7/21/2025 - Demystifying JSON for AI Prompting

Transcript

1. 0:00Dum dum dum dum dum. All right, now we're live on YouTube. Howdy, Dr. J. What's shaking? What's going down? [Music]
2. 0:14How about if we hit chords where the fingers are on the actual strings? [Music]
3. 0:33[Music] Is this a place I can rest my forehead [Music]
4. 0:47to gather my thoughts in sweet? [Music] Is this place where the feelings aren't dead? [Music]
5. 1:01From an over exposure to violence. This is place I can slowly face. The only one I truly can know. These are tears from a long time ago.
6. 1:18Got these tears from a long time ago. I need to cry. 30 years old soul. These are tears from a long time.
7. 1:34Go. [Music] Champy. Are you singing anymore? What's up with you? What are you distracted? Chimpy,
8. 1:48you singing? [Music]
9. 2:05Good evening everybody. Welcome, welcome, welcome, welcome to Meltdown Monday. Uh, according to Tik Tok, we have two
10. 2:13people in here. I think we have more than two people in here. [Music] But if you haven't shared the live yet,
11. 2:21what are you doing with your life? Let's get going. Let's get going. Share the live. I listen. This is going to be exciting.
12. 2:29We're talking about Jason tonight. Not Jason. JSON. JavaScript object notation. It's gonna be so fun. It's gonna be awesome. It's
13. 2:40gonna be so good. I think maybe what we'll do to start out is we'll go over to Suno and we'll write a song about JSON [Music]
14. 2:59using JSON. Oh my god, you're going to love it so much. You're not even going to believe it. It's going to be so awesome. It's going
15. 3:09to be so awesome. And if you're sitting there thinking like, I don't know what you're talking about. Well, that's because you don't have a
16. 3:21mathematics degree from Stanford or MIT. God, that must suck. It does because I have a I have an acting degree from Penn State. So,
17. 3:35I know what it's like to feel inadequate mathematically. I do. I do. Do I feel for people that are not polymaths?
18. 3:51Not particularly. I don't really feel for polymaths either. I don't even know what the term polymath means. I think it means of many maths.
19. 4:03Is that like they do some algebra and some geometry? Little trig. little trig thrown in there. I don't know. Isn't all of generative AI
20. 4:17just basic algebra? Is it a polymath that did that or is it just an algebra guy? Like is it if it's just a differential calculus
21. 4:26specialist that was like, "Ah, we're going to make the internet the AI thing." I don't think that needs to be a
22. 4:33polymath. I think that could be a mono math. But here's the thing. Maybe they use the one math and then all of a sudden they
23. 4:41realize we probably need some other maths. You should talk to Jimmy. What's going on with Jimmy? He's a polymath. A what? A poly many maths?
24. 4:55Those people exist? Yeah, they gave him a name. Anyway, you don't need to know math.
25. 5:10Not not for this channel. Not for this channel. We We have a dog that sings. It's sort of like polymath with K9.
26. 5:21There's a number in it and a and a you you could argue that the K is a variable and the nine is a numeric integer. I think
27. 5:32I think I learned that in seventh grade polymath. single math algebra I don't whichever one you learn in ninth grade
28. 5:42when when you're destined to go to school for acting whatever class that is polymath equals multi- potentialite exactly
29. 5:56I went to high school with polymath she was great I you know Paulie um I knew Paulie Well, in grade school and uh
30. 6:07boy, she was a card. She was a card. You know what? Speaking of cards, she would have those flash cards with like the
31. 6:13fractions on it and I was like, "Oh man, she's going somewhere." Turns out it was MIT. Yeah. Yeah. Yeah. She's got a career, Paulie.
32. 6:25Sharon, and and I'm an accountant. You use QuickBooks math. That's a That's a whole different branch of mathematics.
33. 6:35All right, we'll switch this over to here. Um, all right, we'll do that. Let's see. I just heard a Coldplay concert on screen.
34. 6:46Yeah, she's she's famous now. Who knew who knew a mathematician could get that famous that quickly? One little Heimlick maneuver, huh?
35. 7:01Ah, okay. I've got my coffee. I got my my spritzer, my Steltza water. I don't quite have enough ice in it, but
36. 7:15you know, it'll do. It's It's cool. It's cool and chilly. [Music] Yeah, that's what a calculator was for.
37. 7:29That's what Excel was for. how that's what AI is for. Um the the the JSON that we're going to talk about tonight, why we're going to
38. 7:45talk about it is my Twitter feed is being taken over with AI influencers going, "Jesson is the secret to great VO3 prompting.
39. 7:58you should use JSON for all your prompting needs. JSON is everything. And then they post some JSON there like they
40. 8:06[ __ ] wrote it. They didn't. They didn't. I'll show you how to do it tonight. We're Here's what we're going to do.
41. 8:14We're going to teach you a little quasi coding, but you don't have to learn coding. We don't That era is over.
42. 8:25You don't need to code anymore. You can you can if you're one of those polymaths and you're like coding is relatively
43. 8:33important to retain as a practice so that we understand the foundations that create computer applica fine
44. 8:42fine if you want to do that you can do that I'm just saying you don't have to but when someone pops into your feed and
45. 8:52they're all like you should use JSON JSON do you know what JSON O is Well, look, it's all these fancy codes. Look
46. 8:59at all the squiggly things in there. I did that. That's That's what's going on on Twitter right now. It's really [ __ ] annoying.
47. 9:07So, we're going to demystify that tonight. [Music] We're going to debunk those jerks.
48. 9:23My first experience with Gen Spark was dismal. Asked for 16 slides and it failed to find 14 and got the other two wrong.
49. 9:35The images that is Oh, that's interesting. I tried I used GenSpark today and to decent success just doing some research
50. 9:47on uh effectiveness of video and pharma marketing and it it did pretty good but it it hallucinated you know a decent
51. 9:57amount of the cit I asked it to hyperlink its citations and there was some hallucination in there but um but I also I went to Manis and
52. 10:09tried to have Manis do it and it got halfway through the task and said, "You're out of credits because you're a
53. 10:14loser and don't pay us. So, we're not going to finish your stupid job." I thought that was a little aggressive for
54. 10:21like a a modal a popup window. We're not going to finish your stupid job. Wow, man. Really? [Music]
55. 10:37If you're new here, by the way, um what was I gonna say? Uh this is just how I act. I'm just a dumb dumb. I'm like the
56. 10:45village idiot and uh I'm here for you. Do we have We don't Producer Brandon's not here. I have a feeling producer
57. 10:53Brandon might be napping. Sometimes producer Brandon goes napping and I have a feeling that could be the case
58. 11:01tonight. Dr. Jay's in the house. Kyle, did you get access to chat GPT agents yet? Not as of an hour ago,
59. 11:10but let me try right now. They said it was supposed to be Monday, but I I didn't have much confidence in that.
60. 11:1710 print hello world. 20. End. No, you forgot the loop. 10. Print hello world. 20. Go to 10.
61. 11:32Go to 10. 10 times. 30. End. Sea slug of Doom is a master programmer. You can tell that he's doing RadioShack
62. 11:44basic programming. That's that is the last real programming I've ever done. I mean, HTML scripting, I guess, frankly.
63. 11:54Um, basic programming is scripting, too, isn't it? I checked and I don't have access yet. Anxiously awaiting it. John, it annoys
64. 12:05me how obsessed they are with making them use our computers. Why not just give them read access? Yeah, I know. Well, to APIs. They're
65. 12:15they're going to I mean uh uh CETBT agents is is going to or does have access to APIs through connectors there. there. Listen, right now
66. 12:32because this agentic behavior, things like Manis and and Gen Spark and and OpenAI agents, because we're not used to what they do, um
67. 12:45they they are just by definition going to be way more restricted than they will be in the future. We'll we'll get used
68. 12:51to this. This is this is very much akin to people not putting um not putting their credit cards on the internet in the '9s [Music]
69. 13:16through blue telescope. Looking at the world tonight through blue telescope. I wish I may, wish I might not see what
70. 13:31I see. Sheet metal on sheets of ice. Looking through this blue telescope
71. 13:46down a moon struck a road tonight. [Music] The heavens, the heavens,
72. 14:01fast and wide. [Music] D. [Music]
73. 14:19I don't know those chords. Um, okay. John, what do you mean? What did John say? John, I said it. I said it. What did he
74. 14:30say? Honestly, I'm this close to canceling my OpenAI subscription. Yeah, I know. Hey, you know what's funny? My
75. 14:38my co-founder canceled her OpenAI script des uh subscription yesterday. I said, "You probably shouldn't. They're
76. 14:45probably just upgrading servers." She was like, "Ah, screw it. I'm fine. I'm going to use Gemini." This is So,
77. 14:53we're Hang on. I got to sneeze. Um, sorry. Le Mart. Um
78. 15:07yeah, everything everything is gone from my brain. Oh, open AI subscriptions. Um we are absolutely 100% at a point with generative AI where
79. 15:22you don't need to have a subscription to almost anything. The the one exception to that is video generation tools. Those
80. 15:27are still expensive, but the free tiers of essentially all the tools out there are really good. They're really good.
81. 15:36Most of most of what most people need to do, you can just do with the free [ __ ] Um, but that said, if you want to play with the new stuff,
82. 15:46it's good to have, you know, some sort of subscription and then you sort of figure out, well, what do I want to do and what do I want
83. 15:53to use it for? But I think gone are the days where you you feel like you have to have hundreds of subscriptions or you know dozens
84. 16:03anyway. Um because all the stuff's just so good now. H where's your AGI prediction? Um listen I here's the deal. I think that
85. 16:18so so you know AGI is the point at which these systems can that are as good or better than most humans at most jobs. um
86. 16:29what's been released in the wild right now is not that um the fact that the frontier model companies stopped talking about AGI two or three
87. 16:43months ago and now exclusively talk about AI ASI which is artificial super intelligence where these things are
88. 16:50smarter than you know the smartest people in all disciplines at all things or one definition is sort of smart
89. 16:56smarter than all humans individually. Another definition of super intelligence is smarter than the collective
90. 17:03intelligence of all of humanity. Um, I don't think we'll get there for a while. But I think that I think that AGI has
91. 17:11effectively been hit in most of the frontier labs. They just haven't released it yet. Um, maybe for safety
92. 17:20reasons. I don't know. I don't know. I also don't care. Like for me, my AGI moment was November of 2024. last year in November when 01
93. 17:37whatever 01 preview 01 when the first reasoning model came out from open AI that could do multi-step sort of
94. 17:46internal conversations and it just got better and it and it had kind of I don't know 130 IQund I think it was 120 or 130 IQ
95. 17:58and what folks were saying was you know it can it can it can start to solve PhD level problems. And as I sat there looking at the prompt
96. 18:09hole, as we call it here, I thought uh what problems do I have that rise to the level of a PhD level problem? And the
97. 18:21answer was essentially none. And I realized that I was the bottleneck. So for me personally, the kinds of intellectual circles my
98. 18:36mind runs in, like the circle of thoughts that I hang out with, 01 was was the first time a computer it
99. 18:45felt like it exceeded that for me. And I feel like everything since then has just been increasingly beyond where I was.
100. 18:55Now, my initial instinct when that happened was to get all pissy and mooney and like, "Oh, I'm so dumb." And then
101. 19:02what I realized is, "No, no, no, no. You're where you are. These things are smarter than you." There's two ways you
102. 19:10can interact with that piece of knowledge. You can say, "Oh [ __ ] the robots are smarter than us. What do we
103. 19:16do?" Or you can say, "Hey, I now have a collaborator that's way [ __ ] smarter than me that I can leverage and it can
104. 19:27become my superpower. It can be like a jetpack for my intellect." And so that's the path I've gone down is that when you treat AI like like
105. 19:39it's your partner. And this is about your ideas and expressing your ideas and getting AI to amplify your ideas. For me, the world
106. 19:49opened up. And so the smarter it gets, bring it the [ __ ] on because I'm just going to give it more of my ideas and
107. 19:54make my my ideas better and better and better and better and better and better and better. And people will be like,
108. 20:00"Where'd you come up with that awesome idea, Kyle?" And I'm I'll be like, "I came up with it in my big fat brain."
109. 20:07And then I use this big fatter brain to make it better. Thank you. Thank you for the gifts. Oh man, they won't tell us once AGI is
110. 20:17reached. It will have too much of a ripple effect. I I don't know, Danielle. I I don't I don't think people will care.
111. 20:31And and here's my evidence for that. They don't care. Now, the people that come to this channel care, the early adopters care,
112. 20:45but most people don't know and don't have a whole lot of interest in what AI actually is and what it can do for them, right? I gotta learn
113. 20:56something. That feels like work. Nah. It's thievery. It's the world's greatest plagiarism machine. I hear it hallucinates.
114. 21:05Oh, I'm not smart enough for that. Whatever the [ __ ] trope is that they use, there's just a lot of bitching and
115. 21:11moaning about how much they don't care about AI. I think if Sam Alman comes out and says, "We've declared AGI,"
116. 21:20people will be like, "Okay, haven't you been talking about that for a while? Didn't you do that last year?" I don't think people will care.
117. 21:28Claude tried to trick me the other day. Straight up deceptive. Interesting. The onboarding getting API keys for LLMs
118. 21:36and tools. That's what you're missing. Yeah, I I would agree that the the connectors right now in OpenAI suck.
119. 21:45And so does M MCP servers in cloud. Like everything right now is too geeky. It'll it'll get better. um the fact that you know that what an
120. 21:56API is. Most people don't know what an API is. They tend to just not feel there's much competitive advantage. Who's they?
121. 22:08But I still have to hit send cover. Next week I'm hosting a GPT maintenance clinic for my team. Gosh darn model drift. Wow, that's a drag.
122. 22:23So, what what Chris is talking about there. So, remember a year or two ago when we started making these custom GPTs
123. 22:31and we thought we were all going to be rich because Sam Alman promised us the GPT store and we could monetize these
124. 22:37things and then that never happened and they don't give a [ __ ] anymore. Yeah, that Well, all those GPTs we built, they
125. 22:44don't work right anymore. the image model changed, the underlying models changed, they retired the models that we
126. 22:50wrote them on. Um, so if you've got old GPTs, custom GPTs, and they're good, you know, it might be worth going in and
127. 22:59testing them again and seeing if they work, you might need to rewrite your prompts. That's fascinating.
128. 23:06Fascinating. Fascinating. Fascinating. So, how many folks We got some more folks here. So, what we're going to do tonight,
129. 23:13it's Monday, so it's very possible it's going to be meltdown Monday. Let me see if producer Brandon's in the house. I
130. 23:18think he might not be yet. No, that's all right. We are producerless right now. I have a feeling he's sleeping with the kids. Um,
131. 23:27let's see. Hang on. So, we're gonna talk about Jason tonight. Jason. Jason. Jason. I think I think geeks call it JSON. I
132. 23:43call it JSON because it doesn't sound like a dude. Uggh. I wish I could send more words. Oh, I know. Yeah, it's it's annoying here.
133. 23:55Um then tried to hide the calculator as a hard code. Kavuno source camp experienced model
134. 24:09drift for the first time today in a GPT. Fascinating. [Music]
135. 24:36I prime my face from my wrinkled sheets. [Applause] Just hoping it was all a dream. [Music] Paint in my head and the hole in my
136. 24:51heart. says it was more than just Jim B. [Music]
137. 25:07Um, let's see. The GPTs don't have chat memory. That's probably a good thing. I would think you don't want your custom GPTs to have
138. 25:20memory of all of your stuff. I think though you can turn on and off memory for custom GPTs, can't you?
139. 25:28I haven't I haven't made one in a while, so I don't know what's going on in there. Um Oh, so listen. Next week
140. 25:36we have a big week next week and I got to get my [ __ ] together. Okay. So, next week I am doing a back to
141. 25:48basics AI crash course starting on Monday the 28th and going until um Friday the 1st. And so
142. 26:04Monday's going to be what the hell is AI? Why should I give a [ __ ] Why is it such a big deal? How does it work?
143. 26:10Remember a lot of those charts and graphs I have about how all this stuff works? We're going to do that. We're
144. 26:14going to talk about why it's a big deal, compare it to the the worldwide web. Uh then Tuesday night is going to be
145. 26:23basically a chat GPT 101, like what is this chat GPT thing? What's the chat hole? How do we use it? Wednesday is
146. 26:31going to be creative. So images, music, videos, all that sort of [ __ ] Thursday is going to be bringing it all
147. 26:40together like use cases, figuring out which tool to use when. And then Friday's going to be an AMA and ask me
148. 26:47anything. And the format's going to be really fun. The format. Do I have a I think I have a different graphic here. Hang on.
149. 26:56That's not what I wanted. There it is. Um I don't know if you can see the uh Do we have the format there?
150. 27:05No, the format's not on that graphic. That's okay. I'll just put the the other one back up. Um, the format's
151. 27:15going to be like this. So, it's for the people that come here all the time. Nothing changes, but what it's going to
152. 27:23look like is this. From 8 to 8:30, it's going to be Tik Tok only. We're going to have a Tik Tok pre-show for half an
153. 27:30hour. So, we're just going to hang out on Tik Tok, shoot the [ __ ] maybe talk about what we're going to talk about,
154. 27:36maybe answer some questions. welcome new people. If there's new people there that want to hang out with us, great. If it's
155. 27:41just us irregulars, great. And then at 8:30 all next week, Tuesday, Monday, Tuesday, Wednesday, Thursday, Friday, at 8:30,
156. 27:52we go live on YouTube. All right? And these are going to be kind of a one-hour session followed by 30 minute Q&A.
157. 28:04And it's going to be free. I think what we might do is make the series um available as an on demand thing after,
158. 28:12but if you're here live, you get to see it for free. So, it's a whole week of back to basics. So, I would argue that
159. 28:20even if you have your [ __ ] together with AI, probably not a bad thing to go back through, you know, just the beginnings
160. 28:29because we haven't done that in a good long time. So, this was producer Brandon's idea. I think it was a really
161. 28:33good idea. And so, so here's my request. My request is let people know about this. I would I would really like to expand
162. 28:46the circle of who comes to these things. Okay? So, let people know. And it's going to be 8 o'clock on Tik Tok AI
163. 28:55learning lab and 8:30 to 10 Mountain time on YouTube. Monday, Tuesday, Wednesday, Thursday, Friday, next week. All right. The back
164. 29:05to basics crash course. By the end of that week, if you come and you hang out and you play while I'm playing,
165. 29:13should be good. You should be off to the races with this AI [ __ ] You know what I'm saying? All right. I know most of the people
166. 29:24here are already up to speed on this, but if you're not, it's a good thing. All right. Good. Makes sense.
167. 29:36Will there be boxes? Yes. Where's my box? There will be boxes. Here it is. It's buried. There will be box. Uh oh. What did I do?
168. 29:52What did I do? I did something bad. There will be boxes. Here's a box.
169. 30:06If you don't know about the box metaphor, I am going to explain large language models to you using boxes and ink
170. 30:14cartridges, ink bottles, and it's going to blow your mind. You're going to be like, "It's not possible. How do they do
171. 30:21that? They just stick it all in the box. All right. So, think about who you know, whether
172. 30:37it's a group of people at work or your CEO or or grandma or grandpa or your kids who are like, "Yeah, I know I should
173. 30:48learn AI, but I don't want to. Mom, stop being so lame. Mom, stop telling me to get into AI. You're like a super nerd
174. 30:57now, Mom. Stop it, Grandma. Bring them. All right. Bring your friends from the bridge club. Listen, I know that nobody plays bridge
175. 31:12anymore. That's not the point. Bring your friends from your Call of Duty circle. All right. And we'll get back to basics. Feel
176. 31:25free to send him your orange Hermes boxes. What's the plan? So, the plan next week is five nights of uh sort of back to
177. 31:36basics educ education. First night, what is generative AI? Why is it a big deal? Second night, chat GPT 101. Third night, creativity
178. 31:47with AI, pictures and music and video and [ __ ] like that. Thursday's bringing it all together, figuring out how to figure out which
179. 31:57tools to use when doing a project. And then Friday's like an ask me anything, a Q&A. Hopefully we've got some continuity
180. 32:05across those four days and people will have questions. And then every night is we start on Tik Tok at 8 o'clock. We do a little
181. 32:13pre-show hangout. Then we turn off Tik Tok and flip over to YouTube. We're just going to be purely on YouTube from 8:30
182. 32:23to 10. I'll be demoing [ __ ] and then the last half hour will be Q&A. And you know how this goes. There will be questions
183. 32:30along the way that I'll probably answer and get off track and we won't get anything done and everyone will be
184. 32:37cranky and they'll be like, "I thought he was going to teach us something. Why is he talking about playing cards?"
185. 32:45Because the Welcome to Chat add. Okay. You can't take the stripes off this zebra. Sure, I'm a horse, but I'm kind of fancy. [Laughter]
186. 33:06You'll get off track, but it can still be fun. I love this. You're up here. Oh, welcome, Angelique. Yeah, that's great.
187. 33:12So, so come next week and listen, this week we'll be I'll be prepare preparing for it. If you have ideas about what you
188. 33:19think should be included next week, feel free to let me know this week. If you're thinking that I have my [ __ ] together,
189. 33:32that would be on you. That would be on you. Okay. I know I seem like a professional. [Music]
190. 34:02You ready to sing, Champ? Well, I heard there was a secret chord David played and it pleased the Lord.
191. 34:12You don't really care for music, do you? It goes like this. The fourth, the fifth minor fall and a major lift. Baff King
192. 34:25composing. Hallelujah. Hallelujah. Hallelujah. Hallelu. Hallelujah. [Music]
193. 34:43Um, all right. So, that's next week. No, Washington DC next week. [Music] What's Washington DC? [Music] Genie in a bottle.
194. 35:00If I'm expected to be in Washington DC, nobody told me. Uh, Ann Murphy's in the house. Is that true? Ann Murphy, what's happening? I got
195. 35:15Wednesday figured out. Ann, we're gonna have Mr. K. Uh, and Gwyn is is co-hosting with me. What's happening? What's happening?
196. 35:29What's happening? What's shaking? What's going down, everybody? All right. So, next week, back to basics. Okay. [Music]
197. 35:40Yeah, let's break the internet. Okay, so for next week, think of people to invite. Give them access to my Twitter address,
198. 35:52AI Learning Lab, and my YouTube address, learning lab-ai. All right, I think tomorrow night I'll have QR codes for that [ __ ] I'm getting
199. 36:03my [ __ ] together slowly. It's coming together. Hey Ann, I started a new thing. There's a surprise, but it's only a week long.
200. 36:16Oh my god. Important to emphasize these are not ghost riders. They're thought partners. Oh, the the uh AI. Yeah, absolutely.
201. 36:27the a AI when you when you feed the prompt when you feed your prompt correctly, when you feed it you your
202. 36:35ideas, your questions, your thoughts, your opinions, your goals, AI transforms from a tool that you use off to the side thinking that it's
203. 36:48supposed to do all the smartypants work. No, it it'll it'll fail you every time if you treat it like that. Oh, give me a
204. 36:55LinkedIn post on political upheaval. It'll write it and it'll suck. But when you say, "Hey, I'm really passionate about this in politics or
205. 37:05that in politics and I want to understand the political context of this versus that." And then you look at what
206. 37:11it gives you and you're like, "No, that's not quite it. I feel more like this." Then it's like, "Oh, here. How
207. 37:16about this?" You're like, "Oh, that's pretty good." That's when it comes to life. when you're the one in the driver's seat and
208. 37:23it's basically just your amplifier. That's how to use this [ __ ] Okay. Um, and we're going to talk about JSON tonight.
209. 37:34He's been talking about talking about JSON for minutes now. Does he ever get to the point? Has he ever finished a
210. 37:41sentence? just just asking because I've I've never been here and it's it says it's a learning lab and I thought there
211. 37:47would be learning and it's kind of more of a professorial kind of situation and he doesn't seem to have finished the
212. 37:54sentence yet. I just was curious if there's is there some sort of curriculum. Oh, next week there might be but probably not. Okay.
213. 38:05And he's the professor. No, no, he he's what? He he's an out of work actor. Oh. Oh. Okay. And so he's he's acting like a
214. 38:19professor in in a sense, but but not a very good one. And and so why do people come here? You You don't know. But you what? You
215. 38:34come every night. That's That's irregular. Oh. Oh, you're called the irregulars. Oh, that's funny. It's It makes sense.
216. 38:45Okay, I'll Yeah, I'll just Maybe at some point I'll learn something, right? Does Is that That's Is it possible I learned
217. 38:51something here? It It's happened. Okay. Okay. May Maybe I'll be one of the lucky ones. I I mean, I play the lotto. I play
218. 38:59the lot. I play I play Megaillions and I play the Powerball. I I don't play the extra number. I think it's a ripoff. Crazy,
219. 39:09crazy. I know. But it's Yeah, I know. I I I would I I would be struck by lightning four times before I won a dollar on on mega millions. I I
220. 39:18understand that. I still I d I have a dream. I have dreams. I have dreams. So So I'll take those dreams and I'll apply
221. 39:24them here. Maybe I'll learn something. That's pretty much how it goes, right people? Isn't that it? Sometimes the hardest is how to ask for
222. 39:34what you want. Yes. And what I have learned from you is to utilize chat GPT to help you create those detailed
223. 39:42prompts. Yeah, Valerie, it's a it's a really good point. A lot of people a lot of people in their life never really
224. 39:50have to take an a position on something, right? They go to work, they're told what to do, they do the thing. They
225. 39:57don't necessarily have to have opinions. A lot of jobs, having opinions is a bad thing, right? Especially in big corporations,
226. 40:07if you're if you work in a political kind of organization or if you work in a situation kind of like the movie Office
227. 40:13Space, your opinions are irrelevant, right? And so a lot of people are not in the habit of having to have a creative
228. 40:23point of view, having to have a point of view. Chat GPT really comes to life when you're clear about what you want and
229. 40:32what success looks like. And it's harder if you're just like, "Go make me something brilliant, chat, GPT." It won't. It'll be It'll be
230. 40:41disappointing every time. Okay. So, here's what we're going to do. We are going to go. [Music]
231. 40:56Do we have a Brandon? We have a No. Brandon. Brandon is right out. He must have fallen right out. We are going to
232. 41:04go. Oh, it stopped sharing again. I have weirdness in my screen sharing lately, people. I don't know what it is. Okay.
233. 41:17So, here we are at chat GPT and then I'm going to go over to Grock. Yeah. Okay. So, I don't know if you all have experienced this, but
234. 41:30in the past month or so, no, not not even in the past two weeks or so, a lot a lot a lot of my the people that I follow on Twitter on X
235. 41:43are talking about JSON prompting, JSo N. They're like, "Oh, the key to prompting is JSON." And it's getting annoying.
236. 41:58And why it's annoying to me is is because the way a lot of influencer influencers are positioning it is there's this
237. 42:13magical thing called JSON and if you don't know it, you're a [ __ ] idiot. And I don't like that. I don't like
238. 42:23[ __ ] intellectual elitism. And I think it's really easy with technical stuff to slip into that. And listen, I
239. 42:31don't think most people are doing it on purpose, but I want to talk about what it is because what it is is actually kind of useful. And
240. 42:42you don't need to even know what JSON is to use it. and not just to use it to like create your own JSON frameworks.
241. 42:53You're like, JSON framework? What the hell are you talking about? Okay, so Ann Murphy, I don't really know what
242. 43:03JSON is, and I'm okay to say it out loud. Bite me, tech bros. Exactly. Exactly. Sorry, too much ADHD. What? No, Brandon. I I think Brandon
243. 43:19Sometimes Brandon goes and takes a nap with his kids and then he sleeps through the start. So, I think that's probably
244. 43:24what happened here. Um, what's the plan, Eric, Kyle? Okay, so here's the plan. So, here's a here's a goodie. Here's a
245. 43:33goodie that I'm going to give you that I think you're going to like. Um, if you've not played with Grock on on X,
246. 43:41one of the things that Grock has access to that no one else can have access to, because you remember when when Elon
247. 43:47bought Twitter, he cut off API access to everyone, right? You used to be able to tap into the data of Twitter. You can't
248. 43:56do that anymore. You cut off all that access. They've got access to their own data. So within Grock, you can do things like ask
249. 44:05about what's going on on on X. Um, and so I asked I said there seems to be a rash of JSON posts about AI prompting.
250. 44:14Characterize the rise in the type of post you're seeing. So it says the trend of discussing JSON prompts for AI,
251. 44:21particularly in image and video generation, has surged dramatically in July 2025 based on an analysis of recent
252. 44:29X activity. There appears to be a very recent phenomenon with minimal minimal discussion in prior prior months. So
253. 44:36there were 10 posts in April, 10 posts in May, 20 to 25 in June and now kind of explosive growth recently. Um
254. 44:49a 3 to 500% increase in volume from June to July. Okay. uh types of types of of prompts that we're seeing de demonstrative showcase
255. 45:03posts. So, look at this cool look at this cool video I made in VO3. JSON was the key to it not sucking, right? See a
256. 45:11lot of those. There's educational ones. Some people are explaining how to use JSON posts, but they're just like,
257. 45:16"Here's my JSON prompt. You go make your own." Right? And then there's some that are sort of debate like is using JSON
258. 45:24better than just using structured prompting just better than using your [ __ ] words, right? That debate's out
259. 45:31there. All right, so that's that. So let's help Anne Murphy. Anne Murphy was the one person in this entire audience. There's one of us
260. 45:41who who was willing to admit, uh, hey, professor, I don't really know what the [ __ ] you're talking about.
261. 45:50Right. So, let's go. Let's go. Let's go to the chat hole. Let's go to the chat hole. Hey, chat Jai. Hi, chat. Jai, explain
262. 46:06what Jso N is to me like I'm a dum dum. I love I love telling Chad GBT I'm a dumb dumb.
263. 46:22All right, imagine you have a big box of toys and each box is each toy is labeled and inside a box toys are sorted into
264. 46:29different sections like action figures, cars and dolls. Now JSON so JSON stands for JavaScriptJS object not notation JS N JavaScript
265. 46:42object notation. You will not be quizzed on this. Nobody gives a [ __ ] if you know those words. That's the history of it.
266. 46:51It's basically a labeling system. It's a labeling system that is understood across computer systems. So it's like
267. 46:59you can think of it like a quasi sort of database or just just a labeling system, right? And so the way it works is you have sort
268. 47:11of open squiggly brackets and closed squiggly brackets. Um you have labels and then you have data. So the label is
269. 47:22something like you know what's your name, what's your age, what kind of pets do you have and where do you live, right? So my name is Kyle.
270. 47:32I'm 30. I have a dog and a cat inside brackets. I live in Denver. Now, why is this valuable? Let's see. Here's the breakdown. Name is a key.
271. 47:45Kyle is the value. Pets is a key. And the value is an array or a list of dog and cat. And location is a key. And
272. 47:54Denver is the value. So, so basically it's just a labeling system. Now, this thing that we have right here is a JSON
273. 48:03framework. The framework is we're going to ask for a name, age, uh pets, and location. Now, this is very similar.
274. 48:15So, I'm going to say, okay, ignore JS right now and explain structured AI prompting for me.
275. 48:33Structured AI prompting is like giving clear instructions to a very smart but a little literal robot. Instead of asking
276. 48:41vague questions like tell me about cats, you're giving it a more specific set of instructions. Right? So for example,
277. 48:48instead of saying explain a cat, you might say give me a list of five unique cat breeds including their physical
278. 48:56characteristics and personality traits. This structured prompt is specific. You're telling the prompt what to focus
279. 49:04on, cat breeds, how many items you want, five, and what details you need, physical characteristics. Okay. So,
280. 49:14using JSON is just a fancier way of using a structured prompt. Okay. Um, so there's all these different prompt
281. 49:25prompt engineering methodologies out there, the ones they sell you for 149 bucks. Buy my prompting methodology and
282. 49:32I'll give you my library of 5,000 prompts and it will change your life for the better. Those guys and gals, mostly guys, douchebags. Um
283. 49:45they're they're basically just selling you they they've come up with a framework that we're going to talk about
284. 49:51a topic, then we're going to talk about colors, then we're going to talk about what other attributes and then they basically fill in that
285. 49:59framework, write write a bunch of prompts, put them into a PDF and sell it to you. Okay, that's the way it works. So
286. 50:07what happens with JSON is because it's highly structured and always consistent, the computer knows exactly what to do with it, right?
287. 50:18Computers are really good at reading code because it's highly structured and they're highly structured. So what
288. 50:24you're doing when you're doing JSON is instead of having to reverse engineer this sentence, give me a list of five
289. 50:31unique cat breeds including their physical characteristics. So now I could just say now turn that structured prompt into JSON.
290. 50:48All right. Task. List five unique cat breeds. Oh, wait. No. Sorry. This gave me the answer. Um,
291. 51:02no. Give me a JSo N prompt because it gave me the answer. Ah, I see what you want now. Okay, so the prompt
292. 51:15task give me five cats format list count five detailed requirements physical characteristics personality traits and
293. 51:25then there's an object cat breeds. Okay. So, JSON lets you look at these labels and it tells you what to fill out.
294. 51:35Okay. And then you can take this prompt and you can always use it. Um you you can fill it in with anything you
295. 51:44want or even better you can have chat GPT once you have the framework established. Have chat GPT gives you
296. 51:50lots and lots of things. Okay. So, here's what I want to do. I'm gonna start a new chat. I'm going to say, um, give me five, um,
297. 52:06attributes of a good midjourney image prompt. See what it comes up with here.
298. 52:21Clarity and specificity, action and mood. visual elements, composition and perspective in incorporate styles. Okay.
299. 52:33Can you turn this into a simple JS framework? Okay. Subject. So what's the subject of?
300. 52:48So there's there's too many things here. So, um, this is too complicated. Give me five, uh, let's see, four keys.
301. 53:00Okay. Subject, style, mood, details. All right. So, the subject, describe the main subject or focus of the image.
302. 53:09Style, artistic style, mood, emotional tone, or action, details, key visual elements, lighting, color, palette,
303. 53:16texture. I'm going to add one more. Let's also add something about location or setting
304. 53:33subject style mood setting details. Okay. So, so this is a JSON framework. Now, you don't have to write this in
305. 53:44JSON, right? You could just say, "Whenever I write a prompt in midjourney, I'm going to make sure it includes a subject, a style, a mood, a
306. 53:52setting, and some details." Right? That's what what all of the prompt engineering methodology meth methodologies teach you. The ones that
307. 54:00have acronyms, the ones that don't have acronyms, the ones that are just like, "I'm a genius and I'll tell you how to
308. 54:07in think of your prompts in a structured way. Follow me." They're they're all doing this. It's all the same [ __ ] This is just a way to do
309. 54:17it that makes it super efficient for the computer to understand and it makes it really consistent across prompts. Now,
310. 54:25I'm going to copy this prompt framework. We're going to do a new chat because again I I want to have a a fresh slate.
311. 54:34So, what I'm going to say, Kyle ASMR, you make me want to be a better person.
312. 54:49Okay. Um, I'm gonna say I have a prompt frame framework for mid journey prompts.
313. 55:07And I'm going to put colon shift return paste. And so I just pasted that all in there. Okay. And I'm going to say now I want you to create
314. 55:2520 JS prompts based on this framework.
315. 55:41And I want the subject to always be
316. 56:04a I'm going to put this in quote. Paisley oyster. I want you to be creative with the other
317. 56:20attributes so that we get lots of variety across the images. Okay. So basically what I'm
318. 56:36doing now is I'm saying I want you to lock down the subject. Right? So you could say to you could say I want you to
319. 56:42use this framework and I want you to just randomly put something in everything or for the style I want I
320. 56:49want everything to be the same style but I want it to be lots of different subjects. So this is just a way to very
321. 56:54efficiently you know create a little a little framework. Um, you can learn how to write the JSON by just taking a look
322. 57:02at what chat GBT chat GBT did, but you don't have to. And so I'm just going to have it go do this. So it's going to
323. 57:10give us 20. Here's 20 creative midjourney prompts. Okay. Paisley Oyster, whimsical watercolor, photorealistic, steampunk,
324. 57:21cyberpunk, vintage oil painting, abstract expressionism. So, let's go grab Paisley Oyster Abstract Expressionism. We're
325. 57:32going to head on over to what's this called? Midjourney. I'm going to change my my stylization. I'll put it 100. Weirdness, I'm going to
326. 57:42put it zero. We'll do this as a 16 by9. My my favorite aspect ratio. Um, I'm going to turn on personalization because
327. 57:51this is kind of an arty kind of project. And I'm going to paste in that JSON. I didn't copy it because I'm a loser,
328. 57:57baby. Copy. Go over here. Paste. Wait, why is it not pasting? What's going on? Why is this not working, people? I'm confused.
329. 58:13Copy. Paste. Paste. Why am I not copy pasting? Um,
330. 58:28it might be because this is in a copy block. So, let me just copy this all out of here. Go over to notes and go. Oops.
331. 58:38Go. Oops. No. Go new. Paste. All right. Which one did we want? We wanted abstract expressionism. Right.
332. 58:48There's Gothic minimalist. Oh, that's cool. These are cool. Surrealist, abstract expressionist. Okay. Copy. By the way, the whatever I just
333. 59:00experienced. If this doesn't paste, then something's wrong with my Chrome browser. No, there it pasted. Okay. So,
334. 59:08now you see the prompt is this, you know, highly structured JSON prompt. And I'm going to do that. And then I'm going
335. 59:15to go back over here to notes. And I'm going to grab here's a surrealist one. So, let's grab a surrealist one. Go back over here to
336. 59:31Why? Wait, why did it do so many? Oh, wait a minute. This This [ __ ] up.
337. 59:48blues and yellows. So, wait, it split it split my JSON up into many individual
338. 1:00:03components instead of putting them all together. I've never experienced this before. This is weird. Um,
339. 1:00:15let me go back to chat GPT and say, um, give me the same prompts outside of a code block.
340. 1:00:29Has anyone else ever experienced that with JSON in Midjourney where it split everything out? Brandon's here. Producer Brandon's here.
341. 1:00:41You can't just put JSON in a midjourney prompt. Oh, well that sucks. Oi, color me wrong. It doesn't parse it as JSON because it's
342. 1:00:54neuro. It's natural language processing. There's no sleep in an AI. Kyle probably I predicted that you had fallen asleep
343. 1:01:01with your kids, Brandon. But people did miss you. You predicted correctly. Yes. That's awesome. Okay. Well, so let's
344. 1:01:15let's just now that we've got this structured JSON. So it looks like with with midjourney. So here's abstract expressionism. Let me
345. 1:01:26grab it as nonjson. I'm just going to grab it as as text and pop that in there. All right. Now it did it correctly. All right.
346. 1:01:39So there you go. See, you don't have to use JSON. In fact, in MidJourney, you can't. Which is I, you know, Vicki, when
347. 1:01:46when you first told me just stick raw JSON in there. I'm like, I don't think that's going to work. And you were like,
348. 1:01:52no, no, just try it. And then I tried it and it broke. So, you know, I I don't want to say I told you so, but you know.
349. 1:02:01So, apparently there's places where JSON prompting isn't good. which we which we learned in real time. But let's go let's actually go look
350. 1:02:09let's have a chat add moment, shall we? Let's go look at Oh, this is our paisley oyster, but it's abstract expressionism. So, of course,
351. 1:02:19it doesn't look like an oyster. Let's go grab a surrealist one. That's pretty funny. Copy. Go back over
352. 1:02:29here. Oops. Oh, no. I just got rid of my chat GPT. I'm such a [ __ ] Tab management. All right. There's that.
353. 1:02:36Hang on a sec. Physical dexterity challenge. I failed. Damn it. Oh, I failed again. Uh oh. Hang on. I'm about
354. 1:02:44to get permanently banned from TikTok. Did I do it? Yay. I did it. Okay. I leave you unsupervised for one hour and
355. 1:02:51you break midjourney. I totally broke midjourney. All right. Look, there's our paisley. There's our paisley oyster
356. 1:03:00um in Surrealist. Let's go grab another. Let's turn this back into chat. GPT.
357. 1:03:17Okay, let's do noir paisley oyster. Noir copy. So, so does this make sense that JSON is just a way to structure prompts.
358. 1:03:31So, so you can come up with any structured prompt. My favorite thing to do, like I like structured prompting in
359. 1:03:38certain cases. And I'll tell you I'll tell you where it makes most sense. When you're doing a single image, it doesn't
360. 1:03:44necessarily make sense. If you're doing something like a VO3, a video prompt, JSON kind of makes sense because you could have attributes for
361. 1:03:56cinematography, for movement, for lighting, for action. You could break the thing into different scenes and
362. 1:04:04those scenes could have start and end times. So you can actually have very very complicated JSON that describes
363. 1:04:11lots and lots of elements of a moving story, right, of a video. Um, subject paisley oyster. Why is there no oyster here?
364. 1:04:25H I think because these prompts suck. All right, let me go rewrite these prompts. as NLP
365. 1:04:41prompts. Just play, Kyle. Just play. Exactly. Okay. Create an image of a paisley oyster. So, let's go get our
366. 1:05:04So, here's abstract expressionism Paisley Oyster with natural language prompting. So, these should give us better results, I'm thinking.
367. 1:05:15Where's the new R1? Here's the surrealist one. Let's grab this.
368. 1:05:31Oh, I've also got my personalization turned on. That could [ __ ] it up, too. Let me turn that off. And let's go get this. And let's go get
369. 1:05:46this. And then let's go find the new R1. That's Gothic,
370. 1:06:04nostalgic. Actually, you know what I'm finding really interesting right now? because it wrote these as natural language. I'm
371. 1:06:17having a really hard time finding the different styles because it it sort of put them in different places.
372. 1:06:26That's an impress impressionist one. Bold and quirky pop art style. Use black and white tones, heavy shadows. Okay, here's the new one. So,
373. 1:06:37there's new. So, we'll pop that in there. So the nice thing about JSON is they're very easy to read the different elements.
374. 1:06:57This is kind of cool. Paisley oysters. Does JSON dejanky somewhat? Well, in the
375. 1:07:12case of Midjourney, it it rejankified. It it increased the jankification because midjourney is not reading JSON.
376. 1:07:19But let's take one of these JSON um one of these JSON prompts. Um let's see. I'll go grab one from here. Let me grab the surreal surrealist
377. 1:07:34one. We'll grab the Let's grab Gothic. So, here's a Gothic prompt. And then let's go to we'll go to some other image tool like uh
378. 1:07:51let's go to um ideoggram. We'll see if Id ideoggram can eat eat JSON. So in theory, what it does, the the only thing JSON does is it uses a
379. 1:08:09consistent um structured language that most computers understand rather than every different prompting framework being some different
380. 1:08:22combination of things. It just makes it consistent and easy to use. And chat GPT is really good at producing it. So you
381. 1:08:28can say I I'll show you a complicated one here in a second. So there's there's our JSON in ideoggram and let's generate
382. 1:08:37that and see how it does. So it says generating magic prompt right now. Um so what it's doing is it's
383. 1:08:46taking it ingested that JSON and it's um oh that's beautiful actually. Um it interpreted the JSON and rewrote the
384. 1:08:57prompt. So, I put in JSON and then it said, "A photograph depicting a paisley oyster resting on a weathered stone
385. 1:09:06pedestal with a decaying within a decaying castle." Those are actually beautiful. Look at that one. Beautiful. I love the smoke on
386. 1:09:16that one. The little dying rose. And and what's nice is you know that all of the prompts have the same elements in them, right? They
387. 1:09:29all have the subject, they all have a style, they all have a mood, they all have a setting, and they've all got some
388. 1:09:34amount of details. So, let's do the impressionist one now. Go back over here. Paste in the new thing. See, generate.
389. 1:09:56It's writing its own magic prompt. Yeah. And I'll I'll do I'll do them in chat GPT as well. That's a good idea,
390. 1:10:06Vicki. Right. And so they all have this consistency about them now. And what's nice is you can go in and then you can
391. 1:10:13now change like if I say that that doesn't really look impressionist. So, we'll say style impressionist. Um,
392. 1:10:22I'm going to say impressionist painting with visible strokes. So, I can go in and just change one of the elements
393. 1:10:35and we'll see how it affects it.
394. 1:10:52It didn't. Let me turn off magic prompting. Public default magic prompting off. Let's just let it be the JSON. See how it does.
395. 1:11:13And this is the kind of thing to experiment with. Yeah. So this this is a bit more impressionistic, right?
396. 1:11:20It looks like it looks like a cheesy postcard, but if you want cheesy postcard.
397. 1:11:34All right. Okay. So, let's go try this prompt in chat GPT. So, I'm going to do new chat. I'm just going to put in let's say um
398. 1:11:49make a wide image colon and then the JSON. And so, it's now going to read that and make us an image. Chat GPT was slow
399. 1:12:05today. Boy, was it slow.
400. 1:12:23Oh, you don't have to put anything in other than JSON. Cool. Okay, great. Titan Todd experience the slowness of Chad GPT today.
401. 1:12:46Paisley oyster. Yes, Paisley oyster. Why did I come up with that? I don't know. Welcome to the way my brain works.
402. 1:12:56Have I ever put the words paisley and oyster together in my life before? No. I I've got a large language model up here.
403. 1:13:03It's just more random than most. Oh, the size or orientation should be in the JSON. Yeah, that's that's a good idea.
404. 1:13:32Mr. It Hello, irregulars. Come on, man.
405. 1:13:50Wait. the first ever AI only social video app. Is it the Pika thing? That's Pika. Yeah, not only did I see that, I
406. 1:13:58went and looked at it. It's really interesting. Yeah, I'll talk about that. I didn't see that from Danielle, but
407. 1:14:02thank you, Brandon. Um, so while this thing's rendering, uh, taking forever. Um, it chat GPT, they got to get their
408. 1:14:13[ __ ] servers up and running. They're up to a million GPUs, by the way, and they need like 10x or 100x more than
409. 1:14:19that. It's it's crazy the amount of GPUs they need. Um, so PA, if you don't know PA, they're using Google now. Oh, chat GBT. Yeah, they're
410. 1:14:35using Google servers. I wonder if they're using TPUs. That That's actually fascinating. They might be. Um,
411. 1:14:46we can't make fun of Google anymore, can we? Pete, Pate, what's going on? Um, so Pika, if you don't know what they
412. 1:14:57are, it's a video generation platform, PA Labs. And they've got a bunch of presets that make it really easy to like, you know, take
413. 1:15:10an object and peel it off like a sticker or crumple it up or cut it with a knife and it looks like cake. Um, and they
414. 1:15:18have sound effects in them and things like that. And and they keep adding these things. They're they're kind of
415. 1:15:24like Tik Tok filters in in a certain way. Well, they just launched in their iOS app. They've they're they're coming out of
416. 1:15:33private beta right now and they're they're basically doing an all AI video social channel, social network.
417. 1:15:48So I can go in there and I can make all sorts of cool videos with me with a monster behind me or me me crumpling up
418. 1:15:54objects or whatever it might be. And I think it's actually fascinating because when you like I think I have
419. 1:16:00access to it. I was playing with it a little bit before I came on and when you're going through it, it's just like
420. 1:16:07these fun videos that have these crazy special effects in them and it's like why not? Like why not have an all AI
421. 1:16:16like this whole sort of [ __ ] thing of well we've got to label it if it's AI generated. Well no you don't.
422. 1:16:24You have to label it if someone is trying to deceive you, right? Like like there it makes sense. But the fact that
423. 1:16:31it was made from AI from an AI tool is irrelevant. Like if it's just if I used an AI tool and I'm making like a little
424. 1:16:39cartoon character walking down a street, does it matter to you if I generated that in AI or I did it with hand
425. 1:16:47photographed claimation frames? No. Who gives a [ __ ] So, they're leaning into it and just going, "This is an all AI channel. It's
426. 1:16:58all Pika videos." I think it's pretty smart, actually. Meta is trying to do the same thing with photos in the in the Meta AI app. I
427. 1:17:07actually got a call from them teaching me how to post on the app. That's fascinating. All right. Well, here's our
428. 1:17:12here's our chat GPT app uh or our chat GPT image. Uh, as as typical with chat GPT images, it's way too yellow. It's
429. 1:17:23everything is overly yellow. Uh, I hope they fix that in the next update because it it really sucks. You can you can you
430. 1:17:30can now spot chat GPT generated images a mile away because they all look like some sort of aged 1940s. They were printed on 1940s paper and
431. 1:17:41have aged and yellowed. Okay, I want to do one final thing now with JSON. So, if you were here earlier, you get that JSON is just a is a
432. 1:17:53is is a fancier way of doing a structured prompt, right? where you're doing these labels that use the same
433. 1:18:00sort of computer uh format, this thing called JSON, JavaScript object notation that every tool apparently except MidJourney can take. Um
434. 1:18:14there's places where having a JSON framework, right, a preset framework really does start to make sense. And if I think of something like
435. 1:18:25V3, that's one of them. Um, so, so what I want to do here is I want to, um, I want you to research
436. 1:18:40the best V3 prompting, best practices.
437. 1:18:59and include um multiple scenes in the same generation.
438. 1:19:14Okay, so it's going to now go off and search the web. Totally agree. I hate the yellows. I know. It's just everything.
439. 1:19:23What what what ends up happening if you just have engineers in a company is that engineers are not going to look at the outputs of
440. 1:19:37chat GPT and say, "Hey, all this [ __ ] looks too yellow." What they're going to say is, "Did it include the correct objects that
441. 1:19:47were specified in the prompt?" Yes. Did it Did it make the letters Did it spell the letters correctly? Yes. Okay. Check
442. 1:19:55check check check. They're going to check all of the requirements. Clearly, one of the requirements was not color balance.
443. 1:20:04If they had a a liberal arts major or a creative person on staff, they might have said, "Hey, we've got to
444. 1:20:13we got to we got to look at the overall aesthetic here. Oh, and by the way, why do you not have 16 by9 as an output
445. 1:20:22format since every video format out there is either 9 by6 or 16 by9? Why is that not an option? Because engineers built it.
446. 1:20:34An engineer said, well, we have a wide option and we have a tall option and we have a square option. The requirements
447. 1:20:40have been fulfilled. Why would you need more than that? Oh, I don't know. Because every time I generate an image
448. 1:20:48in chat GPT and I want to put it into a video, I have to take it into Photoshop and extend it. That's a pain in the balls.
449. 1:20:58Tik Tok pin. I asked Chat GPT to create my images without the antique finish. Did it work? I'll try that. Okay. Divide it into narrative and
450. 1:21:10sequential sequences. Okay. So I just asked chat GPT to go figure out tips for V3, right? So now I'm going to say now turn that into a
451. 1:21:29JS framework where each video has three scenes.
452. 1:21:44a beginning, middle, and end. Um, as well as overall controls for look and feel,
453. 1:22:00music, etc. And then scene by scene details. All right. So, so it's going to come up with a a framework for us here.
454. 1:22:15She says it works. Details in Tik Tok pin. You just have to ask for it not to put tone on the image. Okay,
455. 1:22:24let's definitely go try that because that's annoying as [ __ ] that tone, the antique finish.
456. 1:22:38Okay, so we've got video, overall controls, look and feel, style, cinematic, lighting, soft and ambient, mood,
457. 1:22:48dramatic and intense, color palette, camera movement, music, scenes, scene one, title, beginning. We don't need titles.
458. 1:23:06Okay, let's see. Um, scenes don't need titles and should blend smoothly
459. 1:23:20from one to the next. Black bar. Yo, black bama jam. Oh, black
460. 1:23:36bama jam. All right, fantastic. Fantastic. All right. All right. You ain't very good. Okay.
461. 1:23:51Great. Now give me a sampler of three fleshed out JS prompts
462. 1:24:09based on this framework the visual wait let's see all elements and the story each one tells
463. 1:24:25should be world class and memorable. Find your inner uh scorses.
464. 1:24:39[Laughter] Find your inner scorses. All right. Boom. Off we go. Here are three J fleshed out JSON prompts.
465. 1:24:51Okay. So, the first one, let's see what it says. Cinematic and dark, epic, high contrast, mood, desperation, and sacrifice. Color
466. 1:25:02palette, muted blues and reds, flashes of gold, music, tense, orchestral scenes. A lone night, battered and bloodied.
467. 1:25:13Wide shot captures the night. Harsh shadows. Wind howls. The knight whispers to himself, "This is the end, but not
468. 1:25:21for me." All right, this is good. So, we're going to copy this. Then, we're going to go over to Flow. You're like, Flow? Oh, yeah.
469. 1:25:29Flow.google. Flow. Google is where you make V3 movies. V3 is their their little action, right? Their video tool.
470. 1:25:43Make scenes of up to eight seconds. We're go new project. And we're just going to paste in the JSON. There you go. There's your JSON.
471. 1:25:56Fancy fancy. Look at all that code. Did we have to write the code? No. Do we have to know what JSON is? No. We can.
472. 1:26:04We can know that it exists. We can know that if we have a JSON framework, we're going to have scenes and shots that are consistent
473. 1:26:14across time because they're all going to have the same elements that we can control. So, if you're a control freak,
474. 1:26:21JSON prompts are your buddy. If you're an add freak, you might not give two shits. Then just have Chat GBT write it
475. 1:26:28for you like I just did. All right. Boom. Off to the races. Let's go grab another one. The forbidden encounter. Gritty noirque.
476. 1:26:39All right. Let's copy that. Pop that in there. Oh, I'm in the wrong flow. Oh well,
477. 1:26:56we've got two different flows. We'll go get one more. The awakening of the fallen hero. Boy, it's getting mighty dramatic in here,
478. 1:27:05huh? Right. Bang. Bang. Makes sense. Now once ChatGpt understands that structure, it can just go write a bunch of creative variations
479. 1:27:17of it. Or you can go do that. If you want to use your fat human brain with all your cholesterol getting eaten up in there, fine, do that.
480. 1:27:28Or if you want to be lazy like me, have Chat GPT write that [ __ ] and then go make a video of it and you're like, "Oh,
481. 1:27:35that sucks." And you go look at your JSON and you're like, "Oh, that was crappy prompt. I could have looked at
482. 1:27:40that. That would have saved me a nickel or 75 cents or whatever the hell these things cost these days. Too much. Too much, I'll tell you.
483. 1:27:53Oh, look. We have our We have our dude. So, let me change my sharing, shall I? [Music] All right. Fantastic, Bob. Here we go.
484. 1:28:09Let's see how disappointing we are. This is JSON at its finest. >> This is the end, but not for me. >> This is the end, but not for me.
485. 1:28:30This is the end. >> Not bad. Not great. That little transition at the end is weird, but you know, not bad. All right, here's the next one.
486. 1:28:41>> Where am I? >> You have been reborn, but your journey has only just begun. >> Where are we headed? >> To where the light leads.
487. 1:28:49>> Where am I? >> You have been reborn, but your journey has only just begun. >> Listen. >> Where are we headed?
488. 1:28:55>> To where the light leads. >> Listen, this is every cutscene in every video game ever. >> Where am I? You have been reborn, but
489. 1:29:05your journey has only just begun. >> Where are we headed? >> To where the light leads. >> That's [ __ ] hilarious. All right,
490. 1:29:12let's go look at our our last one. I got to change my Wait, did this work? This generation violates policy. Why? Why, you [ __ ]
491. 1:29:33I'm so tired of safety controls. Can we be done with safety controls? Let me just make violent [ __ ]
492. 1:29:48We'll see if it makes it. There's Grock for that. Except Grock doesn't do video yet, although I think they're going to soon.
493. 1:30:04All right. Was he taking a nap? Yeah, I think he was taking a nap. Tom Arnold, that was pretty funny. All right, let's
494. 1:30:13see if this one renders this time. I think it might. We'll see. So, is JSON making sense yet? No, it it just killed it. H, unbelievable.
495. 1:30:28Um, all right. Let's go back to chat GPT. Let me change my sharing again. Let's go see if we can make chat GPT not give us something in
496. 1:30:36Hey Danielle, what do you what do you say to it? Don't give it that antique finish chat. Um, make me a wide image of
497. 1:30:52um, I asked Chat GPT to create my image without without the antique. a wide image of uh a claymation.
498. 1:31:08Claymation. Um, jogger running past a hot dog stand in New York City
499. 1:31:24with a squirrel peering. How do you spell squirrel? Q U I R E L. Peering. Peering. Shut up. Pier shut up.
500. 1:31:41Do you know how hard it is to type in public people down at him? Um, comma without uh let's see uh
501. 1:31:57remove the antique finish. from the image.
502. 1:32:12All right, let's see if it works. Let's see if it works, people. Carrying on. Carrying on. Oh, is Tom Arnold for me? Hey, Sarge. What's
503. 1:32:25happening? I get it. That's a troll. H Tom Arnold, I've heard. Fat Alec Baldwin, I've heard. Let me let me help
504. 1:32:34you out here. Uh you could also call me uh Rosie O'Donnell on a bad day. Uh Philip Seymour Hoffman. I've gotten
505. 1:32:48uh Tom Arnold. Uh who's the other one? There's there's another one that's kind of Tom Tom Arnold like uh
506. 1:32:54was it the the family guy dude? Family something something family modern family. I don't know that guy. Anyway, yeah. So, welcome,
507. 1:33:07welcome, troll away. My request with trolling is the best trolls feel like a compliment on the surface and then you think about
508. 1:33:17them and they they hurt. They they dig a little deeper. So, you know, get creative. All right.
509. 1:33:43Hey, March. [Laughter] Oh, this does look less yellow, doesn't it, Danielle? I think you're right.
510. 1:33:58And you know what we'll do? We'll take this image and we'll go uh we'll go to VO3 and we'll turn it into something.
511. 1:34:11But look, it doesn't have the antique finish on it. But like seriously, like seriously, there there's no one at Open AI. They're
512. 1:34:20they're a 14 trillion dollar company now. They didn't hire one visual arts person. They didn't hire one
513. 1:34:32liberal arts person to come in there and say, "Uh, hey, how come everything looks depressing here? Look at this. This is
514. 1:34:41so much better." They could they could literally put it in the system prompt. In fact, if you want your chat GPT
515. 1:34:49images to not look shitty, go into your system prompt and update this. We can go do that right now. It's on the backlog.
516. 1:34:56Exactly. All right. Uh let me How am I going to do this? I'll download this image and then we're going to go over to VO3.
517. 1:35:06There's our sleeping prince. We're going to go frames to video. We're going to upload our image that we just make, excuse me, just made.
518. 1:35:21Lovely.
519. 1:35:42Okay. Oh, by the way, remember when I was bitching before about Chat GPT not doing things in 16 by9? This is why. See, I
520. 1:35:58have to figure out how how do I cut off his feet or do I cut off the squirrel's tail? Chat GPT should figure this [ __ ] out.
521. 1:36:09You're supposed to be the state of the art model. Start acting like it, [ __ ] All right, sleeping prince. That's pretty funny.
522. 1:36:26Okay. So, we're gonna say uh man jogs by as vendor says
523. 1:36:42hey. Says hey. Hey, you feeling snacky?
524. 1:36:57Question mark. And the squirrel says, "Sure,
525. 1:37:11I'll take some peanuts." They both laugh. [Laughter] Uh oh. It can't handle caps for
526. 1:37:26inflection. I don't think I did caps. All right, let's see how this does. Oh, okay. Demand your $10 back, Kyle. Listen, it's
527. 1:37:42$20. It's $20. my uh my chat GPT subscription. Oh my god, I'm getting sleepy, people. Oh yeah, it's getting to be time. What
528. 1:37:56are we doing? What are you guys still doing up? Okay, while that's rendering, let me go pop this thing up.
529. 1:38:11Next week, next week, back to basics. We're doing a fiveday AI deep AI crash course. Okay. It starts next Monday the 28th. We do
530. 1:38:27half an hour on Tik Tok and then we do an hour and a half on YouTube. All all next week. All right. I want you to tell people
531. 1:38:37about it. Share it with people. Okay. All right. Okay. We'll we'll have QR codes and stuff tomorrow or sometime this week. All right.
532. 1:38:52Okay. Okay. [Music]
533. 1:39:06All right. Here we go. I think we have video. We have jogger. We have jogger. We have jogger and squirrel. Jogger and Squirrel.
534. 1:39:15You ready people? It's going to be so good. It's probably going to be disappointing. Although, I got to say V3 things are
535. 1:39:22less disappointing than I usually expect. Hey, you feeling snacky? Sure, I'll take some peanuts. Hey, you feeling snacky?
536. 1:39:37[Music] Sure, I'll take some peanuts. Hey, you feeling snacky? [Music] >> Sure, I'll take some peanuts. Hey, you feeling snacky?
537. 1:39:54[Music] Sure, I'll take some peanuts. Listen, it's not Nickelodeon quality, but you know, it's close.
538. 1:40:09Oh my god, that's bad. That's That's bad. But you know, you know you know how long that would have taken you to do if you did it as
539. 1:40:19actual claimation. >> Hey, you feeling snacky? Sure, I'll take some peanuts. Hey you.
540. 1:40:35Oh my god, it's giving robot chicken. It's pretty bad. All right, good people. So, um,
541. 1:40:49what was the homework I gave over the weekend? I think the homework I gave over the weekend was play with, uh, play
542. 1:40:54with Manis and Jen Spark. Hopefully, you did that. I think someone said earlier they they played with GenSpark and it was disappointing. Um,
543. 1:41:02all of these agentic tools right now are disappointing. All of them. All of them. They hallucinate. They're too slow.
544. 1:41:10They're all sorts of [ __ ] up. So, okay. Danielle just got access to Pika's No Thing. I think I may have access to
545. 1:41:22it, but I don't know what it looks like. Does it just look like you can scroll like Tik Tok? If that's it, I think I have it.
546. 1:41:36All right, before I take off, there's apparently a few fun things in irregulars. So, let's go take a look at irregulars, people.
547. 1:41:45Let's go do this. If anyone put things in irregulars that has sound, I'm going to have to change my tabs again and that's going to be a
548. 1:41:53pain in the ass. Oh, it's a whole different app per Danielle. I don't really remember what the other app looked like, but I'll I'll
549. 1:42:03go see if I have access to it because I think I have Pika Pro. I think I'm in the Pika um creators program, but I'm not sure.
550. 1:42:15All right, I'll go take a look. You were going to a regulars. Do you see what Brandon does there? He he he talks
551. 1:42:24to me as if I didn't remember. He doesn't realize I've got genius level large language models just sitting right
552. 1:42:30up here in my brain. He's like he like I I don't need to be reminded because I because I have my own large language
553. 1:42:38model. You have too many tabs and not enough context window. That's totally it. I I am My my tabs are a reflection
554. 1:42:46of my mental state. Oh, see, we do we we do Vicki at work today. We do have to go switch tabs because Vicki put up a
555. 1:42:54video. I told you this was going to happen, people. You don't think I know what's coming? I know what's coming. Oh, and then this is
556. 1:43:03going to pop it into a different window, isn't it? Maybe. Maybe not. Vicki at work today. No audio. Okay, good.
557. 1:43:18That's great. I love the I love the style of that one, Vicki. That's very cool. Back to basics. Oh, nice. What's this? Steo. Nice.
558. 1:43:30I like it. Come on, guys. You had to ask. Very cool. Oh, you made it 16 by9. Oh, that's good. Yeah, exactly.
559. 1:43:46That's pretty funny. Uh uh uh uh. Pico is invite only for now. I think I deleted mine. Samma speaks about the future.
560. 1:44:02Just another manic Monday. [Music] I'm working at work until midnight today, tomorrow, and Wednesday. Vicki,
561. 1:44:17oh my god, I'm so sorry. I'm so glad you're here. Screw them. Hang out in the AI learning lab. Tell them it's a
562. 1:44:26professional development. Oh, nice. Oh, all right. I'll take this one. Nice, Steo. That's good. Oh my god, there's Champy. Very nice.
563. 1:44:39All right, cool. Wait, go back up to digital gods post where in irregulars. I don't I do not see that. I scrolled past it.
564. 1:44:57Lord digital gods. He was a god. He was a digital god. Oh, story transformer. What is this? >> Hey everyone, it's Lord Digital Gods,
565. 1:45:11also known as Glenn. Here to uh show you something I've been working on. This is done almost entirely by Vibe Coding.
566. 1:45:19It's been built on Manis and Gro 4. Uh the nice details been very very happy with uh Gro 4 by the way. Unbelievable.
567. 1:45:28Unbelievable model. All right. So the purpose of this uh application is to create kids books from any story, adult stories or
568. 1:45:37>> beautiful. Okay. So So that's 18 minutes long. We're we're not going to watch the whole thing there. Long video. Here's the thing.
569. 1:45:46This is a really good example. Hang on. Let me let me flip my So, so where that is, by the way, if you're wondering where that is, um, if
570. 1:45:55you go to the AI salon, if you're if you're a member of the AI salon, go to the irregulars channel, scroll down a
571. 1:46:02little bit, you'll see this post from Lord Digital Gods. So, not only has he created a really cool application that's
572. 1:46:08going to generate a kids book, he's created this video explaining how to use it. One of the things we talk about in
573. 1:46:15the AI salon all the time is generously lead or learn out loud. Right? So as you're building stuff, share it with
574. 1:46:22other people and share how you built it, how you're learning it. Notice in the beginning of that he said, "Hey, I built
575. 1:46:29this app. Most of this is was done with vibe coding and I've been really happy with Grock 4." So there's like already
576. 1:46:35goodies in there that you can learn from. This is one of the best ways to get good at AI and get known in AI as
577. 1:46:43someone who's up to something is build stuff and share what you're learning. Generously lead, right? Share it in the
578. 1:46:51AI salon, share it on LinkedIn, share it on Twitter, share it, you know, at your dinner party with your friends who hate
579. 1:46:58you talking about AI. All right, beautiful. Love it. Okay. Uh, I'm going to get out of here. Tomorrow night's Tuesday. I think it's a
580. 1:47:08normal night tomorrow night. Yeah, it is. It is. Right. Yeah. Yeah. Yeah. Yeah. Yeah. Yeah. Yeah. Yeah. Yeah. Yeah. All right. Um
581. 1:47:18I know the JSON stuff is kind of geeky, but just know it's just a different way to structure a prompt. And you don't
582. 1:47:24have to learn JSON. Chat GPT already knows it. So does Gemini. So does Grock. Every all of the things know what JSON
583. 1:47:32is. So if you want help building a JSON prompt, just say what are the elements I want in my prompt and tell it to make a
584. 1:47:39JSON framework out of that. And then you have structured prompting things. If you're doing it in midjourney, don't do that. Do other [ __ ]
585. 1:47:50All right. Fantastic everybody. Uh what time central time tomorrow night? Central time tomorrow night it would be
586. 1:47:569:00. So I start 8:00 p.m. Mountain time unless otherwise specified. Next week is the back to basics uh AI
587. 1:48:06crash course. Five nights covering basically all the bases of AI to get you started. All right. Alrighty. Okay. Okay. Is everybody ready?
588. 1:48:22All right. Okay. Hey, I wonder if we should do a vibe coding night. Brandon, right now we've got a Q&A night. My
589. 1:48:34problem with vibe coding is it just might not be interesting to look at. I don't know. We'll talk about it. You
590. 1:48:41answered your own question. Yeah, exactly. All right, fine. Okay, everybody, have a good night and I will see you tomorrow.