8/14/2025 - Exploring Pika Social, GPT-5 Pro, and Vibe Coding

Description

Chapters

Transcript

0:00 Maybe
0:02 sing
0:08 him
0:11 like slow.
0:13 [Music]
0:18 His beard was a looking guy.
0:26 She sat on a stool and he said, "What do
0:29 you want?"
0:33 She said, "Give me a love that don't
0:36 freeze up inside."
0:41 You said, "I have melted some hearts in
0:44 my time, dear."
0:47 [Music]
0:49 But to sit next to you, well, I shiver
0:52 and shake.
0:57 If I knew love, well, I don't think I'd
1:00 be here.
1:04 >> Asking myself if I've got what it takes.
1:08 [Music]
1:12 I do your blue heart.
1:20 has done.
1:22 [Music]
1:23 Turn what's been frozen for years
1:29 into a river of tears.
1:36 [Music]
1:52 Thursday night. Good evening, good
1:53 people. What's going down? What's
1:55 shaking? What's happening? Steo, welcome
1:58 from Down Under. Good to see y'all. Hope
2:00 you're all doing well.
2:04 Oh man, I've been having fun with uh
2:09 with Pa Pab Social. Pika Social.
2:14 That's been a that's been a fun
2:16 [Music]
2:36 I have some thoughts on on Pika Social
2:40 from a development standpoint.
2:43 [Music]
2:47 And from a uh how we're going to adopt
2:50 AI standpoint, I think there's some I
2:53 think there's some really interesting
2:55 things to learn from what they're doing.
2:58 So, we'll play with we'll talk about
2:59 that tonight. Champy's in fine form. He
3:01 is in fine form tonight.
3:05 What a week. Looking to get my dose of
3:07 laughter tonight. Well, hopefully
3:09 hopefully I got some comedy in me. Last
3:11 night I was just a cranky little [ __ ]
3:18 for no reason.
3:20 [Music]
3:25 Oh, I forget. What does Pika cost? I
3:27 think Pika is like one of the normal
3:29 video things. They probably got a $10
3:32 tier that's useless and it's probably a
3:34 $30 tier where you get some decent
3:35 amount. Um,
3:38 the Pika social app. So, there's Pika's
3:41 got two apps. They've got the Pika app
3:44 where you do video generation and then
3:46 they have Pika Social which is in
3:49 limited preview right now. You can only
3:51 get in if you have an invite and if if
3:53 you get invited then you get three
3:54 invites. It's one of those things. It's
3:57 iOS only so it's the iOS app.
4:00 Um and they're kind of
4:06 I don't want to say they're competing
4:07 with Tik Tok. They have created a Tik
4:09 Tok like app.
4:12 Um, and right now it's free. They're I
4:16 think they're just trying to understand
4:17 what they've built and what they've got.
4:19 Um, I think it's a really interesting
4:23 lesson in
4:26 [Music]
4:29 in um resisting the urge to
4:34 add more and more features. It is a It
4:36 is a frustratingly limited feature set
4:41 and I I actually think it's kind of
4:43 smart. And I'll I'll tell you what I
4:45 mean in a bit. I'll show it to you. In
4:47 fact, let me make sure my iPad's charged
4:49 so
4:52 we can play later.
4:54 We will play later. Yes. Yes. Yes. Yes.
4:59 Yes. Yes. Danielle's been having fun
5:01 with it. Yeah. It's really fun. It's
5:02 really they're they're really on to
5:04 something here. Mr. IT, what's happening
5:09 [Music]
5:24 through Blue Telescope.
5:27 Looking at the world tonight through
5:30 blue telescope.
5:33 Wish I may, wish I might not see what I
5:37 see.
5:39 Sheet metal on sheets of ice.
5:46 Look through this blue telescope.
5:49 [Music]
5:51 Down a moon struck a road tonight.
5:55 [Music]
6:11 Yeah.
6:23 Yeah.
6:27 [Music]
6:39 Every time I see you now, get that look
6:42 in mine.
6:43 [Music]
6:44 Every time I see your mouth, I hear that
6:48 smile.
6:51 The early misty morning light I heard
6:54 the engine turning in the old f outside.
7:00 [Music]
7:04 You were leaving me
7:07 again today.
7:10 You will convince me
7:13 again today.
7:16 You're leaving this hard time looking
7:18 for someone else's cold and rain.
7:23 [Music]
7:27 So long
7:30 [Music]
7:33 now. Don't you cry.
7:36 [Music]
7:41 So long, Susanna.
7:44 [Music]
7:47 Don't you cry for me.
7:51 [Music]
7:57 [Music]
8:03 Do you know that Chad GPT5 Pro is now
8:07 smarter than 99.9%
8:10 of the humans on the planet? Did you
8:13 know that
8:20 is true?
8:22 [Laughter]
8:26 Oh man. Welcome everybody. Welcome,
8:29 welcome, welcome.
8:33 Hope you're doing well tonight.
8:36 [Music]
8:59 Wraith
9:01 desperately hating his old place.
9:04 Dreaming to discover a new space.
9:09 [Music]
9:12 desperately hating his old place. Dream
9:15 to discover a new space. Buried himself
9:18 alive
9:20 inside his basement. Tongue on the side
9:23 of his face. Working away on
9:26 displacement. What it would take to
9:28 survive.
9:32 Cuz when you're done with this world,
9:37 you know the next is up to you.
9:41 And for once in his life, it was quiet.
9:48 As he learned as he learned how to turn
9:51 with the tides
9:53 [Music]
9:54 and the sky was a flare and came up for
9:57 air.
9:59 [Music]
10:02 All right. All right. Enough bad song
10:05 playing and bad song singing. Played
10:08 with Pika AI today. So much fun. Yeah,
10:11 Pika. Pika. I'll tell you what Pika is
10:13 doing well.
10:15 Um, and I think their social app really
10:18 exemplifies it.
10:21 Um,
10:23 most of the frontier model companies,
10:25 most of the video companies, let me
10:27 switch my videos here. So the people the
10:30 good people have a good view.
10:34 This simal casting on Tik Tok and
10:36 YouTube is special.
10:39 It's something we've chosen to do.
10:42 Who is not smarter other than Pate?
10:48 Exactly. I I would say Pate Pate's
10:51 probably right up there. Um
10:56 uh what was I talking about? What was I
10:58 talking about? Oh, Pika. What's Pika
11:00 doing? Well, okay. So, most of the
11:03 frontier model companies, most of the
11:07 creative companies, the image generator
11:09 companies, the video generator
11:11 companies, the song generator companies,
11:15 um,
11:17 are kind of assuming
11:20 that their audience wants to geek out.
11:28 I don't know.
11:31 Oh, don't worry. Okay. Um,
11:40 they they assume that their users are
11:42 going to be people that want to geek out
11:43 on prompting, that want lots of
11:45 controls. If you look at a tool like
11:47 Leonardo.ai,
11:49 it's got all sorts of [ __ ] in it, right?
11:52 and all sorts of different features and
11:54 dials and it's like the cockpit of a 747
11:56 and you can tweak it and you can make
11:58 luras and
12:00 styles and whatever videos all that
12:04 stuff.
12:06 What Peak is doing is they're kind of um
12:11 they're experimenting with like the
12:14 video generation equivalent of casual
12:16 gaming. So, so remember we used to have
12:19 the these big AAA video games, right?
12:21 And then like the iPhone came along and
12:24 the app store came along and this whole
12:26 idea of casual gaming where you can just
12:28 have a game that you can just play with
12:30 your thumb while you're standing in
12:31 line, just casual [ __ ] around, right?
12:34 You know, shoot basketballs at a basket
12:36 or ski a ball down a hill, whatever it
12:38 is. Um, PA is essentially pre-prompting
12:44 um, their video model and putting it
12:46 behind buttons, right? And so you can
12:49 just go in and you type in a subject and
12:51 then you can say, "Turn this into a
12:53 video where you peel the subject off
12:55 like a sticker or you crumple it up like
12:58 paper and they've added sound effects
12:59 in." So they're so they're basically
13:02 taking a few tools, duct taping them
13:05 together with a clever prompt and
13:07 putting a button on front in front of
13:08 it.
13:10 Um,
13:13 what they're doing with Pika Social
13:18 is you've got a thing that looks like
13:21 Tik Tok
13:28 pika social pa.
13:33 All right, I've got
13:35 14% so I should be okay.
13:38 Okay. So, you've got
13:43 [Music]
13:45 >> right,
13:47 you got your share button, you got your
13:48 comment, you got your heart, you got
13:50 your plus button, you got your profile.
13:52 Everything's just like Tik Tok, right?
14:01 >> All right.
14:03 So,
14:05 what's different? Well,
14:10 it's 100% AI generated and it's 100%
14:16 photo to lip sync
14:20 or like photo to dance.
14:23 So, you can't even record yourself. You
14:25 can't record video of yourself and have
14:27 it modify that. And I think they're
14:30 doing that on purpose. I find it really
14:31 fascinating. So, if I go plus,
14:34 it's it's saying take a picture of
14:36 yourself. So, I'm going to be like,
14:40 you know, I take a lovely picture of
14:42 myself. I can't record video.
14:46 There's a little animate button at the
14:48 bottom. There's a little audio thing at
14:50 the top. So, again, just like Tik Tok.
14:52 And then these things on the side.
14:55 AI speech. So, you can have AI
14:59 synthesize it or you can record it.
15:02 change the look and add a performance.
15:05 Now, if I add the performance,
15:09 there's lip sync, fast rap, emotional
15:12 singing, TED talk,
15:15 sad, playful, sassy, and then you can
15:17 describe your performance. So, what
15:20 these do if you click on them is they
15:24 well they in the background they add a
15:26 uh a prompt, but you can also just type
15:29 in a prompt.
15:34 And then if you go to looks edit look
15:37 error
15:39 still in still in preview.
15:44 Let's go. Come on.
15:47 You can do it. Network connection was
15:49 lost. Oh, this keeps happening on my
15:51 iPad. I think it's my shitty Wi-Fi.
15:54 Um, let me pop out of that app. Pop back
15:58 in. Go there.
16:00 Let me quit the app. Come back in.
16:06 Do plus. Take a picture.
16:09 Okay. Edit. Look. Okay. So, there we
16:13 are. And then I can, you know, I can
16:15 just go like hair color. Um, pink. Bang.
16:20 And it's going to add pink hair color.
16:24 Tell the boys to get off Call of Duty.
16:30 So,
16:34 so that's that done. And then I can go
16:37 add sound.
16:39 And I can do AI speech. And let's see, I
16:42 can do southern dude.
16:44 >> Hello, this is your clone voice. Thanks
16:47 for using Pika.
16:50 >> Or I can just record something.
16:56 Well, you're not just special. Look at
16:59 me with my little pink hair. Do you like
17:00 my pink hair? I like my pink hair. You
17:02 know what I should do? I should put on
17:04 my pink bow with my pink hair. All
17:07 right.
17:09 So, that should be awesome.
17:11 >> Well, ain't that just special? Look at
17:14 me with my little pink hair. Do you like
17:15 my pink hair? I like my
17:17 >> solid performance. That's like That's
17:20 award-winning acting. I I have
17:22 >> And I'm just special. Sh. calm down.
17:27 I have I have a I have a fine arts
17:29 degree in acting. So, you can tell that
17:33 there, right? And so, now we we recorded
17:35 that sound. We've got that picture and
17:38 we'll just animate it and now that's
17:40 going to turn it into a thing. This is
17:41 the whole app. So, the whole app is you
17:44 take a picture of yourself,
17:47 modify it somehow, add some sound to it,
17:49 and it does the lip sync.
17:52 You can also take anyone else's video.
17:54 If you like their video, you can use
17:56 their look, their sound, their motion,
17:59 or all three. So, you basically just
18:02 say, "I want to you I want to make a
18:04 version of me in their video." And then
18:07 you take a picture of yourself and it
18:08 applies all their crap to you, right?
18:12 And then off we go. Um, so here's what I
18:15 think's interesting about this.
18:20 They're training us.
18:23 They're training us to get creative
18:27 with AI generated video. And at the same
18:30 time, they're training us to accept that
18:34 an all AI video channel
18:37 isn't necessarily a bad thing, right?
18:40 One of the tropes in society right now
18:44 is well, was it AI generated
18:48 with the with the pursed lips? Listen
18:51 here, little mister. Did you use AI to
18:54 generate this? Cuz if you did, that's
18:55 what you're an awful person,
18:58 right? And they get that shitty look on
19:00 their face and they're all judgmental
19:02 and you're like, "Uh, I was just [ __ ]
19:05 around and having fun." Well, aren't you
19:07 just the art thief of the century,
19:09 stealing from all those artists, world's
19:12 greatest plagiarism machine? Aren't you
19:15 proud of yourself? Right.
19:17 >> Well, and I'm just special. Look at me
19:20 with my little pink hair. Do you like my
19:22 pink hair? I like my pink hair. You know
19:24 what I should do? I should put on my
19:25 pink bow with my pink hair.
19:29 >> Well, and I'm just special.
19:33 >> That's pretty good.
19:37 So, I'm going to put the the caption.
19:39 Where's my pink bow?
19:41 Where's Where's my pink bow?
19:50 And then we're going to post that.
19:53 And so now, if you're on the Pika social
19:55 app, you may run in.
19:56 >> Ain't that just special? Look at me with
19:59 my little pink hair. You like my pink
20:00 hair? I like my pink hair. You know what
20:02 I should do? I should put on my pink bow
20:04 with my fake hair.
20:07 >> It's
20:07 >> Well,
20:08 >> it's actually the the the lip sync
20:11 animation is actually really good. Um,
20:14 okay.
20:19 So, so you've got all these people
20:23 who are not using AI
20:27 clutching their pearls about was it AI
20:29 generated or not? And listen, I
20:31 understand the fear of where that comes
20:33 from is
20:36 if someone puts out a video of me that's
20:38 a deep fake
20:42 that is disparaging or misleading or
20:46 whatever it might be. That's dangerous,
20:48 right? So, I think the sentiment for
20:51 where the concern comes from is fair,
20:54 but it completely discounts all of the
20:57 other possible use cases of this kind of
21:00 technology, like making a stupid silly
21:03 video or making a social network where
21:07 people could just make silly AI videos
21:09 and share them with one another and have
21:11 some fun,
21:14 right?
21:16 Play first. There's also a lesson here
21:19 in we all now have access to APIs and
21:24 incredibly powerful models where you
21:26 could put together a social network that
21:29 is a combination of Instagram and Tik
21:33 Tok and AI generated stuff and non-AI
21:36 generated stuff and just mix it all
21:38 together.
21:40 What they've chosen to do is just say
21:41 no, we're going to lean into it. We're
21:44 we're an AI video generating tool. We've
21:47 got this cool lip sync feature. What if
21:49 we just did an app that takes that one
21:51 feature and just does it really well.
21:55 Now, are they going to expand over time?
21:57 They might,
21:58 but right now that there's something
22:00 about the simplicity of that
22:04 they could easily add
22:07 text to video, video to video, image to
22:10 video,
22:12 but they're not. They're just saying
22:14 photo to um to lip sync. In fact, you
22:21 can only take a selfie. You can't upload
22:24 an image unless tell me if I'm wrong on
22:27 that, Danielle, but I don't think you
22:28 can. The way I did it last night was I I
22:31 made a really cool person in Midjourney
22:34 and then I basically just flipped my
22:37 phone around to my screen and I took a
22:39 picture of the MidJourney image like it
22:41 was a selfie. Um, saw this online. Don't
22:44 be a prostitute. Yeah, exactly. All I
22:47 know right now, James Dutder, all the AI
22:50 customer service sucks. Oops. I mean,
22:53 pretty bad. Oh, you can say sucks here.
22:56 Yep. Plagiarized an episode of Seinfeld,
22:59 I think. Just kidding.
23:02 Hey, Kyle, I was watching The Irregulars
23:04 on Netflix earlier. Oh, that's that's
23:06 really funny.
23:10 Um, man, I can't read. Oh, there we go.
23:14 That's better.
23:18 I looked up and saw the Pika video.
23:22 That Pika video, the animation in that's
23:24 really good, isn't it?
23:29 [Laughter]
23:35 Well, ain't just special. Look at me
23:38 with my little pink hair. You like my
23:39 pink hair? I like my pink hair. You know
23:41 what I should do? I should put on my
23:43 pink bow with my pink hair.
23:50 I'm a professional.
23:52 Um,
23:59 oh man,
24:01 couple of people still trying to find
24:02 the app. Okay. So, if you go in the app
24:05 store, let me go in the app store and
24:06 find it for you. It's iOS only. So, that
24:09 means Apple Apple mobile devices only.
24:16 Where's my app store? Oh, there it is.
24:24 P I K A
24:27 search.
24:29 Okay. So, if you look here, you see that
24:33 one that says Pika AI video? That's not
24:36 the social app.
24:42 These two
24:44 that say Pika Art and whatever, Pika
24:48 something Labs, those are knockoff apps.
24:54 All right. The one you're gonna look for
24:56 is the one with the the Chinese I I
24:59 assume Asian woman.
25:02 And notice how the little pika rabbit is
25:05 in 3D Chrome and it says Pika Social AI
25:09 video. That's the one. Now, you're going
25:12 to need a code for that. If you're a
25:14 member of the AI salon, which you should
25:17 be,
25:20 and you go to the irregulars,
25:26 which is down in clubs and hubs.
25:31 Yeah, I changed my tab. You can share.
25:35 Um, Danielle, three posts down, Danielle
25:39 did a post
25:41 that's got a code in it. I don't know if
25:43 this code is still live, but if it is,
25:46 Android, no, no Android. Not right now.
25:50 Just iOS.
25:53 So, not only do we get blue bubbles and
25:55 you get green ones, which is
25:57 embarrassing. You don't get Pika Social.
26:02 You got You got to be on the iOS train,
26:04 people. It's It's a class thing.
26:10 It's it's digital classism.
26:15 It isn't actually a rabbit. It's a pika.
26:18 A small aminal
26:20 that looks like a mix between a mouse
26:22 and a rabbit. Fantastic. I did not know
26:24 that. How do they monetize the app or
26:29 what do you think their game plan is?
26:31 That's a great question, David. I um I
26:36 mean right now this this is in sort of
26:39 like alpha testing or like you know
26:41 private beta testing or whatever they
26:43 call it. Um
26:46 so right now they're just testing it. Um
26:50 they make money on their um video
26:53 generation site by charging
26:55 subscriptions.
26:56 Um this social app might just be an
26:59 experiment. It might be a kind of thing
27:02 like they roll commerce into it. So
27:05 imagine,
27:08 so they've got all these templates. Like
27:10 one of the things Peak is really good at
27:11 is they figure out some prepackaged
27:14 things to be able to let you just push a
27:18 button and it makes magical results.
27:20 Well, imagine them being able to do that
27:23 with social shopping where you could
27:25 just take a snapshot of yourself and
27:27 then you could put yourself in seven
27:29 different locations
27:31 holding up a product selling that
27:33 product. So, just like people sell [ __ ]
27:35 on Tik Tok, there could be a whole model
27:37 around social social selling. Um, could
27:41 be advertising. I I doubt it's going to
27:44 be that, but it could be. Um, they'll
27:47 likely also take what they're learning
27:49 from this lip lip syncing in this little
27:51 social app, migrate that back to the
27:54 web, so you can do more sophisticated
27:56 kind of uh animations for normal video
27:59 projects, whatever normal is moving
28:01 forward. So, I don't know. I don't I
28:03 don't I don't know a ton about their
28:06 business model, but right now they're
28:08 making money with subscriptions.
28:10 All of these companies I am I I would be
28:14 shocked
28:17 with maybe the exception of Midjourney.
28:20 I would be shocked if any of these major
28:24 frontier model companies or major video
28:28 image sound companies if they're making
28:31 any money. I'm sure they're all bleeding
28:33 money right now. Right. They're buying
28:35 as many GPUs as they can. They're
28:38 getting as many users as they can.
28:39 Right. right now it's the it's the
28:42 audience grab. They're trying to get as
28:44 many audiences hooked on their platforms
28:47 as possible.
28:48 Right. One of the one of the smartest
28:50 things that OpenAI's done in the past
28:53 two or three months is they turned on
28:55 memory. Right? So Chat GPT now knows all
28:59 of your chats for the past year or so.
29:02 So you can say, "Hey, based on what
29:03 we've talked about in the past year,
29:05 what are some ideas that that I've
29:07 brought up that I never pursued?" and
29:09 it'll give you a list of them. Um, once
29:12 that's turned on, you're not going to
29:14 want to stray very far from that. You
29:16 want to keep adding to that memory,
29:18 right? So, so they're all trying to
29:20 figure out what's their sticky um
29:23 advantage that people won't leave. And I
29:24 think that social app is an interesting
29:26 one where if people really get into
29:29 making these crazy AI videos and it's
29:32 just a world of crazy AI videos, then
29:35 there's none of that weirdass [ __ ]
29:37 Was this AI generated? Yeah. All of it's
29:39 AI generated. And who gives a [ __ ]
29:43 Who gives a [ __ ]
29:47 right? The only reason you should give a
29:50 [ __ ] if something is AI generated is if
29:52 someone is trying to fraudulently
29:54 deceive you. And in that case, there are
29:58 actual laws on the books
30:01 that prevent that. It's independent of
30:04 technology. People can deceive you using
30:06 non AI, right?
30:08 I mean, we've The last 20 years has been
30:11 nothing but bizarre [ __ ] strange ass
30:15 misinformation.
30:18 Oh man. Keep them. Lick them in into
30:22 their ecosystem. Exactly. It's the
30:24 corner drug dealer phase. Yeah. Give out
30:26 some freebies. This is totally it. You
30:28 want a little bump? You want a little
30:30 bump? Some pretty good stuff this week.
30:34 Yeah, exactly.
30:37 Lock them in. That's it. Source camp.
30:40 I'm loving OpenAI's memory. No more
30:42 searching individual chats. I know.
30:44 Thank God. That was a mess. They will
30:46 eventually have cloud storage memory and
30:49 sell it at a premium. I agree with that.
30:52 Actually, I'll tell you probably a
30:54 really good business
30:56 is, and I'm trying to think how you
30:59 would do this. I guess you'd have to
31:00 have all sorts of your own
31:04 data storage and APIs and I I don't
31:07 know, you probably do it with web
31:09 scraping, but
31:11 imagine a service that you could
31:13 subscribe to that actually provides
31:17 memory across services.
31:20 Like wouldn't it be amazing that I could
31:22 not only search through my chat GPT
31:24 history but Claude and Perplexity but
31:27 also things like Midjourney and you know
31:30 I could be like hey which video service
31:32 did I did I do that video with the Civil
31:35 War guys? Oh that was in Lumalabs.
31:37 Here's the link to it. Right. That that
31:39 would be huge because
31:42 one of the things we talk about here in
31:44 the AI learning lab is play right and
31:47 just go play with as many tools as
31:49 possible. play. Just figure out what the
31:52 stuff does. Figure out what the
31:54 capabilities are.
31:58 If you're anything like me, you've got
32:00 crap everywhere. It's just like social
32:02 media. Like, how many social media
32:04 accounts did you start over the past 20
32:07 years that are just orphaned out there
32:09 swinging in the wind, right? Same with
32:11 all these AI tools. But with something
32:14 like universal memory, maybe it's a
32:16 blockchain enabled memory where you can
32:18 sub, you know, you can um opt in for
32:21 your different services. That that would
32:23 be pretty slick. Maybe an agent could do
32:25 that memory across platforms thing.
32:27 Yeah, exactly. And that's why, you know,
32:31 a lot of people right now are like,
32:32 well, you know, it's just going to take
32:34 all of our jobs and what do we do? Well,
32:39 you now have the equivalent of like PhD
32:42 level employees that work for you, and
32:44 those employees are going to get smarter
32:46 and smarter and more and more autonomous
32:48 where you can just turn them loose to go
32:50 do stuff. So, as these tools get better
32:54 and smarter and can do all this stuff,
32:57 you're going to have ideas like, "Huh,
32:59 wouldn't it be cool if blank
33:04 go start that business?" I think
33:05 entrepreneurship
33:07 is is going to skyrocket.
33:11 It's going to skyrocket
33:14 because a lot of people are going to get
33:16 laid off. They're going to be like,
33:18 "Holy [ __ ] what do I do? Let me go
33:19 apply for jobs. Oh, there are no more of
33:22 my jobs. Well, now what do I do?" Well,
33:26 maybe I should look at this thing that
33:28 caused my
33:30 my unemployment,
33:32 my undermployment.
33:36 and learn about it. Oh, I didn't know it
33:39 could do that. Oh, I didn't know it
33:41 could do that. Hey, wait. If I take my
33:43 skill of the thing I used to do and my
33:45 dream of the thing I've always wanted to
33:47 do and I use these AI tools as an
33:50 amplifier of my skills, holy [ __ ] I
33:53 could start a new business that I would
33:55 love.
33:58 We're going to see a ton of that.
34:01 And those companies are likely going to
34:03 hire
34:06 trusted people that can do things that
34:09 the founder isn't good at. And sure, the
34:13 founder can use AI to do a bunch of that
34:15 stuff, but that can get kind of lonely.
34:17 So what we're likely going to see is
34:19 two, three, four, 10, you know,
34:22 somewhere between two and 10 person
34:24 companies
34:26 that are operating like they're 150,
34:30 200, 500 person companies. It's going to
34:33 be insane.
34:38 Now, the fact that you're here watching
34:41 this means you're ahead of people. But
34:42 there's a lot of people right now just
34:44 sitting on the sidelines like, "I don't
34:46 like it. I don't like it. The robots are
34:49 going to kill us. I don't like it one
34:51 bit." Taking the jobs is stealing from
34:54 the people.
34:59 The
35:02 people being pissed off right now at AI.
35:05 I completely understand it.
35:08 Um,
35:13 even the gray and not so gray areas of
35:16 how these models were trained that was
35:19 like unethical and they shouldn't have
35:22 done it, but they did it.
35:26 Like you can be pissed off that they did
35:28 it, but they did it. And because they
35:31 did it, these models now exist. And
35:34 because these models now exist and
35:36 they're this good, they're not going
35:38 away. There is there is like just
35:43 ridiculous amounts of money being pumped
35:46 into this sector.
35:49 Ridiculous amounts. 24 year old
35:52 engineers getting $250 million signing
35:55 bonuses.
36:00 Uh, I do not see a pinned comment right
36:02 now.
36:05 All the faces tonight are absolutely
36:08 classic.
36:10 I just dropped a code in irregulars for
36:11 the chat social app. Wait, in Ire's chat
36:15 for the Pika social app. Okay, so look,
36:19 for those of you who went and got the
36:22 Pika social app, but you don't have a
36:23 code, Silverf Fox is your best friend.
36:28 She put it in the chat. So go into the
36:30 Irregulars channel, Irregulars AI
36:32 Learning Lab.
36:35 And then
36:39 let's see, Lori Blair,
36:42 should you get it, use her TTWV.
36:47 All right. Use my code. I think she's
36:49 got three. If she's like a normal human
36:52 being like me, she got three. All right.
36:57 We create whatever the hell comes to
36:59 mind and eventually something shakes
37:01 out. Yep.
37:03 It can have my job if it will also pay
37:06 my bills so I can do the thing that I
37:08 want to do as a kid. Have fun. I think
37:10 that starts to be where we go. Um the
37:14 transition period's going to suck. Like
37:16 to be clear, people are going to lose
37:18 their jobs. They're going to be more
37:20 pissed off than ever.
37:22 Um, and over the next five years, we'll
37:25 start to figure it out. People will
37:26 adapt. People adapt.
37:29 We adapted to the industrial revolution.
37:31 We adapted to the steam engine. We
37:33 adapted to the the tractor on farms that
37:37 displaced 80% of farm workers.
37:40 Now, that took four decades.
37:44 And what we're going to see is probably
37:45 going to be half a decade, maybe a full
37:49 decade
37:51 where we're going to see similar kinds
37:52 of displacement,
37:54 but we're also being given this this
37:58 knowledge tools been democratized and
38:00 given given to us all. So,
38:03 um
38:06 I'm exactly as normal as you, Kyle.
38:08 Exactly.
38:13 Oh man. Okay. So, let me let me share
38:16 something here.
38:19 Um, this image
38:30 this is
38:35 as of today
38:38 I think.
38:40 Um, this is tracking aai.org. So, so
38:44 basically what this organization does is
38:46 they take every new large language model
38:48 that comes out and they have it take the
38:51 the Mensah Norway quiz, the IQ test,
38:55 and they do it multiple times per model.
38:58 And you can see most models got to the
39:01 sort of center of the hump here, which
39:03 is basically the human average IQ of a
39:06 hundred,
39:07 right? And then you've got distributions
39:09 that go down and you've got
39:10 distributions that go up. This thing all
39:13 the way out to the right here
39:16 is Chat GPT5 Pro and it hit an IQ of 148
39:25 today.
39:28 148.
39:29 Smarter than 99.9%
39:31 of people on the planet.
39:34 So, all of the people that said large
39:35 language models can't get us here,
39:38 um, were wrong.
39:41 Uh, it got us here. Now, GPT5 Pro, it's
39:44 200 bucks a month and, you know, it uses
39:48 massive computer resources for every
39:50 prompt. One of the things Altman said is
39:52 he thinks he's going to give a handful
39:55 of GPT Pro um,
39:59 instances to $20 a month people. like
40:02 maybe we get five a month or something
40:03 like that. But if you've got a really
40:06 tough problem that you want to crack,
40:08 throw it into GPT5 Pro and it can, you
40:11 know, it can likely be
40:15 148 IQ. They they said Einstein Einstein
40:19 was in the 160 to 180 estimated range.
40:23 So, we're not quite at Einstein yet, but
40:25 we're at the 99.9%
40:29 of of all people's intelligence, right?
40:33 Um, so that's that's kind of crazy. So,
40:36 that happened today. And then I want to
40:38 show you another thing which is just
40:41 bonkers. Um, I got to find it.
40:46 [Music]
40:48 Here it is. Okay.
40:52 You can hear that, right? Right.
40:54 Brandon,
40:58 >> do you need to join the irregulars? No.
41:00 If you if you go in and join David, if
41:02 you go in and join the AI salon, if so,
41:04 if you go to community.thesalon.ai.
41:09 So,
41:11 that URL with community at the beginning
41:13 of it, community
41:16 salon.ai.
41:17 That'll take you to the community site.
41:21 um just sign up for it there. If you
41:23 scroll down the lefth hand side, you
41:25 should see the irregulars under clubs
41:27 and hubs and then if you just click on
41:30 that, you go in there, that's where you
41:31 can see those posts that have the codes.
41:33 Okay,
41:35 beautiful. So, what I'm what I'm showing
41:38 you right here is a little minutong
41:40 movie
41:42 about the China's in in Beijing today,
41:46 they kicked off the humanoid robot
41:49 games.
41:52 Yes. Yes, they did.
42:09 [Music]
42:16 soccer. Oh no.
42:29 [Laughter]
42:33 boxing.
42:35 [Laughter]
42:50 [Music]
42:55 Now,
42:57 strange as that is,
43:01 >> I don't think it's going to be strange
43:02 for long,
43:04 um,
43:09 I don't think it's going to be strange
43:10 for long at all.
43:12 the the the
43:17 rapid progression that we've seen in
43:19 large language models over the past two
43:21 and a half and and three years. We're
43:23 going to start to see in um
43:28 in humanoid robots and they're going to
43:30 get cheaper and they're going to you
43:32 know they're going to be marketed to
43:33 consumers. China's going all in on this.
43:36 So, so part of what this event is is to
43:39 get people used to seeing them,
43:43 used to seeing their capabilities. This
43:46 to me feels like one of those videos
43:48 that we will look at five years from now
43:51 and they'll have these robots that are
43:53 running faster than any human, right?
43:55 That they're this will become a
43:57 technology capability thing. And I could
44:00 see there being different countries
44:02 competing like showing off their
44:04 technology, right? I can absolutely see
44:07 this being a thing that gets really big.
44:09 Um,
44:13 if if where your mind is right now is it
44:16 will never be normal to see a robot
44:18 walking down the street.
44:21 Well, if you've played with chat GPT
44:24 three years ago, you would probably
44:26 argue it would never you would never
44:28 find it normal that you can just say to
44:32 a computer, "Write me
44:35 you know, an epic poem in amic
44:38 pentameter about quantum physics and
44:40 it'll just write it for you. Or it'll
44:43 just you just tell it lovable to write
44:45 you a program
44:47 of a of a video game that your kid came
44:50 up with the idea on a beach and you get
44:52 back home and you're like, "Hey, make
44:54 this game for me." And 10 minutes later
44:56 your kid is playing a video game that
44:58 the computer programmed. We have
45:01 normalized so much remarkable capability
45:05 in the past two and a half years just on
45:07 this channel alone.
45:10 Like
45:13 I don't know about you, like I do this
45:15 every [ __ ] night.
45:18 Every night. And sometimes when I do the
45:21 I I just do the the basic parlor trick.
45:23 Write me a poem about quantum physics
45:25 and it just goes
45:30 It's remarkable, right? Same thing's
45:34 going to happen with robots. The first
45:36 one you're going to see, you're going to
45:37 be like, "Oh my god." Everyone will be
45:38 taking a picture of it. And then two
45:41 months later, there'll be 10.
45:44 And then a year later,
45:46 we'll be shopping with those robots. And
45:50 there'll be funny videos of a robot
45:51 losing its mind and
45:54 tripping and knocking down a bunch of
45:56 cans in the grocery store. Yeah,
45:59 self-driving cars in San Francisco was
46:01 weird until it wasn't.
46:04 Now Teslas are driving around Austin.
46:06 It's weird until it wasn't.
46:10 So, it's just crazy. Just crazy what's
46:13 coming.
46:15 Um, it sounds like we've got some new
46:18 folks here, which I'm really excited
46:19 about. If you're new here, first of all,
46:21 welcome. My name is Kyle Shannon. Um,
46:24 let's see. Ryan kind of makes sense.
46:26 Humanoid is more familiar as opposed to
46:28 the robot and Interstellar or HAL. I
46:31 think we'll see specialty robots. The
46:33 thing about humanoid robots is
46:36 our entire world is built for us, right?
46:40 So while you could make specialized
46:42 robots that are really, you know,
46:45 efficient at a particular thing, just
46:48 like large language models being
46:50 generalized intelligence, generalized
46:52 knowledge, um having a humanoid robot
46:55 can be generalized, you know, physical
46:58 physical work. So we've got knowledge
47:01 work with the LLMs and with with all the
47:03 other models out there. And then we've
47:05 got this for physical work.
47:08 Um,
47:10 okay. Let's see. Um,
47:14 Whimo Whimo driverless cars are all
47:16 something. That's what God says. Yeah, I
47:20 don't think this is a God thing. I I
47:22 really don't. The whole thing about, you
47:24 know, we're acting like God and if this
47:26 is sentient beings and things like that.
47:29 I don't think it's that. I I understand
47:32 the fear, right? And I understand
47:37 it's very easy to demonize the
47:39 technology companies and go, "Oh,
47:40 they're playing God."
47:43 No, what they're doing is they've
47:45 discovered
47:47 a technology that can provide us
47:50 remarkable access to knowledge
47:53 and
47:56 do that in this generalized way. And the
47:59 way I think about it is not like us in
48:02 competition with that intelligence, but
48:05 more like
48:07 us augmenting ourselves with that
48:10 intelligence.
48:12 So like my experience here and and the
48:14 experience of a lot of the people in the
48:16 AI salon is not that AI is this awesome
48:20 thing that like a vending machine like
48:22 you put in your prompt and out comes
48:25 genius results. That's not how this [ __ ]
48:28 works. you put in your prompt and [ __ ]
48:30 comes out of that thing most of the
48:32 time.
48:34 The way you do it is you say, "Okay, if
48:37 it's not the genius, then it's still
48:40 about my intention and my meaning."
48:44 Um,
48:45 okay. But I worry about the people that
48:48 control the prompts.
48:51 We control the prompts. You mean you
48:53 worry about the frontier model companies
48:55 that are making the the uh the models
48:58 the Oh, the one above it. I meant in the
49:00 context that it was something to work
49:03 for us.
49:09 Well,
49:12 yeah. I mean, I think I think ultimately
49:14 the humanoid robots will work for us.
49:18 Um, it is it is in it is in all of the
49:23 robot building companies interest to
49:27 sell as many as possible. And so if they
49:29 if they just use the robots to repress
49:33 us, then they don't sell as much [ __ ]
49:36 So So they want us to want robots. They
49:39 want us to to use these things. Tik Tok.
49:42 Um, I want a robot that can do my dishes
49:45 and remind me why I just walked in the
49:47 other room. Exactly. Exactly. And and
49:50 like that's the kind of thing that I I
49:52 think we'll get. Again, I think that's
49:54 the kind of thing I think we'll
49:55 normalize. Um
49:57 I don't disagree with you that you are
50:00 you are very right to be skeptical
50:05 that open AAI or X AI or Google or Meta
50:13 or any of them
50:15 are going to inherently do the right
50:17 thing. Whatever the right thing is, we
50:20 should be skeptical that they're not
50:21 going to do that. But there's something
50:23 happening
50:26 at scale that I don't think I've ever
50:29 seen before.
50:31 We saw it a little bit with the
50:33 internet, but
50:36 every time a frontier model company puts
50:38 out a model, we get an open-source model
50:41 either from China or from the US that is
50:44 nearly as good or as good and sometimes
50:47 better than that model from from the
50:49 Frontier Company.
50:51 um fully open sourced.
50:54 If you've got the fully open source
50:56 models with the weights, you can train
50:59 them to be however you want. So if you
51:03 think that the the the safety measures
51:07 that OpenAI is putting into their
51:09 models, you don't agree with with their
51:11 politics or you don't agree with their
51:13 definition of safety and you don't like
51:15 anthropics. Well, go to one of the open
51:19 source models and train it yourself. do
51:21 it yourself
51:22 and put it in the marketplace. I mean,
51:24 th this is
51:32 one of the things that drives me a bit
51:34 crazy right now is people that are on
51:38 the sidelines with AI are talking about
51:41 AI like it's being done to us.
51:46 It's only going to be done to you if you
51:49 don't use it.
51:52 Right?
51:54 It's not like they're making all these
51:56 discoveries and and all this incredible
51:59 progress and they're keeping it behind
52:01 the firewall and only giving it to their
52:03 highest paying customers, right? That's
52:07 the way AI's been for decades. What
52:10 they're doing is they're putting it in
52:12 our hands. So, we get to play with this
52:14 stuff. The 98.5% of people that are not
52:17 engineers
52:19 have access
52:21 to the best models in the world.
52:26 And
52:28 way more than 50% of people, at least in
52:31 America, are on the sidelines going, I
52:33 don't like it. I don't want to deal with
52:35 it.
52:38 It's one of the most profoundly powerful
52:40 technologies in the history of humanity.
52:42 It's not going away.
52:45 And you have two choices. If it and
52:47 listen, I could be wrong. Maybe it is
52:49 going away, but I I just I can't I can't
52:53 see a path where it goes away. So, if
52:55 it's not going away, you have two
52:57 choices.
52:59 There's two. It's binary.
53:02 You deal with it or you don't.
53:05 And if you're sitting on the sidelines
53:07 pissed off about it and you're just
53:10 watching this bullet train [ __ ] take
53:12 off,
53:14 AI is going to happen to you.
53:18 If you're like, I don't like it. I don't
53:20 like it one bit, but you know what? [ __ ]
53:22 it. I'll show up at the AI learning lab.
53:24 I'll listen to this [ __ ] lunatic rant
53:26 about how it's inevitable,
53:30 and maybe I'll try it. Maybe I'll make a
53:32 a kids book for my kid. Oh my god,
53:34 that's so cool. Can it make pictures? Oh
53:36 my god, it made pictures. It'd be nice
53:39 if that were a song. Wait, you can make
53:40 that into a song, too? Holy [ __ ]
53:45 Right.
53:49 What what what
53:52 we regularly experience in here
53:57 is not this adversarial relationship
53:59 with AI, but this relationship where
54:02 it's like I still get to have my agency
54:04 as a human being to say, "Here's what I
54:08 want in the world. Here's the audience I
54:10 want to do this for. Here's how I want
54:11 to impact them. Here's an idea that I
54:14 have. Here's an idea that I had when I
54:16 was seven.
54:18 that when stepdaddy told me I wasn't
54:20 talented, I sort of shut that idea down.
54:23 And you know what? I just remembered it.
54:26 And let let me go to chat GPT. Hey, I
54:29 had this idea when I was seven. Oh my
54:30 god, that's a great idea. Here, let me
54:32 help you articulate that and expand it.
54:35 Holy [ __ ]
54:37 And all of a sudden, people are are in
54:39 here reinventing themselves. The amount
54:42 of people who have come out of
54:44 retirement in the AI salon and in the AI
54:47 learning lab is remarkable.
54:51 They were like, "Yeah, I'm done." And
54:53 they're like, "Holy [ __ ] this is the
54:54 most exciting. I feel like I'm seven
54:56 again.
54:58 [ __ ] amazing." All right, Tik Tok
55:00 pin. Oops. And black bar. I can't
55:03 believe you didn't yell at me for the
55:04 black bar.
55:07 Will Smith is sitting over there saying,
55:09 "Hold my beer." Exactly. Exactly.
55:15 Um,
55:20 at a minimum, you owe it to yourself to
55:22 just play with these things. Don't take
55:25 them seriously.
55:27 Here's the thing. If you try AI
55:30 and you hate it,
55:32 cool.
55:36 But, I mean, really try it. Like, really
55:38 try it on for size.
55:41 Like go in and take something that
55:42 you're really passionate about,
55:44 something you really know. Like maybe
55:45 you're into gardening and go in and see
55:47 what it knows about gardening.
55:50 And then take something that you're
55:52 really shitty at. Like maybe you're
55:54 really shitty at math. Like I'm really
55:56 shitty at math. Like really shitty at
55:59 math.
56:01 But I can go in now and it can do math.
56:04 And you might have heard from people, oh
56:06 AI is not good at math. No, AI wasn't
56:09 good at math and now it is.
56:14 It wasn't good at math and then they
56:16 figured out ways to make it good at
56:18 math.
56:19 Like it can write its own Python code,
56:22 which is code that you use to do math,
56:25 among other things,
56:29 and it can write and execute its own
56:31 Python code and do complicated math.
56:34 Now, is it always perfect? No. But I'll
56:37 tell you something that'll blow your
56:38 mind. If you take something where you
56:40 think you you know in your heart you're
56:42 horrible at that thing. I could never be
56:45 a coder because my mind just doesn't
56:46 work like that. I don't have the
56:48 attention span for it. And you go to
56:50 lovable.dev.
56:53 Let's go there right now.
57:10 archetypal two YouTube comments wasn't
57:13 good at math until two weeks ago when it
57:15 won the gold medal in the math
57:16 competition. Exactly. There's basically
57:19 a 0% chance that everything goes fully
57:21 right or or wrong direction. It boils
57:24 down to who is the most effective and
57:26 cooperation always wins wins game theory
57:30 in the long term.
57:34 The here's here's a really cool
57:38 dynamic.
57:40 Oh, do I have to change my tab? Yes.
57:42 Here's a really cool dynamic. Um
57:50 the frontier model companies are
57:53 populated with engineers and
57:55 mathematicians, right?
57:58 Heavy leftbrain,
58:00 logical, intelligent
58:03 process systems, math, coding,
58:07 right?
58:09 Almost none of them have any creative
58:12 people in them.
58:15 none.
58:18 They are building tools
58:21 that are called large language models.
58:25 What I can promise you right now is that
58:27 the engineers that are building these
58:29 tools have no idea how the 98% of us
58:34 that are not engineers and heavy geeks
58:37 are going to use these tools. We have a
58:40 disproportionately
58:42 impactful
58:44 um uh uh uh influence on how these tools
58:49 are going to be used than we may
58:51 realize. No, we didn't build them and
58:54 no, we're not writing the prompts that
58:55 control them and make them safe and all
58:57 that sort of stuff, but we're using them
59:00 and we're building things and we're
59:01 putting things in the world.
59:04 And
59:06 if rather than just sitting on the
59:08 sidelines pissed off, we go, "Oh, you
59:10 know what would make it better? If we
59:12 made some things out of these AI tools
59:15 that really thought about humans first
59:17 and technology second, because the the
59:19 frontier model companies certainly
59:21 aren't doing that. We can do that."
59:24 Well, how do you do that? I don't know.
59:26 You just use things to make [ __ ] that is
59:29 more human centric. take whatever you're
59:31 passionate about and use the tools to do
59:34 that. Right? So like this is not we are
59:39 not passive bystanders
59:43 at all. Right?
59:47 At all.
59:50 In fact, Sam Alman talks all the time
59:52 about they put things in the world early
59:55 so that they can see how people use them
59:56 to go, "Oh, we didn't know people would
59:59 use it that way. Let's figure out how to
1:00:01 support that.
1:00:03 All right. Um, relevant spotlight
1:00:06 comment. Open AAI figured out
1:00:08 immediately that iterating hundreds of
1:00:11 millions of times beats the pants off a
1:00:13 few hundred engineers trying to guess.
1:00:15 Yeah, exactly. The reinforcement
1:00:16 learning thing. I was grilling an econ
1:00:20 professor of mine about um,
1:00:24 black skulls theory pretty hard. I don't
1:00:27 know black skulls theory, but we could
1:00:29 go there and do it. Anyway, let me show
1:00:30 you. We We can go to to uh to chat GPT.
1:00:34 I just don't know what that is. I've
1:00:35 heard of it before, but I don't know. I
1:00:37 don't know it off the top of my head. Um
1:00:40 and at one point he just made an
1:00:42 analogy. I'm just a carpenter.
1:00:45 That's interesting. If the tool works,
1:00:47 it works. Yeah.
1:00:49 And this is so um David Shapiro talks a
1:00:54 lot about the industrial revolution and
1:00:57 the steam engine augmented our physical
1:01:00 strength and generative AI is augmenting
1:01:03 our intellect, right? Our our our
1:01:07 uh intellectual strength. Um both, you
1:01:11 know, massively transformative. Um,
1:01:16 we didn't think we would get displaced
1:01:19 like this, you know, just like I'm sure
1:01:22 the farmers in in the late 1800s didn't
1:01:24 think they were going to get displaced.
1:01:26 Um, Dana, they they said the same thing
1:01:29 when computers came into homes and now
1:01:31 everyone has one in their pocket. Yeah.
1:01:34 Like we've got crazy ass computers in
1:01:37 our pocket now. Okay. If you've not seen
1:01:40 Lovable before, and this is this is the
1:01:43 the thing of, you know, if you've told
1:01:45 yourself that you can never do
1:01:47 programming, um I'll just do my favorite
1:01:49 one. Um, I want
1:01:52 you to build me
1:01:56 an exact
1:02:02 clone of the
1:02:05 original
1:02:08 Space Invaders
1:02:10 video game with
1:02:13 sound effects and scoring
1:02:17 and play ability.
1:02:22 Um, I want you to make the color
1:02:27 version,
1:02:30 not just the black and white. All right.
1:02:33 So, I just typed in what I would
1:02:35 consider a really crappy prompt.
1:02:38 And now, if I click the code mode,
1:02:42 you can actually watch this thing. So,
1:02:43 it's set up this whole environment for
1:02:45 all of the different files. And for
1:02:47 every one of these files, it's actually
1:02:49 writing code for them.
1:02:53 Um, it's not copying and pasting this
1:02:55 code. It's actually writing the code.
1:02:58 So, let's Here's index.html that's
1:03:00 already written.
1:03:03 Um, so it's spinning up the preview.
1:03:06 So, what did it say here? It said, I'll
1:03:08 build you a color colorful Space
1:03:10 Invaders clone with classic gameplay
1:03:12 elements. the features in this version.
1:03:15 Player ship with left right movement and
1:03:17 shooting grid of colored alien invaders,
1:03:20 collision detection with explosions like
1:03:22 like it knows what I meant when I said
1:03:24 make me a game. It knows what it needs
1:03:26 to build. It knows how to structure an
1:03:30 application.
1:03:32 And in a few minutes here, we're going
1:03:34 to be able to play this game. All right,
1:03:36 Tik Tok pin. For better or worse, I
1:03:39 finally have a tool that can move as
1:03:40 fast as my brain. Yeah, that's the other
1:03:42 thing. The other thing that is so
1:03:43 inspiring about this AI [ __ ]
1:03:46 when you when you get into it, when when
1:03:48 you when you shift out of this mode that
1:03:51 AI is this thing off to the side that's
1:03:53 going to replace me, and you shift it in
1:03:56 front of you and say, "No, no, I'm in
1:03:58 control.
1:04:00 I'm the conductor of this orchestra."
1:04:02 And the more AI tools that I can learn,
1:04:06 the more instruments I have in my
1:04:07 orchestra. And what I can do is I can
1:04:10 just sit there as the conductor and I
1:04:12 can go, "Oh, I have an idea. Let me
1:04:14 start out in chat GPT and organize my
1:04:17 thoughts and let me go over to this
1:04:18 thing and make some pretty pictures. Let
1:04:20 me go over to that thing and make some
1:04:22 movies."
1:04:24 And within
1:04:26 15 or 20 minutes, you can take what was
1:04:29 just a vapor of an idea in your head and
1:04:32 be looking at a TV commercial about it
1:04:35 or playing a video game or have a a
1:04:39 10-page investor deck,
1:04:42 right? Maybe you were at some dinner
1:04:44 party and you spouted off your mouth
1:04:48 some big idea like, "I have this idea
1:04:50 for a thing." And there was an investor
1:04:52 there and the investor said, "Well, do
1:04:53 you have an investor deck?" And you're
1:04:55 like, "Uh,
1:04:57 yeah, okay. Let me send it to me
1:05:00 tomorrow." Uh, okay.
1:05:04 In the olden timey world, you'd be
1:05:08 screwed.
1:05:10 With AI, you can go, "Hey, here's this
1:05:12 idea. I need an investor deck. Please go
1:05:15 do some research. Find all the
1:05:17 competitors in this space. Find the
1:05:19 total addressable market.
1:05:21 do all the investor best practices and
1:05:23 make me a 10 slide deck that that you
1:05:27 know follows Silicon Valley best
1:05:29 practices and 15 minutes later you'll
1:05:32 have something.
1:05:34 Um, we just asked for a Space Invaders
1:05:36 game. Here it is. Let's see.
1:05:45 All right. It doesn't have good logic.
1:05:49 Uh oh.
1:05:51 What?
1:05:53 We're screwed. We're screwed.
1:05:59 Couldn't have that collision detection.
1:06:00 But anyway,
1:06:02 you can't code now. You can.
1:06:06 And you know how you fix all the crap
1:06:08 that's broken and the fact that the
1:06:09 shapes look like [ __ ] You tell it.
1:06:13 You just go, I don't like those shapes.
1:06:15 And the bullet shouldn't go through all
1:06:16 the things. It should do a row at a time
1:06:18 and they come down too fast. And the
1:06:19 stars in the background are too fast. Go
1:06:21 fix that. And it goes and fixes it.
1:06:28 Wild times.
1:06:30 Wild times. Stacy Rodriguez, when they
1:06:33 say you can build software, everyone is
1:06:34 building games and calculator apps and
1:06:36 card games, which is cool. But can we
1:06:39 build software that is actually useful,
1:06:40 like a CRM or accounting app? So Stacy,
1:06:42 let me show you something that I made.
1:06:44 So, if you go to
1:06:45 project-cardboard.lovable.app.
1:06:55 This is a
1:06:57 adh friendly
1:07:00 uh project manager, like a to-do list
1:07:03 kind of thing that that I um that I
1:07:07 built in three hours because I was
1:07:10 trying to get all my projects out of my
1:07:12 head and on a whiteboard. And I figured
1:07:14 out a cool way to organize them on a
1:07:16 whiteboard. And I don't know why it's
1:07:18 not loading. It might be It might need
1:07:20 to spin up the server. I don't I haven't
1:07:22 run this in a while. Um
1:07:27 within three hours I had a functional
1:07:30 project management app. Yeah, you can
1:07:32 make anything you want. Like if the
1:07:35 reason games are easy is you can you can
1:07:37 know if a game works or not, right? You
1:07:40 you know what they look like. You know
1:07:41 what they play like.
1:07:43 Um,
1:07:45 yeah, this is not working.
1:07:49 Strange.
1:07:54 Let me go back to lovable
1:07:58 and log into my dashboard. Maybe I need
1:08:01 to up up update something.
1:08:06 [Music]
1:08:15 [Music]
1:08:17 All right. What's going on here?
1:08:21 [Music]
1:08:23 Oh, it did a security assessment.
1:08:30 [Music]
1:08:38 But yes, you can. Where's the dog? The
1:08:40 dog is over on the bed. Um,
1:08:44 yes, sir.
1:08:46 >> There you go. Uh, first of all, uh,
1:08:48 we're about to lose our Tik Tok bin, so
1:08:49 hang on one second.
1:08:51 >> All right.
1:08:52 >> Uh, now you got the song stuck in my
1:08:54 head. So, thank you. Um, your Superbase
1:08:58 uh, project probably spun down. uh it
1:09:02 caused it because of inactivity since
1:09:04 it's been a while since
1:09:05 >> inactive. Yeah.
1:09:08 >> So that's one thing to call out. If
1:09:10 you're doing anything on any of these
1:09:11 Vive coding sites that's using level
1:09:13 that's using Superbase Superbase if you
1:09:16 don't actively use the project, it will
1:09:18 shut down your server unless you're
1:09:20 paying for it.
1:09:21 >> Got it.
1:09:22 >> Kelly's pins back up by the way.
1:09:24 >> Okay. I came up with a new biz idea and
1:09:28 one hour later I had a full power
1:09:30 powerpoint presentation to I say show
1:09:33 for it. Oh to investors. Yeah, there you
1:09:36 go. Book them, Kyle. Yeah. So, so what
1:09:39 what happened there? Oh, wait. It's
1:09:41 back. It looks like it's back. Maybe it
1:09:43 spun it back up. Um, so I'll do
1:09:46 kylestorymind.com.
1:09:50 Um,
1:09:58 it failed to fetch. Yeah, it's it's not
1:10:00 working. So, uh, my, uh, thank you for
1:10:02 that, Brandon. Yeah, my my, uh, the
1:10:04 database here because I haven't used it
1:10:06 in a while. It spun that thing down.
1:10:08 But, yeah, you can you can do any app
1:10:10 that you want. You want to do a CRM? Uh,
1:10:13 someone someone did a CRM the other day.
1:10:15 Actually, let's let's go let's go do
1:10:17 one. We'll do one. I I thought this was
1:10:19 a really clever thing. Um,
1:10:23 I want to make a CRM
1:10:27 where
1:10:29 um the interface
1:10:33 looks like Netflix
1:10:37 where each project is a movie poster and
1:10:44 um the people are cast members
1:10:50 something like that
1:10:53 make it um
1:10:57 easy to use,
1:10:59 easy to manage and look
1:11:04 absolutely
1:11:07 fabulous.
1:11:10 And and if you wonder, can you say
1:11:13 things like that?
1:11:15 Yes, fabulous.
1:11:17 Yes, you can.
1:11:20 You can just
1:11:24 you can talk like an idiot to these
1:11:26 things. You can have them talk like an
1:11:28 idiot to you. You can talk very sternly
1:11:30 and very direct to it.
1:11:33 It understands language. It understands
1:11:36 intent. It understands nuance.
1:11:40 It understands jokes. If you're joking,
1:11:43 ah, that's funny that you called me a
1:11:46 dum dum.
1:11:49 It's it's remarkable.
1:11:50 >> Does it understand how to read sticky
1:11:52 notes?
1:11:52 >> No. Tik Tok question. Robbie, when will
1:11:56 AI be able to create completely new
1:11:59 types of AI for you? Um, when will it
1:12:01 engineer? I think we're we're we're
1:12:05 at the beginning of that right now,
1:12:07 Robbie. So, okay. So, here are some here
1:12:10 are some hints that that's happening.
1:12:13 For the first two years post chat GPT,
1:12:17 all of the frontier model companies were
1:12:19 talking about wanting to achieve AGI.
1:12:23 And AGI is artificial general
1:12:24 intelligence. And that's the point at
1:12:26 which, you know, an AI system can do
1:12:30 most of the tasks of most economically
1:12:32 viable jobs.
1:12:34 Um about four months ago, five months
1:12:38 ago, they all kind of like they
1:12:41 coordinated, but it it it was within
1:12:44 like a three-week period, they all
1:12:46 stopped talking about AGI and started
1:12:49 talking about ASI, artificial super
1:12:52 intelligence. That's where the the these
1:12:54 things are basically smarter than all
1:12:56 humans at all things.
1:12:59 Chat GPT5 Pro, the the high-end model
1:13:03 today just hit 148 on the Mensa IQ. It
1:13:07 took, you know, they did multiple passes
1:13:09 at it. It hit a 148. Um,
1:13:15 part of the acceleration that's
1:13:16 happening is they're using reinforcement
1:13:19 learning to to um have these things
1:13:22 recursively self-improve. They're
1:13:24 starting to do that process. So right
1:13:26 now in the labs, these things are
1:13:28 starting to teach themselves how to get
1:13:30 better and better and better. I built a
1:13:33 layoffs.fyi
1:13:35 type tracker for higher ed actions in
1:13:37 USA with some knowledge of that's pretty
1:13:40 cool.
1:13:42 with some knowledge of coding. Yeah, it
1:13:45 like having just even a little bit of
1:13:47 knowledge of coding just just
1:13:49 understanding how code works and how
1:13:50 it's structured and it requires
1:13:52 different files and that there's a
1:13:54 server and a database and a front end
1:13:55 and things like that and you know
1:13:58 security and authentication. Um if you
1:14:01 know the basic components of software
1:14:03 you can vibe code.
1:14:06 Now, I think there's going to be a whole
1:14:08 industry. If if you're in here and
1:14:10 you're actually a coder, if you're
1:14:11 really good at coding, I'll tell you an
1:14:13 I'll tell you an industry that I think
1:14:16 is going to be huge.
1:14:18 And it's going to be
1:14:20 um vibe coding cleaning crews.
1:14:25 vibe coding cleaning crews who come in
1:14:29 and some, you know, showoff CEO decided
1:14:33 to vibe code a CRM just like I did and
1:14:36 they're going to launch it and it's
1:14:38 going to they're going to lose all their
1:14:39 data and they're going to want someone
1:14:41 who knows coding to come in and fix it.
1:14:43 I think that's going to be huge,
1:14:46 huge, huge, huge. Mobile first app an AI
1:14:50 analytics suite. So, if we click on one
1:14:52 of these projects, people
1:14:56 Alex Rodriguez, that's pretty funny. But
1:14:58 anyway, here's here's a a a Netflix-like
1:15:01 CRM thing. Four or four projects. I
1:15:05 don't know how to make a new one.
1:15:08 All right. So, so right now this doesn't
1:15:10 work. So, I'm going to say um
1:15:13 make this functional
1:15:17 um without
1:15:20 uh
1:15:22 without a
1:15:24 database. Um,
1:15:27 I'm fine with it losing
1:15:31 losing info,
1:15:35 but I want to be able
1:15:38 to create
1:15:41 new accounts
1:15:44 and new people.
1:15:49 All right. And now it'll go off and do
1:15:50 that.
1:15:55 And I I you can there's just all sorts
1:15:57 of stuff you can do. You can change the
1:15:59 look and feel. You can, you know, one of
1:16:01 the things about vibe coding is a lot of
1:16:03 the apps all look identical. Well,
1:16:05 that's just because the people making
1:16:06 them were too lazy to say, "Hey, I want
1:16:09 it not to look like everyone else's app.
1:16:11 Make it look more like this or that."
1:16:13 Like this already looks different than
1:16:14 most vibe coding apps. They all have
1:16:16 that sort of white interface with the
1:16:18 purple rounded corner blocks, things
1:16:20 like that.
1:16:22 Um,
1:16:24 all right.
1:16:27 Echarium. Let's see. The real. AOL.
1:16:32 Great name. Oh, AOI. When it comes to
1:16:35 issues of security, how can non-coders
1:16:37 implement this? I'm sure you hear. Yeah.
1:16:40 Um, okay. So,
1:16:44 just like
1:16:48 there was a famous um about two years
1:16:50 ago, there was a famous instant where a
1:16:53 lawyer Tik Tok question. Yeah, I got it.
1:16:56 That's what I'm on. Uh a lawyer wrote a
1:16:59 legal brief and handed it to a judge and
1:17:03 the judge looked at the legal brief and
1:17:05 he looked at the citations and he said,
1:17:07 "None of these cases exist."
1:17:11 and the lawyer used chat GPT to write
1:17:14 the legal brief and didn't check the
1:17:17 citations, right? The the prior law.
1:17:21 They were fake cases. That's not on Chat
1:17:24 GPT. That's on the lawyer, right? Same
1:17:27 thing with vibe coding. If you're if
1:17:30 you're going to vibe code a CRM just for
1:17:32 a little fun experiment, you don't need
1:17:35 to worry about security. Like even put
1:17:37 in authentication, things like that. But
1:17:40 if you're going to deploy this app into
1:17:42 your company or into the world, um you
1:17:46 damn well better understand the the
1:17:49 implications, right? So there are
1:17:52 absolutely going to be scores and scores
1:17:55 and scores of people who put apps into
1:17:58 the world. There there's I heard two two
1:18:00 or three stories today. They put apps
1:18:03 into the world. They don't secure the
1:18:05 data. they don't, you know, whatever
1:18:08 you, you know, put it on on uh
1:18:10 specialized servers that, you know,
1:18:12 aren't easily accessible or or they do
1:18:15 backups of raw data, what, you know,
1:18:17 what whatever the the security
1:18:19 obvious mistakes are. Um, and
1:18:22 sophisticated hackers now have AI tools.
1:18:25 They're going to be able to just take a
1:18:26 Manis agent and say, "Go find me
1:18:29 anything at lovable.app.
1:18:32 Go find me any app at lovable.app app
1:18:34 that has the following 10 security
1:18:36 vulnerabilities and go exploit them for
1:18:38 me and it'll just go do that. So
1:18:44 if you don't know what you're doing,
1:18:45 then get someone who does, right? Or
1:18:50 have Chat GPT teach you what you need to
1:18:52 know. The other thing that's going to
1:18:54 happen is the the um the application
1:18:58 developers like Lovable, like they just
1:19:00 raised $200 million on a what was it a
1:19:05 $4 billion valuation. This is an
1:19:07 eight-month-old company. They just
1:19:10 raised $200 million.
1:19:12 Um
1:19:14 these things are going to get better and
1:19:16 better and better at plugging those
1:19:17 security holes for you. But that's not
1:19:19 the stage of the industry we're at right
1:19:21 now. The reason there's so much
1:19:22 opportunity right now is all of the AI
1:19:25 tools suck. They all suck. They're all
1:19:27 janky. They're all broken. They they
1:19:30 hallucinate. They do crappy work. It's
1:19:33 on you.
1:19:35 Now, if you want to wait until these
1:19:37 tools are bulletproof, that'll be five
1:19:40 years from now.
1:19:42 and all of the opportunity,
1:19:45 all of your friends that decided to get
1:19:47 into it now and figure it out and duct
1:19:49 tape stuff together, they they will have
1:19:51 successful companies and you'll be like,
1:19:52 "Damn, I should have done that." The
1:19:55 opportunities right now, but also the
1:19:57 risk is right now. So, you gota you gota
1:20:00 just hop on your horse and get rolling.
1:20:03 Get rolling. Yeah. Get disbarred or Oh,
1:20:06 got disbarred or reprimanded. He
1:20:08 certainly got reprimanded. I don't think
1:20:10 he got disbarred.
1:20:12 Um, Rick Olsson, when it comes to
1:20:16 issues of security, how can coders
1:20:18 implement this? That's what I just
1:20:20 talked about.
1:20:23 It's already happening. I just had to
1:20:24 clean up a beast of a vibe code of that.
1:20:26 There you go. Yeah, I was talking. I I I
1:20:29 I'm serious. Wait, Archinum.
1:20:34 Archenum. Cool name. Um, if you've got
1:20:38 the skills to be able to go in and clean
1:20:39 up a vibecoded app, [ __ ] hang a
1:20:42 shingle right now. And like I like you
1:20:44 could go to Lovable and Vibe Code your
1:20:47 website for I'll fix your vibecoded app.
1:20:53 I'm telling you, man.
1:20:55 All right, these things still don't
1:20:57 work. Completed.
1:20:59 How do I make a new one? Yeah, dumb
1:21:02 dumb. All right, it didn't do it. Cinema
1:21:05 RM
1:21:06 here's stats.
1:21:11 Um, let's let's do this. I'm gonna go to
1:21:13 chat GPT. Watch this. We'll do something
1:21:15 fun for you our keenum. Um Um, here we
1:21:19 go. So, I'm going to go um I'll I'll do
1:21:22 it with with talking. I think I can
1:21:24 dictate to it.
1:21:26 All right. Can you hear me? Yes, you can
1:21:28 hear me. Okay. So, I want to um create a
1:21:32 website
1:21:35 uh that is marketed toward vibe coders
1:21:39 who have built an application using vibe
1:21:42 coding that they're super proud of, but
1:21:45 all of a sudden realize that it's got
1:21:47 big security holes in it or big parts of
1:21:50 it don't work right and they don't know
1:21:52 what to do. Period. I'm an expert coder
1:21:55 and I can go fix vibe coding apps and I
1:21:58 want a website that's super compelling
1:22:01 um super clear and awesome.
1:22:07 All right, let's see if it heard what I
1:22:10 Yes. All right, can you hear me? Yes.
1:22:11 Okay, here we go.
1:22:16 Think longer. Okay, so so we're we're in
1:22:19 auto mode of chat GPT5.
1:22:25 And I'm trying to think where I want to
1:22:27 go do this. Do I want to do this at
1:22:29 lovable or do I want to do this
1:22:30 somewhere else? I could do this right
1:22:33 within chat GPT. Why don't we do it
1:22:35 within chat GPT?
1:22:38 Oh, it's actually just making me the
1:22:39 website. Oh, that's interesting. All
1:22:41 right. Well, [ __ ] it.
1:22:44 Vibe Vibe App Rescue onepage website.
1:22:49 [Laughter]
1:22:51 All right. It's It's writing us our
1:22:53 website now. And we'll be able to look
1:22:55 at this here in a second.
1:22:58 Um,
1:23:00 do you use Claude code at all? I haven't
1:23:02 I haven't used claude code. Um, any of
1:23:05 the things I'm not a real coder, so any
1:23:07 of the things like cursor or or wind
1:23:10 surf um I've used replet agent a little
1:23:13 bit. Um, but but not a ton. Um, I'm more
1:23:17 of a front-end guy and just an idea guy.
1:23:20 So once I get into needing to track all
1:23:25 of the different components myself, I'm
1:23:27 like, "Fuck it. I'm tapping out." Um,
1:23:30 but um,
1:23:33 was it Manis? Yeah, Manis today. Today
1:23:38 added Claude code into Manis and they're
1:23:41 doing all of that stuff for you. So, you
1:23:44 can go into Manis right now into their
1:23:46 developer mode and just tell it you want
1:23:48 it to write an app and it will do all of
1:23:50 the database setup and all all of the
1:23:52 stuff that you normally have to
1:23:54 configure within um within cloud code
1:23:57 and within cla uh within cursor. Um it
1:24:01 just it's doing that for you. So, I
1:24:04 haven't played with that yet, but again,
1:24:06 these these things are going to get, you
1:24:08 know, better and better and better. All
1:24:09 right, let's let's look at our website.
1:24:11 Here
1:24:15 we go. Your vibecoded.
1:24:17 Your vibecoded app ships fast. I make it
1:24:20 safe, solid, and fast. Well, that's bad
1:24:24 copy right there. There's case studies.
1:24:27 Sound familiar? So, that looks like
1:24:28 [ __ ] So, here's what we're going to do.
1:24:30 We're going to close this. We're going
1:24:31 to say, "Chat GPT, that's [ __ ] boring
1:24:34 and awful." I'm going to say, "Why don't
1:24:38 you write me
1:24:40 a creative
1:24:43 marketing
1:24:45 brief
1:24:48 I can give to an app developer?"
1:24:54 Okay. YouTube question. Out of
1:24:57 curiosity, how does it know when to cut
1:25:00 you off
1:25:02 if it's based on compute power it takes
1:25:05 to answer the question? I don't I don't
1:25:08 understand what you mean when it knows
1:25:09 how to cut you off. Um, what's happening
1:25:13 in chat GPT5 right now is it's got these
1:25:17 three different modes. Well, four if you
1:25:19 if you're a pro member. It's got fast
1:25:23 like thinking light. So reasoning light,
1:25:27 normal reasoning and then pro-reasoning.
1:25:29 All of these reasoning models are um
1:25:35 you give it a prompt and it's got sort
1:25:37 of an internal dialogue that it does. So
1:25:40 all of these are basically just the
1:25:42 amount of cycles that it allows itself
1:25:44 to do before it gives you an answer. So
1:25:47 pro it it just goes for a really long
1:25:49 time before it gives you an answer. Mini
1:25:52 is actually a different model. Um but
1:25:54 it's it's just a a smaller lighter model
1:25:57 that does the thinking. Fast is just
1:25:59 like chat GPT40 where you give it a a
1:26:01 prompt and it answers you. The auto mode
1:26:04 is basically dynamically switching
1:26:07 between these. And I I don't like I
1:26:09 don't know exactly what they're doing to
1:26:11 analyze that, but that's what auto mode
1:26:14 does. It's it's a routing model that
1:26:16 basically says based on what the
1:26:18 person's asking for, I'm going to choose
1:26:20 one of these different models to to to
1:26:22 use. Okay. So, here's our creative brief
1:26:26 for the Vibe app rescue.
1:26:29 So, we're just going to copy this
1:26:32 and we're going to go back to Lovable
1:26:36 and we're going to do a new
1:26:40 a new app. And I'm just going to paste
1:26:42 that uh brief into Lovable
1:26:48 and say go. And so now this is we'll see
1:26:51 if this is any better. This should be
1:26:53 better. Lovable is pretty good at this
1:26:54 stuff.
1:26:57 Do you do the same prompt in Grock to
1:27:00 see the difference in the output? Carl,
1:27:01 that's a great great
1:27:04 yes. So, um,
1:27:08 one of the things I love to do is take
1:27:12 this same prompt and and play in a bunch
1:27:16 of different models. So, go into Gemini,
1:27:18 go into Grock, go into um, Anthropic,
1:27:23 Claude, go into Chat GPT.
1:27:26 Um, or if you're doing if you want to
1:27:28 play around with these agentic tools, do
1:27:32 a prompt in Manis, do a prompt in
1:27:33 GenSpark, and do a prompt in chat GPT
1:27:36 agent mode and just see what they all
1:27:39 do. See how long they take. See how
1:27:41 sophisticated the answer is. Like we
1:27:43 just did here. We were just over in chat
1:27:44 GPT. It made us kind of a crappy
1:27:47 website.
1:27:49 Then I just said, "Hey, Chat GPT,
1:27:50 instead of writing me code, why don't
1:27:52 you write me a creative brief?" And so I
1:27:54 just pasted that into Lovable. We'll see
1:27:56 how it does. And yeah, you could paste
1:27:58 this into other vibe coding apps for
1:28:01 sure. That's a great way that that's a
1:28:04 great one of the things we talk about in
1:28:07 the AI salon is this idea of play first.
1:28:10 Play first, mindfully create, generously
1:28:12 lead is the what we call the cycle of AI
1:28:15 readiness. And so that first one, that
1:28:17 play thing is rather than feeling like
1:28:19 you've got to master a tool and you've
1:28:21 got to make your work more efficient.
1:28:23 No, no, no. What you have to actually
1:28:26 understand is what these tools make
1:28:28 possible.
1:28:30 Because it's possible, it's likely that
1:28:33 these things are way, way more capable
1:28:37 than you think is even possible.
1:28:41 And the only way you can find that is to
1:28:43 play outside of the thing that you know
1:28:47 that's this small.
1:28:49 Just push the edges. Do creative things
1:28:52 if you're not creative. do coding things
1:28:54 if you are creative. Do writing things
1:28:57 if you're visual. Do visual things if
1:28:59 you're a word person.
1:29:02 Um
1:29:04 because in doing that, what you'll
1:29:06 realize is, oh my god, I didn't know I
1:29:08 could do that. And then all of a sudden
1:29:11 that's a new possibility in your brain.
1:29:14 So when you get into the next phase to
1:29:15 mindfully create, hey, I've got a
1:29:17 problem I want to solve. Now you
1:29:19 understand the capabilities here and you
1:29:21 understand what you're asking is like
1:29:23 don't these things aren't some of these
1:29:25 tools better at some things than the
1:29:27 other. Yes.
1:29:29 How do you know that? I just you got to
1:29:32 play with them enough to understand. Oh,
1:29:33 okay. Claude's better at this thing. I
1:29:35 like the way Claude does this. I like
1:29:37 the way this one writes. I like the way
1:29:39 this one does research. When it comes to
1:29:41 images, okay, I've got one that I use
1:29:43 for pretty art. I've got another one
1:29:45 that I use for text
1:29:47 and logos,
1:29:50 right? The only way you're going to get
1:29:52 there is to play.
1:29:56 And seriously, think about yourself as
1:29:59 as a
1:30:01 an orchestrator, a conductor.
1:30:04 And the more tools that you can learn
1:30:06 the capabilities of, it's like you've
1:30:08 got virtuoso m musicians all around you.
1:30:14 Right? You need to know where to go and
1:30:16 which musician to point to. But you can
1:30:18 just sit there with your idea and go,
1:30:20 "Oh, I'll go try this tool for that.
1:30:21 I'll try this tool for that. Oh, those
1:30:23 two work really well together."
1:30:26 Right?
1:30:28 All right. What's this thing doing here?
1:30:30 She's making us a stupid website. Come
1:30:32 on, man. Come on, man. Where's our
1:30:35 website, man?
1:30:39 It's got a bunch of JSON. It's got some
1:30:41 tailwind action in it.
1:30:44 Oh, here we're done.
1:30:50 Working on Galaga. Yeah, exactly. The
1:30:53 reason we do video games is video games
1:30:55 are fun.
1:30:57 There we go. This looks like a website,
1:31:00 right? Your vibecoded app ships fast. I
1:31:04 make it safe, solid, and fast. Fast is
1:31:06 wrong here. So, we can actually go
1:31:07 change this. Let me go edit this.
1:31:11 Go edit.
1:31:13 Then I can go I make it make it sa safe
1:31:16 solid. Wait, I make it your vibe coded.
1:31:21 I make it safe, solid, and secure.
1:31:26 Right. Boom.
1:31:30 30 minute triage call.
1:31:33 Problems I fix fast. Security holes.
1:31:36 Performance bottlenecks. Flaky features.
1:31:39 scaling nightmares. Here, let me get out
1:31:40 of edit.
1:31:42 Um, user trust issues, technical debt,
1:31:47 right? From crisis to confidence,
1:31:52 free 30 minute call to triage, how bad
1:31:54 you [ __ ] it up. One to two days, I'll
1:31:56 do a health check. One to two weeks,
1:31:58 I'll secure it and stabilize it. Two to
1:32:00 four weeks, I'll actually make it not
1:32:02 suck.
1:32:04 This is a [ __ ] business.
1:32:07 Boom. done.
1:32:12 I made a a Galaga a Galaga game in
1:32:15 Perplexity Labs. Yeah, that's pretty
1:32:16 cool. Perplexity is pretty good. Um, but
1:32:20 anyway, Arum and anyone else here who's
1:32:23 a coder, I'm telling you, the orphan the
1:32:26 orphan rescue industry in vibe coding is
1:32:30 going to be big.
1:32:32 It's going to be big.
1:32:36 Now,
1:32:40 you should also learn to vibe code,
1:32:42 right? You can go in and you can fix
1:32:45 [ __ ] up vibe coding things, but you
1:32:47 should actually understand which of
1:32:49 these things are the good things. I
1:32:51 would imagine that cursor makes similar
1:32:53 kinds of mistakes over and over again. I
1:32:55 would assume that lovable makes
1:32:56 different kinds of mistakes over and
1:32:58 over again. How do you know that? By
1:33:01 playing with them, by doing it yourself.
1:33:03 At some point, one of these things is
1:33:06 going to fix their vibe coding platform
1:33:09 so that they may not need you anymore.
1:33:11 That would be a good thing to know if
1:33:13 you're basing a whole business on fixing
1:33:15 stuff, right?
1:33:18 Lovable.dev/partners.
1:33:20 There you go. Um, I assume what that is
1:33:23 is
1:33:27 a whole referral system that they've got
1:33:29 for people to come in and unfuck up
1:33:34 Hire a lovable partner. There you go. Go
1:33:36 sign up for this.
1:33:39 Perfect.
1:33:44 Yeah. Yeah. Here's companies. How many
1:33:46 are in here? Six. Eight.
1:33:51 There's eight companies in here that
1:33:53 will go fix lovable things. They just
1:33:55 raised $200 million. They've got a
1:33:57 hundred million. They got to Didn't they
1:34:00 get to 100 million users,
1:34:02 Brandon? I think they did.
1:34:08 No, 200 million. 200 million in
1:34:10 investment, I thought. And I thought it
1:34:12 was 100 million users. Anyway, whatever.
1:34:16 But this is one company. There's only
1:34:18 eight eight people that can fix stuff.
1:34:22 Go sign up. Starting at $100 an hour, 75
1:34:25 bucks an hour, starting from 1,800
1:34:28 bucks, starting from three grand,
1:34:30 starting from a grand. Starting from
1:34:32 $4,500 a month,
1:34:35 starting at $10,000 a month, 200 $200 an
1:34:38 hour, right?
1:34:43 Everyone's going to get laid off. Well,
1:34:45 yeah. a bunch of people that are going
1:34:47 to get laid off are going to start these
1:34:49 companies
1:34:51 and other companies and other companies
1:34:53 and other companies, right? Because
1:34:56 every one of these new things is going
1:34:58 to create new problems that we'll need
1:35:02 to go in and fix. We'll adapt.
1:35:04 Adaptability. Adaptability.
1:35:07 It is the
1:35:10 the skill of the next five years.
1:35:14 Don't get too attached to what you built
1:35:16 that's genius because someone's going to
1:35:18 steal it
1:35:21 or or just AI will just do it, right?
1:35:24 The thing it can't do right now, it's
1:35:26 just going to do.
1:35:28 And then
1:35:30 every three weeks there's new
1:35:32 technology. The technology that you
1:35:35 thought you knew two weeks ago is now
1:35:37 different. It's now different again.
1:35:39 Just adaptability,
1:35:42 being nimble, being zen about, oh, [ __ ]
1:35:47 changed. That thing that was broken, not
1:35:50 broken anymore. Next. Move it on. Move
1:35:53 it on. Move it on. Okay.
1:35:58 [Music]
1:35:59 Who stole your song?
1:36:03 I wonder why someone smarter than me
1:36:04 isn't thinking about different
1:36:06 refrigerant than water.
1:36:11 Oh, nothing creates opportunity quite
1:36:13 like companies that underestimated how
1:36:15 much they needed the people that they
1:36:17 just laid off. That's going to be
1:36:18 another one.
1:36:20 Oh, we [ __ ] up. We shouldn't have
1:36:22 fired all those people, right? I'll come
1:36:24 in with my SWAT team of 10 people and
1:36:27 we'll, you know, we'll write your
1:36:29 company.
1:36:31 That That's a whole another one. Okay.
1:36:33 So,
1:36:35 um I gotta go. My voice is going. I
1:36:37 haven't drunk enough water and I'm I'm
1:36:39 uh I'm crispy. But um I'm really glad
1:36:42 there were some new people in here
1:36:44 tonight. Um go check out the AI salon if
1:36:48 you haven't. Go to community.thesalon.ai
1:36:54 and that's going to take you into our
1:36:56 community. Sign up for it. Here's why.
1:37:02 Nobody right now, nobody
1:37:05 knows what the future's going to look
1:37:07 like, knows how these tools are going to
1:37:09 evolve, knows how the future of work is
1:37:11 going to evolve,
1:37:13 knows what are going to be the new jobs.
1:37:17 The only way that I know that you can
1:37:20 insulate yourself a little bit from
1:37:23 what's coming is to be in a community of
1:37:25 people that are trying to figure this AI
1:37:27 stuff out as well. You don't want to be
1:37:30 hanging out with people that are on the
1:37:31 sidelines just pissed off about AI
1:37:34 because AI is going to hit them upside
1:37:35 the head and it's going to be ugly. You
1:37:38 want to be in in a community with people
1:37:40 that are trying to figure this out so
1:37:41 that as new things come along they're
1:37:43 like, "Oh, I know what to do with that."
1:37:45 And that'll give you an idea. And then
1:37:47 you'll give someone else an idea.
1:37:50 I talked before, play first, mindfully
1:37:53 create, generously lead. The AI salon is
1:37:56 about people sharing what they learn and
1:38:00 and being generous in in how they
1:38:02 support other people. I don't know if
1:38:03 you've noticed, if you're new here
1:38:04 tonight, have you noticed that there's a
1:38:07 lot of people in here that are like
1:38:08 answering questions and doing things if
1:38:10 I miss them? Those people are called the
1:38:13 irregulars. And they're irregular. We
1:38:16 call them that because they're weird.
1:38:20 They show up here night after night
1:38:22 after night after night. A bunch of
1:38:24 stuff that I've talked about tonight,
1:38:26 they've heard me talk about before.
1:38:30 You know what they do when they hear me
1:38:32 talking about stuff they've heard
1:38:34 before? They're off in AI playing with
1:38:36 it or they're in here answering
1:38:37 questions. They're making relationships.
1:38:39 They're building trust with people.
1:38:42 That's what the AI salon's about. Okay.
1:38:46 Next Tuesday, we've got an AI salon meet
1:38:48 and greet. So, go join the online
1:38:50 community. And then next Tuesday we have
1:38:54 uh what we call a meet and greet and
1:38:56 it's a very special one because we're
1:38:58 doing the official launch of the AI
1:39:00 readiness training program which if you
1:39:03 want to check it out the site is now
1:39:04 live. It's are you ready for a.com
1:39:10 and I think we have a banner in there
1:39:12 don't we?
1:39:16 Yes. Is this it? Yeah. If you go to are
1:39:21 you readyforai.com
1:39:23 right there,
1:39:26 that's the AI readiness training
1:39:28 program. This is a very comprehensive
1:39:31 program that was inspired by a bunch of
1:39:34 um we we did a 24-hour
1:39:37 thing called AI Festivus last January
1:39:40 and the the talks were really
1:39:42 remarkable. And so this is all of those
1:39:44 talks, but we we extracted out of it
1:39:47 what were the universal truths, what
1:39:49 were the universal lessons we learned
1:39:52 over that 24-hour period that can get
1:39:55 you into the mindset to be successful
1:39:56 about AI. That's what that that's what
1:39:58 that program is. And then the last thing
1:40:01 I'll I'll talk about tonight, last
1:40:03 little promo
1:40:04 is tomorrow night's what we call Friday
1:40:07 night date night. So, so we'll do one of
1:40:09 these again tomorrow night and and it'll
1:40:11 be normal time, 8:00. Um, it's just a
1:40:14 Friday night. Bring nachos or or hot
1:40:16 pockets, whatever, whatever you enjoy.
1:40:19 Um, but tomorrow at 11 a.m. Mountain
1:40:22 time on LinkedIn is what I call AI
1:40:26 office hours. And it's basically just a
1:40:30 one-hour hangout with people who are
1:40:33 part of the AI salon part people that
1:40:35 are part of this some people that just
1:40:37 come you know from LinkedIn and just
1:40:39 hang out there. It's really a remarkable
1:40:41 group of people. Um so if you want to
1:40:43 meet some people um come tomorrow for AI
1:40:46 office hours. So how you find it is you
1:40:48 go to my uh LinkedIn. So it's Kyle
1:40:52 Shannon on LinkedIn. Uh I'm the CEO of
1:40:55 Storyvine and the AI salon. So you'll be
1:40:57 able to find me. I look like this. My
1:40:59 picture's there.
1:41:03 Um, so yeah. So, so come to AI office
1:41:06 hours tomorrow and then come tomorrow
1:41:08 night for this uh AI learning lab. All
1:41:12 right.
1:41:16 That was a rant tonight.
1:41:23 But thank you all for hanging out. I
1:41:25 really appreciate it.
1:41:27 And uh All right, looks like let me just
1:41:31 look at some comments here.
1:41:38 [Music]
1:41:41 Thank you, Ryan. Thanks for the nice
1:41:42 words. Thanks for the good questions. By
1:41:44 the way, I love questions in here. I
1:41:47 even love questions. You know, sometimes
1:41:48 people come in and like, "Isn't AI just
1:41:50 evil?" Like, you know, they'll come in
1:41:52 and be a little bit trolly. I don't mind
1:41:54 trolly kind of answers or questions. Um
1:41:59 because one of the things I recognize is
1:42:01 that this shit's scary.
1:42:04 And especially if you're not in it, it's
1:42:08 scarier. But even if you do get in it
1:42:11 there, you might have ethical concerns,
1:42:13 you might have safety concerns, you
1:42:14 might have, you know, security concerns,
1:42:17 you might have future of work concerns.
1:42:19 Those are all really valid.
1:42:21 Um, my antidote to that is to be in the
1:42:26 conversation. And so I absolutely love
1:42:29 it when people come in here with those
1:42:31 kind of questions so long as they're
1:42:32 willing to actually like give it a shot.
1:42:36 Like try it on. Like come in, have those
1:42:39 questions, have those doubts, and play
1:42:41 with AI. Play with it long enough until
1:42:44 you have your first what I call Kevin
1:42:46 Mallister moment, right? Remember Home
1:42:48 Alone?
1:42:50 that moment.
1:42:52 Once you have that moment with AI, you
1:42:55 kind of can't unhave it.
1:42:58 Once you actually get
1:43:02 the generative part of AI, the G in GPT
1:43:06 stands for generative.
1:43:09 Once you get what that actually means
1:43:11 and what it can do for your ideas and
1:43:13 and your projects and your work,
1:43:18 it's hard to go back from it. Um, and
1:43:21 that's all I ask is like try it enough
1:43:23 that you have one of those moments. And
1:43:25 then if at that point you're like, you
1:43:27 know what, [ __ ] it. I'm tapping out. I'm
1:43:28 going to the woods. I'm getting I'm
1:43:31 going to dust off my prepper kit and go
1:43:33 move to the woods. More power to you.
1:43:37 But I think what you'll find is it can
1:43:39 be very very inspiring if you get into
1:43:41 it. All right. The real just joined the
1:43:44 salon. Awesome. That's great. Great.
1:43:47 Great. Great.
1:43:51 Oh, I was I wasn't saying that you meant
1:43:53 you meant it in any way negative, Ryan.
1:43:56 I I was just saying sometimes people
1:43:58 come in here and they do, and I'm I'm
1:44:00 fine with that, too. I didn't I didn't I
1:44:02 thought your questions were were great
1:44:03 and very productive. Um, good job
1:44:06 teacher. Thank you, Peter. I appreciate
1:44:08 that.
1:44:11 My grandson thinks AI is creepy and he
1:44:14 works in pest control. There you go.
1:44:18 All right, everyone. Have a good night.
1:44:19 I'll see you at office hours tomorrow at
1:44:21 11 and then I'll see you back here
1:44:23 tomorrow night for Friday night date
1:44:24 night. All right. Peace out, everyone.
1:44:26 Have a good night.