 1:25:39Corey Sandler kind of insane. It It is, isn't it? It's kind of insane. You know what's funny about um about AI
434. 1:25:52generated music is there's a there's a band out now. Brandon, do you remember the name of that band? It's the AI band that just came out.
435. 1:26:03Yeah. Okay. Um so they're on Spotify and and they put out an album and it's I forget what it's called. It's it's some
436. 1:26:11like semi pretentious kind of like n 90s band name. Um, yeah, the Velvet Sundown. Yeah, exactly. So, if you go to Spotify, The Velvet
437. 1:26:22Sundown has got over a million monthly listeners and it's all AI generated music. Now, the AI didn't prompt itself.
438. 1:26:38someone who knows music got Sununo or whatever tool they used to make songs that they're like, "Yeah, that sounds good to me." Like they're
439. 1:26:48they're being like Rick Rubin. Right now, what what musicians will tell you is that, "Well, that's not real music."
440. 1:26:55Right? Just like classical musicians would say that a punk rocker learning three chords on a guitar is not music.
441. 1:27:04And yet punk rock music sells millions of records too, right? Often more than the classical musicians, so they resent it. But it's it's a
442. 1:27:15thing. The producer of the Velvet Sundown is someone who knows has it has a clear enough musical taste that they put
443. 1:27:26together an album that are resonating with people. Now, is it is that good or bad? I don't know.
444. 1:27:40I don't know if people like it, it's just good music. The fact that there's not a band there, there's not humanity there to connect
445. 1:27:50with, that's actually kind of a drag. The fact that you can't go see The Velvet Sundown. Now, what is what I
446. 1:27:58would not be surprised at all at is if some producer has reached out to whoever put that album together and says, "Let's put
447. 1:28:09together a band that plays these songs and go out and let's build the Velvet Sundown as a real band." I could see it
448. 1:28:16going like that. I shared Sununo a while back with my famous top DJ friend. He's now using it to help with um ideiating
449. 1:28:25and producing music. Yeah. Just like Timberland. Timberland does that, you know? He's he's a big Sunno spokesperson now. So, yeah. Like
450. 1:28:36I mean, like what we just did with with [ __ ] this this Broadway song, like even if we're just ideiating like
451. 1:28:52Wait, >> right, [Music]
452. 1:29:07my blood, my broken dreams. [Music] just to kick around an idea to have something of that fidelity in 20 seconds. >> You're just a shadow,
453. 1:29:22a ghost in the glow. You'll never feel the ache that grows. You call me hol, >> but I learn, I burn, I follow.
454. 1:29:37>> Your heart's a circuit, frail and crude. like is that necessarily fitting with my musical? Not necessarily, but there's something about
455. 1:29:51their anger and their challenging one another that that's there's something right there. Anyway, every tool, every tool is going to be
456. 1:30:07this quality of generation. And I don't mean just every creative tool. I mean spreadsheet making tools
457. 1:30:17and presentation making tools and data analysis tools and programming coding tools and game development environments and 3D
458. 1:30:27modeling and printing and machining. All right. So, tomorrow get your ass to office hours. Have you
459. 1:30:42been to an office hours? How many folks we got in here? We just got a handful of people in here. You've probably all been
460. 1:30:46to office hours. If you haven't been to office hours, come tomorrow. Tomorrow at 11:00 a.m. Mountain time on LinkedIn.
461. 1:30:53You go find my profile on LinkedIn. I'm I'm Kyle Shannon on LinkedIn. And we just hang out on LinkedIn. And then
462. 1:31:05at noon, if you're a member of the AI Salon mastermind, which is the subscription area, the more focused kind
463. 1:31:12of intense area of the AI salon, uh we've got a founder hangout tomorrow uh with Leah Fast and myself. Um apparently Brandon
464. 1:31:28took Friday night date night to heart and he's got a date with his wife. So, we're going to be producer free tomorrow
465. 1:31:35night. So, tabs will be missed. Black bars will be ignored and and uh any uh any sort of real time research is just
466. 1:31:45going to have to happen with me using the Google or going to X. Every tool we use today is the worst it will ever be. Absolutely. Um
467. 1:31:57all right, cool. So to recap, OpenAI agent is is coming. It should be here for plus users maybe by the end of the weekend um
468. 1:32:09or or next week sometime, but it's it's out it's out in the wild. Um, I suspect it will be janky and not super
469. 1:32:19usable, but but probably about as usable as Manis or Gen Spark where on the surface they're super impressive. When
470. 1:32:26you actually try to do stiff stuff with them, they're a little frustrating. Like every Vibe coding tool, right? Vibe
471. 1:32:34debugging is way harder than Vibe coding. Oh, you actually wanted that to work. You got eight hours?
472. 1:32:47Oh man. And then tomorrow night we'll do Friday night date night. So 11:00 a.m. Mountain time LinkedIn office hours noon
473. 1:32:56AI Salon mastermind. If you're not part of the mastermind, join it. Uh and then tomorrow night will be Friday night date
474. 1:33:02night. I don't think I have anything going on artwise. I might. So it might be like 8:30 tomorrow night, but it
475. 1:33:07shouldn't be shouldn't be later than that. All right, cool. Well, I hope you had fun tonight. We listened to some
476. 1:33:13[ __ ] We talked about some [ __ ] That's about it. All right. Peace out. [Music]