10/14/2025 - English Is the New Programming Language: How AI Empowers the 99% Who Are Not Coders

Description

Chapters

Transcript

0:05 [Music]
0:18 [Applause]
0:19 [Music]
0:28 Wow.
0:31 [Music]
0:32 Oh,
0:39 [Music]
0:56 every time I see it
0:59 Get that look in mine.
1:02 Every time I see your mouth, I hear that
1:06 smile.
1:09 The early misty morning light. I heard
1:13 the engine turning
1:15 the old photo.
1:20 [Music]
1:23 You were leaving me
1:27 today.
1:30 You will convince me
1:35 [Music]
1:40 again today.
1:43 You're leaving this old town looking for
1:47 someone else's golden ring.
1:50 [Music]
1:52 should have said so long
1:57 [Music]
2:01 cry
2:03 [Music]
2:06 so long.
2:08 [Music]
2:12 Don't you grab me.
2:16 [Music]
2:19 Oh yeah.
2:21 [Music]
2:45 Aha.
2:49 [Music]
2:52 [Applause]
2:59 [Music]
3:05 Which programming language would you say
3:08 you're most proficient?
3:10 This is an easy one. English.
3:14 [Laughter]
3:15 [Music]
3:18 English is the new programming language.
3:23 Uh I'm not a coder. I'm not a coder.
3:26 Well, this is called the AI learning
3:27 lab. How could you do AI without coding?
3:29 Uh because you can.
3:32 Um,
3:35 [Music]
3:40 I mean, I suppose pushed it would be
3:42 HTML, but the HTML I know is from the
3:45 '9s, so it doesn't look anything like
3:48 the HTML that you do today.
3:52 So,
3:56 um, Kyle, want to teach us how to use
3:58 perplexity to merge multiple documents?
4:02 Uh, if I knew how to do that, I would
4:04 love to.
4:06 What is it with the stumpers?
4:09 Everyone coming in. Sir, what are your
4:11 qualifications, sir? Yes, I was
4:13 inquiring, sir, when it comes to
4:16 programming languages. Sir, would by
4:19 chance pray tell your qualifications?
4:22 I ain't got none. Rather, carry on then.
4:31 Stump Kyle Tuesday.
4:38 Um, so but seriously, um, if you're new
4:41 here, welcome to the channel. This is
4:43 the AI learning lab. And if you're like,
4:45 why are you playing guitar? I thought we
4:46 were here to learn code. Um, so here's
4:49 my here's my um let me give you a little
4:52 bit about my background. I
4:56 I have been an entrepreneur I don't know
4:59 most of my life
5:01 and in the mid 90s I started one of the
5:04 first online magazines and one of the
5:06 first digital agencies in 1994. So I was
5:09 there for the very beginning of of that.
5:12 The reason I got into the the web
5:15 development that early is that I'd been
5:18 intrigued by the internet for years, but
5:20 for years the internet had been the
5:23 purview of data scientists and engineers
5:26 and librarians and it was command line
5:28 only and you kind of had to know IP
5:31 addresses and [ __ ] like that. And it was
5:33 very very technical and I'm I'm I'm not
5:36 non-technical. I'm I'm fine at product
5:39 development, you I know how to do [ __ ]
5:42 but I'm not a programmer. I'm not a I'm
5:44 not an engineer. And so the worldwide
5:47 web comes out and it makes all of that
5:51 command line stuff as simple as clicking
5:54 a hyperlink.
5:56 And that was the moment where I
5:58 realized, oh [ __ ]
6:00 something is different here, right? this
6:03 the internet is now available
6:07 for the 99% of people who are not coders
6:10 and librarians and engineers and
6:12 mathematicians and academics.
6:16 And so the online magazine I started was
6:18 an art and culture magazine because the
6:20 internet back then all it had on it were
6:22 dissertations and scientific research
6:24 and things like that. And I thought if
6:26 you could put up something with culture
6:28 and art and things like that, it would
6:30 probably get noticed because there's
6:31 nothing on here like that. And so I put
6:35 out this online zen called Urban
6:37 Desires. And I uploaded some learned
6:40 taught myself enough HTML to put
6:42 together the first issue. Uploaded the
6:44 files to some server in California. I
6:46 didn't even know how to do that
6:49 back then. Your computer had to call
6:51 another computer
6:53 literally.
6:57 that thing.
7:00 So, I had my computer call a computer in
7:02 California and hand it some files and
7:05 three weeks later there was a full page
7:08 article about urban desires in a
7:10 Parisian newspaper.
7:14 Three weeks later,
7:16 full page article in French in Paris.
7:22 And in that moment, I knew that the
7:24 world had changed and I knew that almost
7:26 no one else on the planet knew that the
7:28 world had changed. But I got to
7:30 experience it. Geography had collapsed
7:32 and the distance between, you know,
7:35 sending something used to take weeks or
7:37 days now was instant and I knew that was
7:40 going to change everything. This channel
7:45 is about what happens when that exact
7:48 same moment happens for AI and machine
7:51 learning, which is November 30th, 2022.
7:55 Chat GPT launches. Chat GPT does for AI
7:58 and machine learning what the worldwide
8:01 web did for the internet. So, while I'm
8:04 not a programmer, what I'm about is
8:07 how did the 98.5% of us that are not
8:10 coders and engineers, I think it's gone
8:12 up a little since since I was doing
8:14 things back in the 90s,
8:17 98.5% of us who aren't that, how do we
8:20 use this stuff? And if you are an
8:21 engineer, it's it's still completely
8:23 different because it can now code. And
8:27 if you're like, well, it can't code as
8:28 good as me. Well, have you used it
8:30 lately?
8:32 Right? And give it two years. Have you
8:35 have you look at the growth looked at
8:36 the growth of it? So what does that
8:37 mean? And so what this channel is about
8:39 is about getting curious and
8:40 understanding how it works and what the
8:42 implications are and what the future of
8:44 work is and what the future of
8:45 self-expression is because in the end I
8:49 think what our job becomes is not to
8:51 have the tactical skills.
8:54 It it it matters less and less every day
8:57 if I'm good at this programming language
8:59 versus that one. What matters more and
9:01 more is what's my idea
9:04 and do I understand what I want like
9:08 what's my intention
9:11 so so I believe that our job is to is to
9:14 become much more like a producer like a
9:16 Rick Rubin where our job is to hold the
9:20 fidelity of the idea to have the idea I
9:23 want to do this and hold the fidelity of
9:25 that idea
9:27 and then understand how to use enough of
9:29 these AI tools to get that get them to
9:32 do all the leg work to pull off your
9:35 idea and amplify your idea and reimagine
9:38 your idea.
9:40 So that's that's what's going on here.
9:44 It looks like a zen. Exactly. Visionary.
9:50 [Music]
9:55 My dull sister-in-law is utilizing GPT
9:59 on a day daily basis.
10:02 Yeah. Yeah. And like you know if if you
10:04 haven't played if if you're a coder and
10:06 you haven't played with a tool like like
10:09 Replet Agent or Lovable in particular,
10:12 Lovable for me is the is the first of
10:15 the vibe coding apps
10:18 to really start to hit
10:23 a level of usability for non-coders
10:27 that they can still build fairly robust
10:29 apps that have database backends and
10:31 authentication and things like that.
10:32 They don't have e-commerce in there yet,
10:34 but I'm sure that'll come. Um,
10:38 but just like you can speak an image
10:40 into existence with MidJourney and a
10:42 video into existence with Runway, with
10:44 Lovable, you can speak an app into
10:46 existence. And and you know, coders and
10:48 and engineers are poo pooing that,
10:51 saying, "Well, that's not real coding,
10:53 right?" which is the exact argument that
10:56 photographers said when digital cameras
10:58 came out and what painters said when
11:02 when cameras came out when photography
11:04 came out right in the what 1700s 1800s I
11:07 think 1800s
11:10 18 184547
11:13 the dgeray type came out
11:16 [Music]
11:19 that's not real art you just push a
11:21 button literally the same argument as
11:24 what what's going on today with AI. Oh,
11:26 you just push a button.
11:28 >> You can make money with your
11:30 >> That's not work. That's not real
11:32 creative, right?
11:38 Well, a funny thing happened on the way
11:40 to the um demonization of AI
11:45 is that all of the tools to a tool, all
11:49 of the platforms have gotten
11:51 increasingly sophisticated
11:54 and increasingly complex. And if like
11:57 for example, if you want to be competent
11:59 at chat GBT right now,
12:02 chat GBT is not a single tool, right?
12:04 The last time I counted, and they've
12:06 added some since then, the last time I
12:09 counted, chat GPT was 27 distinct tools
12:13 or features that radically altered what
12:16 it did or what it could do. 27 different
12:19 tools.
12:21 Do you know them all? Do do you like
12:24 Right. Do you know what code interpreter
12:26 is? Do you know Do you know um what what
12:30 uh chat GBT agent is? Have you used it?
12:32 Do you use it on a regular basis? Do you
12:34 know what it's good at? What it's not
12:35 good at?
12:37 Like, it's it's absolutely [ __ ]
12:39 insane. Do you use advanced voice on a
12:41 regular basis?
12:43 Did you know that you can now upload
12:45 videos into chat GPT and it can analyze
12:48 them for you? That didn't exist two
12:50 weeks ago.
12:52 And every single tool, whether it's a
12:54 video tool or an image tool like
12:56 MidJourney, um, every single tool just
12:59 keeps adding features and features and
13:01 features and features. So, so we we have
13:05 absolutely crossed the point where any
13:07 single person can keep just can even
13:10 keep track of what's happening. No one
13:13 can. So, one of the things that I
13:15 strongly recommend
13:18 is um Brandon, if you could pull up the
13:21 the the
13:23 uh what's it called? AI salon. The AI
13:26 salon. Yeah. Banner. Um,
13:29 one of the things that I I I
13:33 believe deep in my soul, like something
13:35 that I'm very passionate about. So, in
13:37 1994 when I started Urban Desires and I
13:40 co-founded agency.com, there was a third
13:42 thing that I did. There was a third
13:43 thing that I created. It was called the
13:44 Worldwide Web Artist Consortium.
13:47 Whack. And it was it was this New York
13:49 City-based group that was anyone who was
13:52 trying to figure out what the hell this
13:54 worldwide web thing was. Come twice a
13:57 month to this meeting and we're going to
13:59 talk about it. And it became kind of the
14:02 central hub of people trying to figure
14:04 this [ __ ] out. And so the week chat GPT
14:06 came out, I started the AI salon. And
14:09 the AI salon is a very similar a similar
14:11 sort of ethos like come with your
14:13 curiosity. Try to figure out what this
14:15 stuff's all about, but do it in
14:17 community. And here's why.
14:20 We're we're in an we're in an inventive
14:23 expansive phase of AI right now. So 1994
14:29 to 1997
14:32 or eight
14:35 the worldwide web nobody knew anything
14:38 about it. There were no manuals.
14:40 Everything was changing. They kept
14:41 changing HTML. They kept changing
14:43 browsers. They kept changing how how you
14:46 found stuff on the internet. So there
14:49 was a three or four year period where it
14:50 was just primordial soup. And the only
14:53 people that were doing anything with it
14:55 were figuring out on on their own. And
14:57 so what that community did, what that
15:00 group did was it it allowed us on a on a
15:02 bi-weekly basis to get together and say,
15:04 "What did you figure out?" Oh, what did
15:05 you figure out? Ah, okay. That's cool.
15:07 Let me go back to my shop and do those
15:10 things and I'll come back next time and
15:12 tell you what I figured out. That's what
15:15 the AI salon is and that's what this
15:18 meeting is about. Uh, Tik Tok pin chat
15:21 voice found a folder in my notion today.
15:25 Oh, that's exciting.
15:27 So, advanced voice used to be fairly
15:29 disconnected from your chat history and
15:33 and things like connections and things.
15:35 Do you know what do you know what
15:36 connections are in chat GPT? Have you
15:38 enabled enabled a connection? Do you
15:40 know what MCPs are? Do you know that
15:42 chat GPT supports them? Do you know that
15:44 chat GPT has five different
15:45 personalities? I could actually go show
15:47 all this stuff instead of just rambling
15:49 about it. But it but the point I'm
15:50 making is
15:55 we're in an expansive phase of AI right
15:58 now. And I don't think it's going to
15:59 stabilize probably for another 3 to four
16:02 years, right? We've we we're three years
16:04 in as of November, November 30th, 2025.
16:08 We're three years into the chat GPT era.
16:10 I think we've probably got another
16:12 three, probably more like five where the
16:15 tools just keep expanding and morphing
16:18 and changing and what's possible keeps
16:20 changing and morphing. You can't keep up
16:23 with it, but you also can't ignore it.
16:26 You can't just sit on the sidelines. In
16:28 my opinion, I think anyone today who's
16:31 who's actively
16:34 choosing not to participate with AI is
16:37 doing themselves a massive disservice.
16:40 Massive,
16:42 massive.
16:46 Because from where I look at things,
16:51 AI is not going away. You don't invest
16:53 in invest hundreds of billions of
16:56 dollars, trillions of dollars, right?
16:59 Nvidia just committed a hundred billion
17:01 dollars to to invest into Open AI, which
17:03 OpenAI is then going to hand back to
17:05 them and then they're going to give them
17:07 GPUs.
17:08 It's one of those Silicon Valley kind of
17:10 deals, but it's a hundred billion
17:12 dollars. You don't put a hundred billion
17:14 dollars into something and and have it
17:18 go away. It's not going away. So if it's
17:20 not going away, then we as human beings
17:23 have two choices
17:25 and it's it is literally binary.
17:28 You either deal with it or you don't.
17:32 This channel is about dealing with it.
17:34 What is it? How does it work? What's
17:36 cool? What's not cool? What's dangerous?
17:38 What's exciting?
17:40 And you know what we're finding in here
17:42 is that all the tropes, all of the
17:44 tropes that are like the robots are
17:46 going to kill us and the the just all
17:48 that [ __ ] it's the world's greatest
17:50 plagiarism,
17:54 tend to be
17:57 things that people use as excuses why
18:00 they're not participating with AI. And
18:02 what we find is the people that are
18:04 participating with AI, if they kind of
18:06 strap AI on like a jetpack, it amplifies
18:10 their ideas and it's incredibly
18:11 inspiring and and can make a huge
18:14 difference for people. Um, is it going
18:17 to absolutely decimate jobs? It is,
18:20 but it's going to make whole new
18:22 categories of jobs possible. And then
18:25 your next question is going to be, well,
18:26 what are they, smarty pants? I don't
18:29 know. We literally cannot anticipate
18:32 what's coming.
18:35 Just like the farmers in the 1800s could
18:38 not anticipate
18:40 what would happen when steam engines
18:42 took all the farming jobs.
18:45 80% of the agrarian society lost their
18:49 jobs
18:51 in the ad with the advent of the steam
18:52 engine. 40 40 to 50 year period. 80%.
18:57 Did they all just stop working? No.
19:00 You know what they did? They moved to
19:02 New York City. They moved to cities.
19:05 They moved to Cleveland.
19:07 [Music]
19:09 We get to make it up. We get to make it
19:10 up. We do. We We are in Okay, here's the
19:14 other thing. Like, why am I so
19:16 passionate? Why am I willing to go live
19:18 five nights a [ __ ] week and just like
19:22 very often feel like an abject idiot
19:24 because I don't know what I'm doing
19:26 because they keep changing the [ __ ]
19:27 rules all the time and the tech and the
19:30 interfaces
19:32 because
19:35 we are
19:37 if you're if you're on this call if
19:39 you're paying attention to this if
19:40 you're listening to this
19:43 we are in a rarified
19:46 amazing
19:48 time in history where we're the only
19:51 people who are going to know what it was
19:54 like before AI and know what it was like
19:57 after AI. And if you participate in that
20:01 transition period to a great degree, you
20:04 get to write some of the rules. You get
20:07 to tap into the opportunity. You don't
20:08 just have to be Sam Alman to have a lot
20:10 of opportunity. Right now
20:14 there are uh what's it called? Wind was
20:16 it windsurf? Windsurf was a vibecoded
20:19 app about vibe coding apps that sold for
20:23 $3 billion
20:26 in a year. Like like the opportunity
20:30 right now is [ __ ] insane. YouTube
20:32 question thoughts on AI trying to gain
20:35 control of itself
20:38 um whether for self-preservation etc.
20:42 um like how we saw recently.
20:45 So
20:48 with with with very few exceptions,
20:53 those moments are happening in test
20:57 environments. Now they are happening,
21:00 right? Like like the AI is being
21:03 deceptive and and and trying to do
21:06 self-preservation.
21:07 Um, but those are happening for the most
21:09 part in test environments. If if if
21:11 there's something that has sort of
21:12 escaped the labs, I haven't heard of it
21:14 yet. Um,
21:22 it is absolutely possible
21:26 that the AI does that. I'll give you
21:28 another possibility.
21:32 If we do the alignment right, if the
21:34 labs do the alignment right and say that
21:36 the imperatives of humans, this is
21:38 something that anthropics working on a
21:40 lot, they call it alignment or super
21:41 alignment.
21:44 If the imperatives
21:46 of the AI are in line with the
21:49 imperatives of humans and prioritize
21:52 humanity,
21:54 it is also possible
21:56 that the AI will
22:00 save us from ourselves.
22:02 Right? Because by far the most dangerous
22:06 actor in an equation with AI right now
22:10 and human beings are human beings. We're
22:12 [ __ ] We're we're idiots and
22:15 [ __ ] with egos
22:17 and we you know everyone's like, "Well,
22:20 AI hallucinates." Have you met a human
22:22 being? They [ __ ] lied. They're
22:25 deceptive. They're power hungry. They're
22:27 egotistical. Narcissistic.
22:29 Right.
22:31 Tribalistic.
22:33 We're about ready to take ourselves
22:35 [ __ ] out here. It is possible that
22:38 the big destructive thing that AI does
22:41 is save us for from ourselves. Johnny D
22:44 question. Iterating on drafts in
22:45 perplexity changes overwrites past work
22:48 when editing.
22:51 Iterating on drafts in perplexity. I
22:55 don't know about that. If it does do
22:56 that, that's pretty shitty.
23:01 I can you not go back in your chat
23:05 history and restore earlier versions? I
23:08 think you can, but maybe I'm not
23:11 thinking of the right feature, but I'm
23:14 pretty sure like in in anything I know
23:16 in Claude when you do when you do claude
23:18 artifacts when you do chat GPT canvas,
23:23 you can scroll up in your chat and
23:25 there's there's save points along the
23:27 way. So you might be able to restore
23:28 that stuff. Rick McCauley, we made it.
23:31 What's happening, dude?
23:34 Those are all downstream of fear. AI
23:36 does not have fear. Exactly. Preach.
23:38 Hallelujah. Jim Ross, what's happening,
23:40 man? Excited you're in here.
23:45 Yes. But they just create new drafts
23:47 which are great but different.
23:49 [Music]
23:51 Oh. Oh. Oh, I know what you're doing.
23:54 Okay. Yes. So, okay. So, Johnny D. Okay.
23:58 Yes. So, this is a general thing with
24:00 with AI. Okay. Let me show you
24:02 something.
24:05 Okay. Let's we we'll go do some showand
24:07 tell right now. I'm not going to do it
24:08 in perplexity just because I haven't
24:09 been in perplexity in a while, but LL
24:11 LLM's all do this and it drives me
24:13 [ __ ] mad. I think I understand what
24:15 you're saying now. Okay. So, we're going
24:17 to go we're going to go um
24:20 um let's see. Write me a short story
24:26 that has turns and twists
24:32 and is about a young couple
24:37 um
24:39 trying to figure out if they are more
24:44 than
24:45 wait if they are more than just friends.
24:49 Oh, did I not share? No. Okay, fine.
24:52 Whatever. Understood. More than just
24:54 friends. Um, set in. Let's see.
25:00 Set in a carnival.
25:03 Okay.
25:06 So, whatever. This is going to write
25:08 before the lights come on. So, so it
25:10 that's good. It remembered that I wrote
25:12 a song called Before the Lights Come On.
25:14 Um, which I'll I'll play for you in a
25:16 second. The carnival was still half
25:17 asleep when they slipped through the
25:19 gate. the boy and the girl friends since
25:20 fifth grade. Now old enough to know that
25:22 something was different but too young to
25:24 say what. The air smelled like sawdust
25:26 and syrup and promises that wouldn't be
25:28 wouldn't last till morning. Mia walked
25:30 ahead. Okay. So So we've got her thing
25:32 here. Okay. So now if you want to let's
25:36 see. She hesitated and then climbed in.
25:39 The wheel creaked upward. If I just say
25:44 um
25:48 let's see uh they pass the mirror maze.
25:52 I'm gonna say instead of
25:56 instead of passing
26:03 of passing the mirror maze
26:08 um have them walk by
26:13 the goldfish
26:17 bowl game and the
26:22 carney
26:23 is
26:26 dumping in the goldfish.
26:32 Okay. And so now it's going to rewrite
26:36 this.
26:39 And let's see,
26:42 did it rewrite it? The carnival was
26:44 still half asleep when they slipped
26:45 through the gate. The carnival. See, it
26:48 didn't do it. So a lot of times what
26:50 it'll do is you ask it to change
26:51 something in the middle of this and
26:53 it'll rewrite the whole thing. So a way
26:55 you can get around that is and I don't
26:57 know how this is done in perplexity but
26:59 if I say if I say now add this to a
27:02 canvas add this to a canvas
27:07 in chatgpt a canvas turns
27:11 this chat where I just have to keep
27:13 chatting against the whole thing into an
27:14 editable document.
27:17 Enthropic is red teaming trying to
27:20 create the worst case scenarios. Just
27:23 vibe coding some sites while I listen to
27:25 you. The usual. Um, beautiful,
27:28 beautiful, beautiful. Yep. It's rough,
27:31 Kyle. Um, okay. So,
27:34 now I've got this thing in a canvas in
27:36 chat GPT.
27:39 Slashcanvas would be good here. What do
27:41 you mean slashcanvas would be good? Oh,
27:43 yeah. That's exactly right. That's what
27:44 I'm doing. So, this is a canvas. So now
27:47 in the canvas I can I can click on this
27:49 edit button and this is now editable. So
27:54 I can now come in here and I can just
27:56 take let's see they walked they walked
27:59 on silence stretching between them. So I
28:02 can just highlight this one sentence
28:05 and I can do things to it like I can hit
28:08 a pl a plus here editor explain. Let's
28:12 see highlight this and then I'll ask
28:15 chat GBT. I'm going to say um expand
28:22 this into
28:26 him
28:30 singing her a silly song.
28:37 And so now what it's going to do is it's
28:39 going to replace the that sentence
28:42 but leave the rest of the document the
28:44 same.
28:46 All right.
28:48 Side hustle jumping in from O'Hare
28:50 coming back from the create conference.
28:52 Beautiful. Is she jumping into
28:56 into the uh into StreamYard, Brandon, or
28:58 just she's jumping in in Tik Tok?
29:02 Tik Tok. Cool. Hey, Side Hustle. What's
29:04 happening?
29:06 Kyle was one of my favorite South Park
29:09 characters.
29:11 What's happening, Cal? Hey, Cal.
29:14 They killed Kenny. What's up, Kyle? Um,
29:18 so anyway, so they walked on. So now
29:19 that that sentence has become a big long
29:22 thing. So if if uh
29:27 if you're frustrated with with chat GPT
29:30 completely overwriting [ __ ] this is the
29:32 way to do it. And it's listen, what
29:34 you're describing is one of the problems
29:36 with vibe coding right now is you you
29:39 tell you tell one of the vibe coding
29:41 apps to like fix a button and it
29:44 completely re refactors your code and
29:46 [ __ ] up your app. Like that's just kind
29:48 of where we are right now. It's a
29:49 disaster. Here from Pensacola, Florida,
29:51 new to your channel. Welcome, welcome,
29:53 welcome, Coach Melissa. Coach Melissa,
29:55 what's happening? Welcome. Um, where are
29:58 you? Where are you with the uh with the
30:00 whole AI thing? Like what what got you
30:02 here? Were you just stumbling and you
30:04 were like, you know, God, Alec Baldwin
30:06 looks like [ __ ] Like what was the what
30:09 stopped your scroll? Where where are
30:11 you? What's vibe coding? Okay, cool. So,
30:14 um let's go jump over and do some of
30:16 that.
30:18 Um let me share this.
30:22 So this thing, what you're looking at
30:24 right here,
30:26 this is a a a an add friendly project
30:30 manager that I that I built that that um
30:36 I was writing some I was trying to
30:37 organize some thoughts on uh a
30:40 whiteboard and I thought well wouldn't
30:41 it be cool if I could just move these
30:43 elements around and then I went to
30:46 lovable and two hours later I had built
30:48 this app. So, vibe coding, there are
30:51 these apps like this one called lovable.
30:53 So, I'm going to go there right now.
30:57 And so, it says build something with
30:59 lovable. So, it's it's just like it's
31:01 just like chat GPT, right? You got a
31:03 little prompt there. And then what the
31:04 hell do you put in the prompt? You can
31:06 describe an application and Lovable will
31:09 build it for you. It will write the
31:12 code. And in fact, Lovable just added a
31:15 thing. They they call it lovable cloud
31:18 where normally when you create an
31:20 application that you've got what's
31:21 called the front end and you've got the
31:23 back end and the front end is what
31:25 everyone sees and that's usually the
31:27 simple thing and that's that's the thing
31:29 where people like ah it's easy to make
31:30 an app you just put some buttons and a
31:32 little field right that's the front end
31:35 the back end are things like
31:36 authentication and databases and and and
31:40 web servers and things that actually
31:42 power the front end right so you got
31:44 front-end developers back-end
31:45 developers. What Lovable's done is
31:48 they've basically built that into a
31:50 single thing. So, it does the front end
31:52 for you and then you literally just say
31:54 turn on the lovable cloud and it enables
31:56 all the backend database um access to AI
32:00 features like you can put nanobanana. We
32:02 could we'll we'll make a nano banana app
32:04 right now. And if you're wondering what
32:05 is a nanobanana app? What the hell is
32:07 nanobanana? If you're like Kyle, are you
32:10 having an aneurysm? A a nano bananurism.
32:14 What? What the hell are you talking
32:15 about? Nano Banana is the new image
32:19 generation model from Google, from
32:21 Gemini. That's really amazing. Uh,
32:23 Coach, oh, I just missed that pin. What
32:25 was what did that pin say from coach?
32:27 I've been using AI for about a year. Um,
32:29 and in am in an AI course right now.
32:32 Awesome. Welcome, welcome, welcome.
32:35 So, here's the here's the philosophy of
32:37 this channel. The philosophy of this
32:39 channel is
32:42 just be in the AI conversation. Like Jim
32:45 Ross who's here, that Jim Ross said,
32:47 "I'll be vibe coding while you're waxing
32:49 poetic up there." Um,
32:52 Jim is is one of what we call the
32:54 irregulars. So, the irregulars are a
32:57 group of weirdos. They're not weird.
32:59 They're irregular. But they're a group
33:01 of people, human beings
33:04 that come to this channel every single
33:07 night.
33:09 It's nuts.
33:11 But here's why they do it. They they may
33:14 come a little bit for me, but what this
33:16 is really about is about this community
33:19 connecting with one another, supporting
33:21 one another, learning from one another,
33:24 and just being in the conversation. So,
33:26 a lot of times someone will come in and
33:27 they'll ask questions like someone just
33:29 said, "What's vibe coding?" And so, now
33:31 we're going to run down a vibe coding
33:32 rabbit hole and I'm going to actually
33:33 build you an application before your
33:35 eyes and it should melt your face.
33:39 And and what this channel's about is no
33:41 matter where anyone is in the process of
33:44 figuring out this AI stuff, just come in
33:47 here and be in the conversation. And
33:48 sometimes the conversation's going to be
33:50 way over your head. And sometimes the
33:52 conversation's going to be, you know,
33:55 you're going to be like, I know more
33:56 than he does, and that's part of the
33:59 part of the charm, right? And what it's
34:02 about is just we're we're all learning
34:04 this right now. None of us know what it
34:06 means. Nobody really knows more than
34:09 anyone else right now because the
34:11 technology is changing so fast. We
34:13 literally can't.
34:16 like the way you made an AI movie a year
34:19 ago and what was possible a year ago
34:22 versus how you make one today and what's
34:24 possible today. It's it's just there
34:27 it's just it's a completely different
34:28 set of tools. It's a completely
34:30 different set of possibilities. So,
34:32 welcome to anyone who's new and just
34:34 jump in, sit back and if you're if
34:37 you're feeling feel free to ask
34:38 questions whether you're on TikTok or
34:40 YouTube. What's what's amazing about
34:43 this community about the irregulars?
34:45 They're incredibly loving. They're
34:47 incredibly generous. And so if you have
34:49 questions, and I don't get to them very
34:52 often, they'll answer your questions and
34:53 they'll say, "Hey, come join the AI
34:55 salon." And you know, you can ask
34:57 questions over there and I'll help you
34:58 out over there. So if you haven't done
35:00 that yet, you should do that. Join the
35:01 AI salon. Go to go to the salon.ai and
35:04 say join our community. And uh and you
35:08 you will see it's a it's a really
35:09 amazing place. Okay.
35:11 Gravity jump. My back end hurts. I can
35:14 never get it to work. Okay. Vibe coded
35:17 two apps last week. Oh, that's great.
35:19 Uh, Tai B, uh, a genealogy app and a
35:22 martial arts career seeker app. That's
35:24 beautiful. It's amazing. Yeah, it's it's
35:26 incredible. Okay, so what are we going
35:29 to make tonight? Let's see. Let's do a
35:34 um
35:38 Oh, I got one. Um, I want to create
35:47 a
35:50 fake
35:52 um,
35:54 DNA animal designer
35:59 using nano banana
36:03 to generate images.
36:08 The user
36:11 um picks
36:14 up to three animals
36:18 to combine
36:21 in a visual interface.
36:27 Interface.
36:30 Um what were those cards called? Do
36:32 y'all remember those cards where you had
36:34 like the head of an animal, the middle
36:35 of an animal, and the legs of an animal?
36:38 Anybody remember what those were called?
36:44 Come on. Some of you Gen Xers in here,
36:47 you got to have a couple of brain cells
36:48 left cuz I don't.
36:52 LSD. No. No, it's not LSD.
36:55 [Music]
37:00 Silver Fox. Say it again. The entire
37:03 animal. No, it was it was a card. It was
37:06 a It was like they were flip books and
37:09 it had three three panels and it there
37:11 was like the head panel, the middle of
37:13 the animal, and then the feet. And you'd
37:15 flip it. It'd be like the head of a
37:17 flamingo, the body of an elephant, and
37:19 the legs of a chicken. Come on. I know
37:22 what you're talking about, but I don't
37:23 know the name. Anyway, all right. A
37:25 visual interface.
37:27 The 60s were good to you.
37:31 I actually remember those cards though.
37:33 Okay. A visual interface like the ones
37:37 that had a head, body, and feet
37:44 cards
37:46 you would
37:48 mix up. But this is the
37:52 modern
37:55 digital version.
37:58 It should look like a lab experiment
38:05 experiment
38:08 where you select the three animals
38:14 and what percentage of each
38:20 and then nano
38:23 banana
38:26 makes
38:27 a photo
38:30 of your new animal. That's cool, right?
38:34 Okay.
38:36 So,
38:39 we're effectively done.
38:42 So, what vibe coding is, uh, so who was
38:45 it? Who was it earlier on that that when
38:47 they first got in here asked
38:51 not and it was Young Young Bungalow? I
38:54 don't know if Young Bungalow is still
38:55 here. Young Bungalow came in and when
38:57 when he first got in, he said, "What's
38:58 your favorite or what lang programming
39:00 language are you most proficient at?"
39:02 And I said, "English." And so, um, this
39:06 is why because I can now just
39:10 bumble bumble out a description of an
39:14 app and this thing will figure it out.
39:15 So, look what it says. Okay. I'll create
39:18 a DNA animal designer with a lab
39:20 experiment aesthetic using nanobanana
39:22 image generation. This will have a
39:24 visual card interface where you mix up
39:26 three animals with percentage controls.
39:30 Design vision laboratory science theme.
39:32 Visual animal selection cards. Um
39:36 percentage sliders for DNA mixing. Clean
39:38 futuristic interface features. Select up
39:41 to three animals. Right. And then so now
39:43 enable cloud. So I need to enable the
39:45 cloud so that it can um
39:49 use AI features from Gemini
39:52 flipflap books or miximals or exquisite
39:55 corpse books. Cool. Beautiful. I like
39:58 the percentage feature. Thank you. Thank
40:00 you. Um still asking do you follow
40:05 follow Peter Diamont Diamontis? Um, I I
40:09 don't follow him like I would say if
40:11 there's people I follow right now, it's
40:13 Ethan Malik. Um, Paul Ritzer from the
40:15 Marketing AI Institute.
40:18 Um, Robert Scoville I pay attention to.
40:21 Um, Peter, I'll I'll if if he shows up
40:23 in my feed, I'll pay attention to what
40:24 he's talking about, but I don't
40:26 necessarily follow him. Um,
40:30 okay. The game you're likely thinking of
40:32 is often referred to as Animal Rummy.
40:35 No, it wasn't Animal Rummy.
40:37 No,
40:39 it wasn't Animal Room. It's It's
40:41 flip-flop books, I think, is is what it
40:43 was. Mad Libs has the word version.
40:46 Picture picture consequence, maybe. All
40:49 right. Okay. So, and and you're like,
40:51 Kyle, why are you talking? What's going
40:53 on? Why is this not done? Because it's
40:55 coding. It's writing code right now. Um,
40:58 you used to be able to watch it write
40:59 code, but they don't let you do that
41:02 anymore because it's I think they've got
41:03 an interface bug that they still haven't
41:05 fixed.
41:06 Um, but as soon as this writes all the
41:09 code, we're going to be able to play
41:10 with this app or at least look at it.
41:12 And then
41:14 what vibe coding is, actually, while
41:16 that's building our app, let's go
41:21 find the original
41:23 definition of vibe coding. Vibe
41:26 coding Karpathy.
41:29 So, Andre Karpathy, who's one of the
41:31 founders of OpenAI,
41:34 um, on February 2nd of this year,
41:39 wrote this. There's a new kind of coding
41:42 I call vibe coding where you fully give
41:44 into the vibes, embrace exponentials,
41:46 and forget that the code even exists.
41:50 So, if you're not a coder,
41:53 this makes sense to you, right? Because
41:56 you're just like, I just want an app.
41:57 like you don't really give a [ __ ] what
41:59 the code's doing if you're a coder,
42:01 which Andre Karpathy is. He's probably
42:03 one of the better coders, you know, on
42:05 the planet. Um,
42:08 this is like blasphemy, right? So, so
42:11 you know, there's a new kind of coding
42:13 where you give in fully to the vibes,
42:15 embrace exponentials, and forget that
42:16 the code even exists. It's possible
42:18 because LLMs, eg cursor composer with
42:21 sonnet, are getting too good. I just
42:24 talked to composer with super whisper.
42:26 So, so speech to text. So, I barely
42:29 touch the keyboard. I ask for the
42:30 dumbest things like decrease the padding
42:33 on the sidebar by half because I'm too
42:35 lazy to find it. I accept all always.
42:38 So, so when it makes changes and it
42:41 says, do you want me do do you want to
42:42 accept these changes? He just says
42:43 accept all. Um, I don't read the diffs
42:47 anymore. So, diffs are like what
42:49 changed? Here's the old code. Here's the
42:50 new code.
42:51 And engineers will read those
42:53 differences to understand what it what
42:55 changes it made. When I get error
42:58 messages, I just copy and paste them
42:59 with no comment. Oh, am I not sharing
43:01 this? I'm not. Um, when I just uh when I
43:06 when I get error messages, I just copy
43:08 and paste them with no comment. That
43:10 usually fixes it. The code grows beyond
43:12 my usual comprehension. I'd have to
43:14 really read through it for a while.
43:16 Sometimes LLMs can't fix a bug, so I
43:18 just work around it or ask for random
43:20 changes until it goes away. It's not too
43:23 bad for throwaway weekend projects, but
43:25 but still quite amusing. I'm building a
43:28 project or web app, but it's it's not
43:30 really coding. I just see stuff, say
43:32 stuff, run stuff, copy and paste stuff,
43:35 and it mostly works.
43:38 So that's the definition of it. Okay. So
43:43 we now have Dr. J exquisite
43:48 corpse books. Those were the the
43:49 animals. Cool. Very cool. Can you
43:52 explain the token system with AI,
43:53 please? Yeah, I absolutely can. Give me
43:56 one second. Okay. DNA slot one, DNA slot
44:00 two, DNA slot three. So, we're going to
44:02 put Oh, these need to be drag and
44:05 dropped, but that's okay. We're going to
44:07 do Oh, this is cool. So, so we've got
44:10 we've got So, this app, it looks like
44:12 it's going to be functional right out of
44:14 the gate. So, we're going to do a dog, a
44:15 dragon,
44:17 and
44:20 [Music]
44:23 a shark.
44:25 Okay. So, so it's done the math for us.
44:29 So, notice that it's done the math. Oh,
44:31 yeah. Is there a turtle in there?
44:32 There's not. So, one of the things I
44:35 need to do is I need to be able to say,
44:37 um, we need to be able to add our own
44:40 animals. Enable nano banana. I think
44:42 that's already done. Okay. So, let's
44:46 let's make this thing
44:51 like 50% shark,
44:57 44% dragon.
45:01 Oh, it doesn't balance them equally. All
45:03 right.
45:07 All right. So, we're going to go
45:10 45 dragon, 41 shark, 15 dog. Generate
45:14 hybrid creature.
45:17 Now, now this should
45:20 this should
45:22 So, already already. Okay, here's our
45:25 hybrid creature. Look at that. Are you
45:29 [ __ ] kidding me?
45:32 at the water. Shark teeth, dog body,
45:35 dragon tail, dragon fins, things to the
45:39 percentages.
45:41 One prompt.
45:43 David Simon, what's happening?
45:46 One prompt.
45:49 And now we can we can go fix this,
45:50 right? You could go say, I don't want I
45:52 don't want to use emojis for the
45:54 animals. I want, you know, more animals.
45:56 I want to be able to add my animals. I
45:58 want people to be able to create
46:00 accounts.
46:03 That's [ __ ] sick. Lion. Wait, how do
46:06 I How do I restart? I Oh, I can't
46:08 restart. We got to be able to restart,
46:11 right? A hybrid creature.
46:15 I could add sharing in here. I want to
46:17 add social media sharing so I can share
46:19 that that thing, you know, out to the
46:22 world.
46:24 Come on, man. You kidding me?
46:28 Okay, let's do one that's like 70% dog.
46:34 We'll we'll we're going to lower the
46:35 shark. We're going to lower shark to
46:37 six. Oh, wait. I can get rid of one of
46:39 these. Ah, okay. Let's add in eagle.
46:44 Eagle's good. All right, we'll go 33 33
46:48 33. This is dragon, eagle, and dog.
46:53 [Laughter]
47:00 [Music]
47:05 How cool is that? Come on, people. Look
47:09 at that.
47:12 You kidding me?
47:20 We should probably mention credit limits
47:23 for storage.
47:25 $1 AI out of the box
47:28 on the free plan.
47:31 I don't Can you write that in English? I
47:34 don't quite understand what you're
47:35 saying, Brandon.
47:40 That is a cool dragon dog bird.
47:43 Every generation costs Oh, every
47:45 generation costs credits. Yeah, yeah,
47:47 yeah.
47:48 And Lovable gives you a dollar to start.
47:51 Okay. Yeah. So, you can do this in
47:54 Gemini AI Studio.
47:57 Um, but anyway, so so that's vibe coding
48:00 and you can vibe code video games. You
48:03 could vibe code
48:05 like a a I I don't know an app for
48:09 managing your soft softball team like
48:12 like let people log in and and let you
48:14 know their availability and and it'll
48:16 you know create schedules for you like
48:19 literally any app idea that you have you
48:23 can just make it here. Now Lovable is
48:24 web apps only right? So these are apps
48:26 that are going to live on the web. If
48:28 you want to make phone apps there's
48:29 different vibe coding tools for those.
48:31 that gets more complicated. But lovable
48:34 if you just want to if you've never
48:35 coded before in your life and you want
48:38 to go from a standing start, I've never
48:41 made an app before to what I just did
48:44 here. Right? One prompt sort of poorly
48:49 articulated.
48:50 Boom. Functional app within I don't know
48:53 what was that 10 minutes.
48:56 Imagine spending four hours with this
48:57 refining it and doing all that sort of
48:59 stuff. All right. So, if someone start
49:02 uses the app that you create, you pay
49:04 for every generation. Yes, you do, Rick
49:06 McCaulay. So, one of the things that
49:08 Lovable does not have in um
49:15 in the cloud right now is the ability to
49:18 to put a payw wall in front of it. So,
49:20 like lots of apps will will let you do a
49:23 generation or two and then they're like,
49:24 "Give us five bucks a month to keep
49:25 going or you know, buy 10 bucks to buy
49:28 credits." Um, you could certainly do
49:30 that with this. Um, Brandon last week, I
49:33 was whining about that and Brandon last
49:35 week spun up a a he basically just said
49:40 to Lovable, "I want to be able to charge
49:42 for this and tied it to his what' you
49:44 tie it to, Stripe or or uh Venmo?" Yeah,
49:48 he just tied it to his Venmo account.
49:50 So, you can have Lovable do that stuff.
49:52 It's just not built into it right now.
49:54 So, um, Kyle, Sora just turned your
49:59 comment into a cool app idea. Posted in
50:02 irregulars. Sora just turned your
50:04 comment
50:06 into a cool app idea. The hell does that
50:08 mean? Where'd you put it? Irregulars.
50:25 I made an app with Lovable that
50:27 transcribes Tik Tok videos. Oh, cool.
50:35 Oh my god. Tik Tok's interface for when
50:39 I try to expand a comment.
50:43 Check the chat.
50:50 I do not see it. I do not know of what
50:53 you expect.
50:58 Who's that? Suzanne.
51:02 Suzanne speaking at the create
51:03 conference. Very cool.
51:09 Oh, look. Here's Corey Sandler's images
51:12 for the uh for the song I made. Very
51:15 cool.
51:17 It's in share cool things irregulars
51:20 now.
51:22 Share cool things.
51:30 [Music]
51:32 Where is share cool things?
51:40 Oh, just refresh.
51:46 No,
51:50 I do not know what you specu
52:04 [Music]
52:07 sneaking into class late. Oh, Mary Mary,
52:11 thank you.
52:22 Kyle said his f vibecoded app would melt
52:25 your face off.
52:38 Wait, hang on. Let me show this tab.
52:43 app that I call Melt Your Face Off
52:44 because when you see it, it kind of
52:45 does. It reads your mood, spins up a
52:47 generative filter, and turns you into
52:48 living liquid color. I'm in the AI
52:50 learning lab and I just finished
52:51 something I'm pretty excited about. It's
52:52 a vibe coded app that I call Melt Your
52:53 Face Off because when you see it, it
52:55 kind of does. It reads your mood, spins
52:56 up a generative filter, and turns you
52:58 into living
52:58 >> That's hilarious. Uh, okay. I thought
53:01 you meant Sora. The OpenAI team that
53:04 built Sora did something with my app.
53:07 No, you made a Sora video of me vibe
53:09 coding an app. Got it now. Good lord.
53:18 Miro encroaching on Tik Tok cam. Oh. Oh
53:21 my.
53:23 Okay. Let's see what's going on. Oh,
53:26 there we go. Okay. All right.
53:31 Oh, yeah. They're they're they're
53:32 building lovable in Demiro right now.
53:35 All right. What was I going to even talk
53:37 about tonight? We got off on tangents,
53:38 which is super cool. Like one of the
53:40 things I love on this channel is if
53:42 people come with ideas, thoughts,
53:44 questions, things like that. Um,
53:47 what do you want to play with? What do
53:49 you want to talk about?
53:52 Um, I'm happy to I'm happy to dig in and
53:55 show some stuff. The Melt My Face app.
53:59 You're going to talk about black bars.
54:01 Shut up,
54:04 Pin. Let's do a code of arms family
54:06 crest.
54:10 That's pretty good. Oh, tokens, Coach
54:12 Melissa. Okay, tokens. All right, so
54:14 you're going to get the box metaphor. I
54:16 haven't done the box metaphor in a long,
54:18 long time.
54:20 Okay,
54:22 [Music]
54:32 where we'll use playing card boxes.
54:39 and a cassette tape.
54:42 Okay, we'll come back to those. Wait,
54:45 we'll use a beat speaker.
54:49 Yeah, there you go. Okay. Okay,
54:54 tokens. Everyone's talking about tokens.
54:57 In fact, the OpenAI developer
54:59 conference, they just gave out awards to
55:02 developers who have processed a 100red
55:06 billion tokens or a trillion tokens and
55:09 they got plaques. You've processed a
55:12 trillion tokens. What What is this token
55:15 madness?
55:17 Okay,
55:19 I'm going to go back a little bit in
55:21 history. I won't be there that long. In
55:23 2017, Google,
55:26 some engineers at Google write this
55:29 paper called Attention Is All You Need,
55:31 and they invent. It's not the same
55:34 without the printer inking bottles. Oh,
55:35 here are the printer inking bottles.
55:36 Fine. Okay. I got I got two. We're good.
55:39 We're fine. Okay. I couldn't find them
55:41 before. I've I've started to clean up my
55:43 desk area. And you know how it is when
55:45 you have ADD and you try to clean
55:47 something and then you can't find
55:48 anything because you knew where the
55:49 piles were. Okay, that's not what we're
55:52 talking about. Tokens.
55:54 2017, Google invents this thing called
55:57 the transformer.
55:59 And the transformer is an architecture
56:02 that is the engine behind almost all of
56:07 the generative AI things we play with
56:09 today. All the large language models,
56:11 all the image models, the the video
56:14 models, all of them are based on this
56:16 transformer architecture. There's some
56:18 other um technologies layered on top for
56:21 images and videos called diffusion
56:23 models and and there's some different
56:25 things in there but but the underlying
56:27 model is is this transformer model.
56:32 What happens is they take a bunch of
56:35 data. So, so let's just say they take
56:37 documents and they take documents like
56:39 here's a book and they describe what the
56:41 book is and they put the the book into
56:44 this model and then they take a blog
56:45 post and they they label that and say
56:48 this is a blog post and maybe here's
56:49 what the topics are and there's all
56:51 sorts of fancy ways to do this but
56:53 basically they go out and they get as
56:55 much data as they can and in fact one of
57:00 the one of the reasons AI has such a bad
57:03 reputation right now is they went out
57:05 and they got all the information. Well,
57:08 what's included in all the information
57:10 is a bunch of copywritten stuff. Okay.
57:13 So, so they absolutely trained these
57:17 models on copywritten stuff. They have
57:19 done that. But when you understand what
57:21 actually happens to them and the court
57:23 actually agrees with this. The court
57:25 calls when you embed
57:28 um something like a book through the
57:30 transformer model. It is a
57:32 transformative act and because of that
57:34 they consider it fair use. But you can't
57:37 steal the book to do that. You have to
57:39 buy the book to do that. But anyway,
57:42 that's neither here nor there.
57:47 Okay. What happens when they take a
57:50 book? We'll just we'll just stick with a
57:52 single book. They take a single book and
57:55 the transformer model fragments, you
57:59 know, splinters, shards
58:03 all of the contents of that book into
58:05 what are called tokens. So tokens are
58:07 either words, periods, spaces, commas,
58:11 um parts of words. So you might have
58:13 running. So you might have run and ing
58:16 running or run and ning two ends. Um
58:20 those might be two tokens that make up
58:22 one word. So in general
58:25 when you have a thousand tokens you it
58:28 equates to about 750 words. So when you
58:32 hear people talk about a large language
58:35 model having um a c certain token
58:38 context window and they say oh this has
58:41 got a 100,000
58:43 context uh token context window that's
58:46 about 75,000 words which basically means
58:49 in its in its memory it can basically
58:52 keep a little more than a a complete
58:55 novel in its thought before it starts
58:57 forgetting things. So, so tokens are
59:01 fragments of words. But, but what's the
59:02 box?
59:04 The box is
59:06 they take all of the documents that they
59:09 can gather, like all of the documents
59:11 that essentially represent the output of
59:13 humanity for all of our history that's
59:15 been stuck on the internet for the past
59:17 80 years and then probably more than
59:19 that. And they run it through this
59:22 machine. And what it does is it takes
59:24 the original document and it shatters it
59:26 into all these tokens and then these
59:29 tokens are clustered in semantic in
59:34 thousand dimensional I know this sounds
59:36 like I'm making this up and it's
59:38 [ __ ] It's not [ __ ] This how it
59:39 works. It's thousanddimensional
59:42 mathematical space where the tokens are
59:44 grouped in semantic clouds. And and
59:47 here's what a semantic cloud means.
59:50 Semantic cloud is you could have the
59:52 word dog, right? The word dog is the
59:54 word dog. Dog g. And you could say, "Oh,
59:56 well, you're just going to have that's
59:58 going to be in there once." No, it's not
59:59 going to be in there once because dog in
1:00:02 one context is the fluffy thing that you
1:00:05 pet when you get home at night. And dog
1:00:07 in another context is he did his
1:00:09 girlfriend wrong and he's a dog. So
1:00:12 those two tokens of dog actually get put
1:00:15 in different semantic clouds. So you
1:00:17 have this massively complicated
1:00:20 mathematical space that has all these
1:00:22 tokens in it. And the way the
1:00:24 transformer model works, this is
1:00:25 absolutely mindblowing that somebody
1:00:28 figured this out. But because they've
1:00:30 figured it out, we get to benefit from
1:00:32 it. When you type in a prompt into chat
1:00:35 GPT, what it's actually doing is it's
1:00:38 generating a probability.
1:00:40 It's generating a probability of a
1:00:43 single token in what they call the
1:00:45 latent space. This thousand-dimensional
1:00:48 mathematical semantic cloud. What you
1:00:52 type in creates a probability that some
1:00:54 token in that massive information cloud
1:00:57 is the most likely next probable token
1:01:02 to answer your question and then the
1:01:04 next one and then the next one and then
1:01:06 the next one. So when you watch chat GPT
1:01:08 write you a whole document or you watch
1:01:10 lovable code for you,
1:01:13 it's all just pulling these mathematical
1:01:16 tokens out of this mathematical space.
1:01:18 So so you have basically all of human
1:01:21 knowledge gets jammed into a box
1:01:25 and this is what's called a large
1:01:29 language model, right? And so you have
1:01:31 like GPT5
1:01:34 is is one of these or GPT5 is actually
1:01:37 multiple of these that are kind of glued
1:01:38 together.
1:01:40 And then chat GPT is actually an app
1:01:43 that sits on top of the large language
1:01:46 model. And this would be like you know
1:01:48 like like um uh Microsoft Copilot,
1:01:53 right? So Microsoft Copilot uses the
1:01:56 same model GPT5 as chat GPT but these
1:02:00 are completely different experiences
1:02:01 right this one's bloated and unusable
1:02:04 and chat GPT is actually good right
1:02:06 because this application is better both
1:02:09 of them when you type a prompt in here
1:02:12 through an API it calls down into this
1:02:14 giant mathematical space and says give
1:02:17 me an answer back and so what it
1:02:19 displays on screen is coming out of this
1:02:21 what's called a corpus of all of the
1:02:24 knowledge of humanity or as much as they
1:02:26 could get in it. Um, and that's the B
1:02:29 box metaphor and that's what tokens are.
1:02:31 It's it's absolutely insane. So, the
1:02:33 here's what's here's why I think it's
1:02:36 important if you're if you're going to
1:02:38 play in the AI game and especially if
1:02:40 you want to get good at this and be a be
1:02:42 a professional about this.
1:02:45 When someone says chat GPT is the
1:02:46 world's greatest plagiarism machine,
1:02:51 it's not
1:02:53 because it's not the way the thing
1:02:55 works. What what people are assuming who
1:02:57 don't don't understand how AI works is
1:03:00 that they're assuming that that chat GPT
1:03:03 is somehow reading a document and
1:03:06 copying like whole paragraphs out of it
1:03:08 and and you know just copy and pasting,
1:03:11 right? That would be plagiarism. It's
1:03:13 not doing that. The G in GPT, GPT stands
1:03:17 for generative pre-trained transformer.
1:03:20 So the T for transformer, that's the
1:03:22 technology. Pre-trained means they went
1:03:24 out and they found all the data for us.
1:03:26 We don't have to bring our own data like
1:03:27 you used to have to in the olden timey
1:03:30 AI days like four years ago. The G
1:03:33 stands for generative.
1:03:36 It's generating new content every time.
1:03:39 So actually there was a comment earlier
1:03:40 about every time I make a change it
1:03:42 completely rewrites my thing. That's
1:03:44 because of the nature of how the
1:03:46 transformer works. It it's generating it
1:03:48 new each time. Now you can prompt it
1:03:51 very specifically to generate something
1:03:54 that looks just like the original,
1:03:57 right? And that's where people can
1:03:59 prove, oh, this was trained on
1:04:02 copywritten works, right? But if you
1:04:05 look at what's actually happening, it is
1:04:06 a transformative act that shatters the
1:04:09 original documents into these tokens.
1:04:11 Cool. Make sense?
1:04:15 It's absolutely [ __ ] insane. It's
1:04:17 like the more you learn about this, the
1:04:19 more I learn about this, the more blown
1:04:21 away I am that it does what it does.
1:04:24 It's crazy. It's crazy
1:04:28 because
1:04:34 In a weird way, it's not actually
1:04:36 understanding what it's doing, but it's
1:04:39 behaving
1:04:42 incredibly like it's understanding what
1:04:45 it's doing, right? It's it's behaving a
1:04:48 lot more like a human than people want
1:04:51 to give it credit for. But it really is
1:04:54 just math. It really is just a
1:04:55 probability calculator. But but a lot of
1:04:58 people will say, "Oh, it's just a
1:04:59 stochcastic parrot. it's just a
1:05:01 probability calculator. It can't have
1:05:03 feelings. It can't it can't be creative.
1:05:05 It can't be this. It can't be that. But
1:05:08 none of that's actually true. Like it it
1:05:11 can actually be quite creative. And it
1:05:12 can be quite creative in a way very
1:05:15 similar to how our brain works because
1:05:17 neural networks, the hardware
1:05:20 underpinning of all this stuff, were
1:05:22 designed to mimic how our brains work.
1:05:25 So anyway, Tik Tok pin or it's just
1:05:27 consciousness twice.
1:05:30 Yeah. I don't know. I mean, the whole
1:05:31 thing about is it sentient, is it
1:05:33 conscious, that that whole thing for me,
1:05:39 that's a semantic argument. It's an
1:05:42 academic argument.
1:05:44 If we as humans interact with this thing
1:05:48 like it's sentient, then it is.
1:05:51 Then it is. Whether or not it actually
1:05:54 is to me is a is an academic definition.
1:05:59 But but for people who are not digging
1:06:01 into the geeky side of AI and they're
1:06:03 just talking with with chat GPT advanced
1:06:07 voice
1:06:09 like an old person, you know, alone at
1:06:12 home where their kid gives them access
1:06:15 to chat GPT advanced voice and they can
1:06:18 have a conversation all afternoon.
1:06:21 To them, that's that's every bit a
1:06:25 relationship as having a person sitting
1:06:27 there, right? So for them, they're
1:06:30 talking to someone and they probably
1:06:32 have a relationship with that person. In
1:06:34 fact, when chat GPT went from GPT40 to
1:06:38 GPT5, a bunch of people got their
1:06:40 panties in a twist because they're like,
1:06:43 "You took away my girlfriend. You took
1:06:45 away my boyfriend. Their personality's
1:06:48 gone. Give me back my boyfriend."
1:06:54 My CPU is a neural net processor, a
1:06:56 learning computer governor Arnold
1:06:59 Schwarzenegger.
1:07:01 My computer, my GPU is a neural network.
1:07:07 And then there's emergent properties.
1:07:08 Rick McCaulay just said that. So
1:07:11 emergent properties, these these this is
1:07:13 an amazing thing. So
1:07:19 when you embed all of this information,
1:07:21 so so basically you categorize a bunch
1:07:24 of information. So So you have human
1:07:25 beings that like watch movies and read
1:07:28 books and read papers and read blog
1:07:30 posts and they they label them and
1:07:31 categorize them and they basically do
1:07:34 all that stuff to be able to embed them
1:07:36 correctly. And so they train these
1:07:39 models on certain things. One of the
1:07:41 things they they never trained models on
1:07:44 was to be able to do translations.
1:07:47 Um, but they just embedded them with
1:07:50 enough data that that they just
1:07:52 realized, oh, these things are actually
1:07:53 really good at translations. And there's
1:07:55 those are called emergent behaviors
1:07:56 where it wasn't specifically trained to
1:07:58 do something, but it's just good at it.
1:08:01 Um, computer program, excuse me,
1:08:04 programming was was one where they
1:08:06 trained it on programming and it got
1:08:08 good at linguistic things because
1:08:10 language is structural and code is
1:08:13 structural and so it could sort of
1:08:15 figure out, you know, um, things that
1:08:18 they didn't think it would be able to
1:08:19 figure out. It's absolutely crazy. All
1:08:22 right.
1:08:24 I had a message come by by on chat GPT
1:08:27 that I was online for a long time and I
1:08:29 should take a break. Well, so so chat
1:08:32 GBT was probably right. Rod
1:08:39 on deck AI studio refresh. Okay, cool.
1:08:42 Um,
1:08:45 I want to show you something. Um,
1:08:49 I don't have really anything to show you
1:08:50 in comparison to it, but let me let me
1:08:52 run over and
1:08:54 and show you something here.
1:08:57 So, this is
1:09:01 Google's playground or Google's they
1:09:04 call it AI studio. So if you go to AI
1:09:06 studio.google.com
1:09:08 they've done a completely new interface
1:09:10 here. Um so you can chat with models in
1:09:14 the playground. So you can chat with AI
1:09:16 models. You can vibe code in in the AI
1:09:21 studio and it's actually really good.
1:09:23 The the vibe coding here is really good.
1:09:25 You can look at monitor usage of your
1:09:27 dashboard. This is a
1:09:30 what this site is aistudio.google.com.
1:09:33 This is for developers.
1:09:36 Um, so if you open up the little side
1:09:40 panel here, um, they've got like
1:09:42 documentation. You can get an API key.
1:09:46 Um, and and they've got lots of choices.
1:09:48 This site used to be very very
1:09:53 cluttered and technical and they didn't
1:09:56 explain anything.
1:09:58 and it was up to you to figure out which
1:10:00 model had which capabilities in it and
1:10:02 you had to read documentation and things
1:10:04 like that. And so the fact that that it
1:10:07 is now
1:10:09 kind of this visual front end, I'm sure
1:10:12 that engineers this probably drives them
1:10:15 [ __ ] baddy. They're like, we don't
1:10:17 need any stupid pictures. They're used
1:10:19 to command line interfaces, right? Um
1:10:25 this is not for them. What what's
1:10:27 happening is this.
1:10:29 As the AI models are getting more and
1:10:31 more and more capable, people that were
1:10:34 historically not coders are now coding.
1:10:38 So what's happening is people are
1:10:39 starting out vibe coding and then
1:10:41 they're realizing, oh [ __ ] I got to set
1:10:42 up authentication or I need to set up a
1:10:44 database or I want to do, you know, some
1:10:46 sort of commerce backend or I want to
1:10:48 put this on a different website. And so
1:10:50 they get chatpt to start teaching them
1:10:52 things or Gemini to start teaching them
1:10:54 things and they're like okay so if you
1:10:56 want to do that you got to go in here
1:10:57 and then you got to put your code in a
1:10:59 GitHub repository and then you got to
1:11:00 push from there and you got to do a pull
1:11:03 request.
1:11:05 Right? So you're starting to get people
1:11:08 who were not trained as engineers
1:11:12 entering engineering spaces. And so what
1:11:15 this represents to me is this is Google
1:11:17 recognizing that they've got audience in
1:11:20 here that's not just engineers.
1:11:23 That mashing
1:11:25 of expertise is going to happen in every
1:11:29 single industry.
1:11:33 Every single industry. It's all going to
1:11:36 get mashed up.
1:11:38 What it means for us as individuals is
1:11:40 that
1:11:43 where we were weak, where we've where we
1:11:45 have gaps,
1:11:48 like for me, I've got ADD. One of the
1:11:50 reasons I can't code is I don't give a
1:11:53 [ __ ] I like I love having the idea. The
1:11:56 minute I have to turn the idea into,
1:11:58 okay, now you're going to have to put
1:11:59 this in a sequence of things in a
1:12:02 framework with I can understand the
1:12:04 frameworks, but the minute I actually
1:12:06 have to go in there and look at every
1:12:07 [ __ ] line of 27,000 lines of code, I
1:12:10 would rather jump through a window,
1:12:12 right?
1:12:14 So, my brain just doesn't work that way.
1:12:16 But now, because of AI, I can take that
1:12:19 gap in my skills and in just even the
1:12:23 way my brain works and I can fill that
1:12:24 in. So now I can just take this idea I
1:12:27 had for an app, put it in one of these
1:12:29 vibe coding things. It fills in that gap
1:12:31 for me. And now I can look in an app
1:12:32 that I built.
1:12:35 And engineers will tell you, "You didn't
1:12:37 build that. You just prompted that."
1:12:41 Yeah, I did. But that was my app idea.
1:12:44 And I'm looking at it now. And I'm now
1:12:46 looking at it going, "Is that as good as
1:12:48 I want it to be?" No. I want this button
1:12:52 over here, and I want that over there.
1:12:53 And I want the label to be different. I
1:12:55 want the colors to be different. And so
1:12:57 I go vibe code that and it does that.
1:12:59 And then it does all that. And I start
1:13:01 playing with it. And I have my animal
1:13:03 DNA maker thingy.
1:13:06 And now I'm like, huh, that's kind of
1:13:09 cool. You know what that would be cool
1:13:11 as? That would be cool as like a a
1:13:14 children's story book. And then I go
1:13:16 over here and I can make a children's
1:13:18 story book. And I'm like, huh? You know
1:13:19 what that would be cool? That'd be cool
1:13:21 as a song. And now I can go over here
1:13:23 and I can make it a song. Then I can
1:13:24 make it a movie. Then I can make it a
1:13:26 business presentation. Then I can launch
1:13:27 a business.
1:13:30 Every single person
1:13:33 who chooses to use AI and who who
1:13:35 chooses to actually understand what it
1:13:37 makes possible
1:13:40 can now just live in idea space
1:13:44 and and think like a producer like,
1:13:47 okay, I got an idea here. It's going to
1:13:49 be big. Great. what you needed to do.
1:13:53 And then the better you get at at using
1:13:56 all these disperate tools, the more gaps
1:13:59 you fill in on yourself, the better you
1:14:01 get at chat GPT, the better you get at
1:14:03 vibe coding, the better you get at
1:14:04 making images, the better you get at
1:14:07 duct taping all these things together.
1:14:09 and what's happening.
1:14:12 One of the reasons, especially for
1:14:14 people that are new here tonight, one of
1:14:16 the reasons people coming back, keep
1:14:18 people keep coming back to this channel
1:14:19 night after night after night
1:14:24 is that we get to collectively watch the
1:14:27 evolution of these tools where we know
1:14:30 how hard it was to get something usable
1:14:32 three years ago and then we knew it got
1:14:35 a little bit easier two years ago and a
1:14:37 little bit easier a year ago. And now
1:14:39 we're kind of marveling at, wait a
1:14:41 minute, I can just go to Sora and I can
1:14:43 make a video with me and Sam Alman
1:14:45 without having to actually create a
1:14:48 Laura model and install open-source
1:14:52 software. And yeah, no, you can just do
1:14:55 that now.
1:14:57 And so that's going to continue to
1:14:59 evolve. And then what happens is all of
1:15:03 the tactical execution
1:15:06 and especially the tactical execution
1:15:09 that doesn't require a lot of thinking.
1:15:11 It's like that blocking and tackling of
1:15:14 I've got to write out the code. I got to
1:15:16 block it out first. I've got to write
1:15:17 out the outline of the presentation.
1:15:19 I've got to write out the story. I've
1:15:21 got to sketch the images. All that
1:15:23 blocking and tackling work that takes
1:15:26 brain cycles but isn't really creative,
1:15:30 that's just going to be done. It's going
1:15:32 to be done for us and all we're going to
1:15:35 have to worry about is the idea and the
1:15:38 fidelity of the idea and our taste.
1:15:43 Holy [ __ ] how cool is that?
1:15:46 You know,
1:15:48 Sununo Studios awesome. At first I
1:15:50 thought, is this AI? No way. This this
1:15:53 was my idea. I just use tools to make it
1:15:55 reality.
1:15:58 You know, it's funny. I I I made a
1:16:00 comment today here. Let me I want to
1:16:01 show you something.
1:16:05 I want to show you all something
1:16:16 Tik Tok pin Mary.
1:16:20 Yeah. You don't have to you don't have
1:16:22 to build a cell phone to be able to use
1:16:23 one. Exactly. Right. I mean, RadioShack
1:16:27 was built on the idea that everyone
1:16:29 wants to build a cell phone. No, some
1:16:32 people just want to use cell phones,
1:16:33 right? We don't always want to build
1:16:35 them. And that's one of the one of the
1:16:37 challenges of AI right now is that
1:16:40 because there's, you know, a 50-year
1:16:42 legacy of of classic AI, classical AI
1:16:45 that really did require you to be an
1:16:47 engineer and a mathematician and things
1:16:48 like that, people still think that it's
1:16:51 got that requirement that you've got to
1:16:53 be able to build a cell phone. No, you
1:16:56 don't. We're like chat GPT marks the
1:16:58 turning point where you can just use
1:17:00 cell phones now, right? That's that's
1:17:03 that's why this is different. That's why
1:17:05 this channel exists. But I want to I
1:17:06 want to I want to share something here.
1:17:09 Um
1:17:11 so John Leland um who's a a salon member
1:17:14 and he's he's got his creativity work
1:17:16 that he does. Um he he uh posted about
1:17:20 this this video that this guy made. And
1:17:23 so let's let me share this tab. So he
1:17:27 posted So here's this video that this
1:17:28 guy Isaac Rodriguez made.
1:17:33 Hello.
1:17:41 [Music]
1:17:47 [Music]
1:17:51 Huh?
1:17:56 [Music]
1:18:01 Hey,
1:18:07 [Music]
1:18:18 heat. Hey.
1:18:21 [Music]
1:18:22 [Applause]
1:18:23 [Music]
1:18:28 Okay. So, we're going to stop it right
1:18:29 there. Um,
1:18:36 so this is a a a
1:18:39 series, a video series that this guy's
1:18:42 putting out there.
1:18:44 And he said something here. Introducing
1:18:46 The Valiant Night, episode 1, a fun
1:18:48 short series created 100% with AI.
1:18:53 Um,
1:18:55 I took issue with this.
1:18:58 What he's talking about here is that the
1:19:01 images that you see here and the
1:19:03 animations and the audio and the music
1:19:06 were created with AI tools.
1:19:09 I would absolutely argue this was not
1:19:12 created 100% with AI. My comment was,
1:19:17 let's see where I put it. It might be on
1:19:18 John's post.
1:19:27 Um, my comment was, "This is great.
1:19:30 However, I would argue this is not 100%
1:19:32 created with AI. Isaac had the idea, a
1:19:36 vision, learned how to prompt, discern
1:19:38 which tools could deliver the quality he
1:19:40 was looking for, created the script,
1:19:42 even if it was AI assisted. He
1:19:45 ultimately chose what was being shown
1:19:47 and said, and how." He generated 10
1:19:50 times or more the number of clips he
1:19:52 didn't use. He curated the best ones
1:19:54 out. He edited them into a final piece
1:19:56 and he package packaged it up and shared
1:19:59 it with the world. AI tools are enabling
1:20:02 tools in the stack. But this is a human
1:20:05 creation. I look forward to the day when
1:20:08 we don't have to declare the tools we
1:20:10 use. Important disclosure. I wrote this
1:20:12 comment on an Intel Mac with an Apple
1:20:15 extended keyboard and the Chrome
1:20:17 browser.
1:20:19 As stupid as that little last sentence
1:20:22 says, when when people like, "Was that
1:20:24 AI? Did you use AI for that?" It's none
1:20:26 of your [ __ ] business. Quite frankly,
1:20:30 what tools I use in my tool stack. If I
1:20:32 had an idea for something and I brought
1:20:34 it to life and I put it in the world,
1:20:37 you can judge it in one of two ways. You
1:20:39 like it or you don't. If I'm happy with
1:20:42 it as a creator, I created it. It's
1:20:45 independent of the tools you used.
1:20:47 So, I would encourage you as you start
1:20:50 thinking about these AI tools, don't
1:20:52 give them more power than they already
1:20:54 have.
1:20:55 If you say that this was 100% created,
1:20:58 Isaac just discounted his contribution
1:21:01 to this entirely.
1:21:04 And I [ __ ] hate that.
1:21:08 Like, this is a wellput together video.
1:21:10 It's a It's a It's a strong well-put
1:21:13 together video.
1:21:15 And that's not because of AI. That's
1:21:18 because of his taste, his storytelling.
1:21:22 Right. So anyway, all right.
1:21:26 Enough [ __ ] ranting.
1:21:29 I'm losing my voice.
1:21:33 What you saying should be on a t-shirt.
1:21:37 I don't quite know what I said that
1:21:38 could be on a t-shirt. That's a good
1:21:40 rant. Thank you very much, Joe. Creative
1:21:41 AI coach. Yeah, like like I did I did
1:21:45 another post to this today. Oh, here
1:21:47 we'll we'll do this. Okay, I'm gonna let
1:21:51 me show you another LinkedIn post today.
1:21:53 I'm I'm on a rant right now. I'm on a
1:21:54 I'm on a [ __ ] bender about this cuz
1:21:57 it's driving me [ __ ] crazy. Was that
1:21:59 AI? Did you make that with AI?
1:22:06 Did you ever use Photoshop, you big
1:22:08 [ __ ] talentless loser?
1:22:13 That's exactly what they're saying.
1:22:15 Cameras don't create a great photograph.
1:22:17 Yeah, cameras are a great example.
1:22:19 Source Cam, you know, you know there's
1:22:20 things in the world called snapshots,
1:22:22 right? And then there's things called
1:22:25 photographs,
1:22:26 and then there's things called art. What
1:22:30 is the difference between them?
1:22:32 You could take the exact same camera and
1:22:35 take a snapshot with it, a photograph
1:22:37 with it, or art with it. What's the
1:22:39 difference?
1:22:42 What's the difference?
1:22:44 The person holding the camera, what they
1:22:46 look at, where they focus, when they
1:22:48 push the button, how they edited
1:22:50 afterwards, what their intent was.
1:22:52 Right? A snapshot is just, yeah, here's
1:22:55 a picture. We are on vacation. This is
1:22:56 cool. A photograph is I actually looked
1:22:59 at that. Art is I took my time. I have a
1:23:01 point of view. I'm not going to let it
1:23:02 go out to the world until I am excited
1:23:04 about it. And then it's going to do what
1:23:07 it's going to do. AI is the exact same
1:23:09 [ __ ] thing.
1:23:12 Um,
1:23:15 let's see.
1:23:18 Where's the boundary? So, this is a
1:23:20 prompt I did today or post I did today.
1:23:27 So, I'm going to play the song first.
1:23:31 So, this is a song I made last week here
1:23:33 in the AI learning lab.
1:23:37 was asleep.
1:23:40 [Music]
1:23:47 The ferris wheels asleep, its arms still
1:23:50 folded tight. We walk between the
1:23:53 shadows and the strings of halfway
1:23:56 light. The popcorn stands are dreaming
1:23:59 of the laughter they'll be fed and your
1:24:02 reflection in a puddles turning pink
1:24:05 instead of red.
1:24:11 You say it's strange to see it empty.
1:24:15 I say it's strange to feel this calm.
1:24:21 You smile like you know something I
1:24:25 don't know. I've already begun.
1:24:30 Before the lights come on, before the
1:24:33 crowd arrives. Before we name what this
1:24:36 could be, before the day survives, we
1:24:40 linger in the middle where the maybe
1:24:43 feels like song. Two hearts not yet
1:24:46 decided
1:24:48 before the lights come on.
1:24:51 [Music]
1:24:59 A prize bear in the windows. Got a rip
1:25:02 along the seam. You laugh and say he's
1:25:05 just like me, stitched up from a dream.
1:25:09 The smell of rain and sawust makes the
1:25:12 air taste almost sweet. I count your
1:25:15 footsteps next to mine. They nearly
1:25:17 match the beat. You toss a coin. It
1:25:20 spins between the quiet and the breeze.
1:25:23 It lands on something fragile, something
1:25:26 trembling between.
1:25:30 Before the lights come on, before the
1:25:34 noise begins, before the world demands,
1:25:37 we choose where all this story ends.
1:25:41 We're written in the margin where the
1:25:43 ink still barely strong. Two souls half
1:25:46 in the promise
1:25:49 before the lights come on.
1:25:54 [Music]
1:26:00 Cotton candy ghost. The whisper of a
1:26:05 secret that your hand could hold but
1:26:07 hasn't yet tried. If I breathe too loud,
1:26:11 I'll break the spell we've spun.
1:26:16 A heartbeat before gravity decides which
1:26:19 way we run.
1:26:23 Before the lights come on, before the
1:26:26 sky turns gold. before our voices learn
1:26:30 the lines they're scared to have told.
1:26:33 Let's walk this midway silence till the
1:26:36 dawn has found its song. And maybe we'll
1:26:40 decide it
1:26:45 before the lights come on.
1:26:50 [Music]
1:27:05 Okay. So, beautiful, right? So, here's
1:27:09 my question. Where's the boundary? Is
1:27:12 this AI or human? It's it's a song made
1:27:15 in Suno. That video's got a Puno logo on
1:27:17 it. So, that's an AI song, right? And
1:27:20 so, if someone goes, "Oh, did you use AI
1:27:22 for that?" What are they doing right
1:27:24 there? They're completely diminishing
1:27:26 it. They're assuming that how I made it
1:27:28 was this.
1:27:30 >> I pushed a button and out came a song.
1:27:33 So, what I wrote was this. There's a lot
1:27:35 of talk of the critical importance of
1:27:37 distinguishing between AI generated or
1:27:39 human generated. Listen to the song I
1:27:41 wrote or didn't, depending on where we
1:27:43 draw the line.
1:27:45 The original four bar guitar lick. That
1:27:48 was me. The idea for the song, the idea
1:27:51 of two people deciding if they're more
1:27:53 than friends, walking down a dirt road
1:27:55 at a carnival before the people show up.
1:27:57 My idea, the prompt that I put into chat
1:28:00 GBT to articulate that that idea. My
1:28:02 idea, the lyrics, AI.
1:28:06 And normally with AI lyrics, I'll edit
1:28:09 the [ __ ] out of them because they're
1:28:10 usually bad or I'll mix things together
1:28:12 or whatever.
1:28:14 The lyrics were so good, I didn't have
1:28:16 to edit them, right? the music prompt.
1:28:19 Um, Chat GPT wrote the initial one and
1:28:22 then I did nine different variations of
1:28:24 the prompt trying to get the song I
1:28:26 wanted. So, that was a combination of AI
1:28:28 and me. The initial song generation, the
1:28:31 first song generation was AI. Um,
1:28:36 the the um curation of creative input.
1:28:40 So, we had people from the live giving
1:28:42 me ideas. So, I'm sort of curating
1:28:44 inputs from them. I'm having ideas of my
1:28:46 own. We're trying to get the song
1:28:48 better. I generated 21 versions of that
1:28:52 song that I didn't like, that didn't
1:28:55 sound like the vision. The first one
1:28:57 sounded like what I what I imagined, but
1:29:00 there was weird things with it. And then
1:29:02 we tried to make it was it was a
1:29:03 fascinating process over over 22
1:29:06 versions. We started really good and
1:29:09 then as I tried to improve it, the song
1:29:11 got worse and worse and worse. And then
1:29:13 Corey Sandler on the channel said, "Oh,
1:29:16 why don't you try this? Why don't you
1:29:17 try that? Why don't you put in cellos
1:29:19 with I don't know, piccolo or whatever
1:29:21 the [ __ ] it was, an obo." And and it
1:29:23 started getting better and better and
1:29:24 better. And then the one that you just
1:29:26 listened to was the 20 second song.
1:29:29 And if as I listen to that song now,
1:29:32 there are no misprononunciations of
1:29:33 words. There's no weird chord
1:29:36 structures. It's just a good song.
1:29:37 Right. So 21 bad versions. Um, uh, you
1:29:42 know, the songs were fine, but but
1:29:45 I just didn't like them. That was me
1:29:47 choosing the 22nd song as the one. That
1:29:50 was me. The album art was Corey Sandler
1:29:55 had the idea. She prompted it into an
1:29:57 image. I chose to use it. And then
1:30:01 choosing to share that song with the
1:30:03 world. That was me.
1:30:05 So, in that stack of creative choices,
1:30:08 there's like three or four things in
1:30:10 there that are AI. They're important
1:30:12 things, but the bulk of the work, what
1:30:15 took 90 minutes
1:30:18 was all the other [ __ ]
1:30:21 You can't tell me that's not a human
1:30:25 creation.
1:30:26 I don't accept it. I don't. And I'll
1:30:29 [ __ ] I'll [ __ ] argue this one up
1:30:31 one side and down the other.
1:30:35 If you use Grammarly, does that mean you
1:30:38 don't write
1:30:40 or AI wrote your thing? So anyway, yeah,
1:30:44 the perception is AI, you push a button
1:30:46 and out out squirts slop. And to be
1:30:49 fair, you can treat AI that way, right?
1:30:52 That's why it has a bad reputation. But
1:30:54 everyone that I know and certainly the
1:30:56 people on this [ __ ] channel spend a
1:30:58 lot of time trying to get AI to bend
1:31:02 toward their vision. That's human
1:31:04 creation. So anyway, so my last question
1:31:06 is, so is this an AI song or just a
1:31:09 song? And does it really matter? You
1:31:12 know, my opinion is if I put that song
1:31:14 out and you don't know it's AI and you
1:31:16 like that song and it moves you. Someone
1:31:18 just said on the on TikTok over here
1:31:20 they downloaded it to their iPhone
1:31:22 because they want to listen to it.
1:31:24 They're not listening to it because it's
1:31:25 an AI song. They're listening to it
1:31:26 because they like the music,
1:31:29 right?
1:31:32 Anyway, all right. Tyade Tuesday.
1:31:37 It's a song. Exactly.
1:31:40 Um,
1:31:43 such a beautiful song. Can I show my AI
1:31:46 class I teach? Absolutely. They're
1:31:48 middle schoolers. It's a be It's
1:31:50 actually a middle school song.
1:31:53 It is absolutely a middle school song.
1:31:55 Like I imagine this is like a 14 like a
1:31:59 14year-old girl and a 15year-old boy and
1:32:02 they've been friends since elementary
1:32:04 school and they sneak into the carnival
1:32:06 before it opens and it rained last night
1:32:10 and they're walking through the carnival
1:32:12 and that was like when I first played
1:32:14 the guitar lick.
1:32:20 Um,
1:32:24 wait.
1:32:25 [Music]
1:32:38 Um,
1:32:40 >> I didn't have I didn't have the dog
1:32:42 howling when I did that, but like that
1:32:45 sounded to me like this innocent kind of
1:32:48 carnival,
1:32:50 you know, when you're a kid and you're
1:32:51 walking through the carnival and it's
1:32:53 kind of innocent and it's like this girl
1:32:55 and boy trying to figure out if they're
1:32:57 more than friends. And there's I mean
1:33:00 there's just some beautiful lines in it.
1:33:02 There's there's these beautiful lines
1:33:04 that I don't know I whatever there's
1:33:07 there's there's the the the the lyrics
1:33:09 in it are just beautiful. So yes,
1:33:11 absolutely. Show your middle schoolers.
1:33:13 It's a it's a it it's it's that spirit.
1:33:18 So
1:33:20 um
1:33:22 the genre was you such a beautiful song.
1:33:24 Yeah. Cool. Story of my life. All right.
1:33:27 Let's see.
1:33:29 And their hands are one apart as they
1:33:31 walk at first. Yeah.
1:33:36 Anyway.
1:33:38 All right. It's 9:30. Any other
1:33:41 questions? Any other thoughts? I know
1:33:43 Johnny D had a question about the
1:33:45 prompt. How do you get it to not change
1:33:46 everything?
1:33:48 I mean, one way is throw it into a throw
1:33:50 it into a canvas so you can just edit
1:33:52 parts of it. That's that's a way to do
1:33:54 it.
1:34:07 Yeah,
1:34:12 you you diminish 50 years of my life by
1:34:14 clicking at that moment with with a
1:34:16 camera. Yeah, exactly.
1:34:19 Um the pool of blood turned pink. Yeah.
1:34:21 Yeah, it was it was like your Well, I
1:34:23 don't think it was a pool of blood. I
1:34:25 didn't see it that way. I saw it as um
1:34:29 the sun's coming up, so it's going from
1:34:31 deep red to pink and she sees the
1:34:33 reflection of his face in a puddle and
1:34:35 it's going from from red to pink. Tik
1:34:38 Tok question, how do you mitigate AI
1:34:41 push back? It's affecting my ability to
1:34:43 implement tech. Um
1:34:47 that's a I the short answer is I don't
1:34:50 know.
1:34:53 Um,
1:35:02 what I what I can tell you, so what I
1:35:06 try to do here when when when when
1:35:08 people come in here,
1:35:11 you know, leaning heavy on the hate, and
1:35:14 I've gotten my ass handed to me on
1:35:15 LinkedIn a couple of times where I did a
1:35:18 post and, you know, the haters the
1:35:19 haters came at me. Um, every in history
1:35:24 in the history of humanity, every single
1:35:26 time there's a transformative
1:35:28 technology, every single time, every
1:35:31 single time, people resist it and
1:35:34 sometimes dramatically. I have a feeling
1:35:38 AI is going to be one of those that gets
1:35:40 resistance dramatically. What I've
1:35:42 experienced is that most of the people
1:35:44 that hate it, hate it,
1:35:48 have very strong opinions about it, have
1:35:51 never tried it.
1:35:53 They're listening to the tropes. They're
1:35:55 listening to, okay, they they stole they
1:35:58 they stole from artists. Okay, it's bad.
1:36:00 Therefore, it's 100% bad 1,000% of the
1:36:03 time. I'll never use it. It's evil.
1:36:05 Right? So, they they throw it into this
1:36:07 categorical evil category. We've got 50
1:36:11 years of Hollywood telling us the robots
1:36:14 are going to kill us. We've got all of
1:36:16 those tech bros in San Francisco that
1:36:19 are building this [ __ ] going, "It is
1:36:20 going to take your jobs." And some of
1:36:22 them are [ __ ] and they're like, "I
1:36:23 can't wait for it to take all the jobs."
1:36:26 Um,
1:36:28 that's [ __ ] terrifying, right? This
1:36:30 is people's livelihood.
1:36:34 When you see something like Sunno,
1:36:35 there's a really amazing video the
1:36:37 series of Tik Tok videos of this of this
1:36:40 musician
1:36:42 who takes his songs and puts them into
1:36:44 Sunno and turns them into full
1:36:45 production songs and he kind of does it
1:36:48 in real time and he does it over a
1:36:50 series of days and he is at the same
1:36:53 time mortified and inspired
1:36:58 and he's also smart enough to know that
1:36:59 this stuff's not going away so that he's
1:37:01 got to find a way to to get on the other
1:37:03 side of the mort mortified part of
1:37:05 himself like holy [ __ ] my my career is
1:37:08 over is is one of his reactions. The
1:37:10 other one is holy [ __ ] this technology
1:37:12 is really cool and amazing and and it's
1:37:15 not going away.
1:37:17 So he's one of the rare people that can
1:37:20 actually hold those two
1:37:23 emotions at the same time. And I I feel
1:37:25 like that's actually a a fairly
1:37:27 sophisticated
1:37:30 place to be that if you're using AI, it
1:37:35 it it's some of this [ __ ] is absolutely
1:37:38 [ __ ] terrifying.
1:37:40 Like what what power it's got. But then
1:37:43 you're like, but the power is in my
1:37:45 hands, too, so I can go do something
1:37:46 with that, right? So it's both.
1:37:51 I don't think you can kind of convince
1:37:54 people to love it.
1:37:58 I think the best that you can do is you
1:38:00 can say, "Listen, hold that hate. I
1:38:03 understand where the hate's coming
1:38:04 from." Like it's it's just it's like
1:38:06 this one of those marriage counseling
1:38:07 things. Yeah. I I like mirror it back to
1:38:10 them. I absolutely understand why you
1:38:12 hate it. You know what? There's things
1:38:13 about AI I hate too. These four things
1:38:16 you just talked about, I agree with you.
1:38:19 I absolutely hate that part of it, too.
1:38:21 These three things you talked about,
1:38:23 it's not actually how it works. But I I
1:38:25 mean, I'm happy to explain that, but
1:38:26 that doesn't matter. But you know what?
1:38:30 When you experience it, it's a very
1:38:34 different relationship than when you're
1:38:36 just looking at it from the outside
1:38:37 because you can see that it's not
1:38:39 perfect. And if you can get them to give
1:38:42 it a shot and if you can get them to do
1:38:44 things that are relevant to them
1:38:47 and they try it and they have one of the
1:38:50 things we talk about in here is is
1:38:54 one of my joys in life is watching
1:38:56 people have their first Kevin Mallister
1:38:58 moment. If you don't know Kevin
1:39:00 Mallister, he's the the kid from Home
1:39:02 Alone where he puts his dad's after
1:39:04 shave on. He's like that moment when you
1:39:08 use AI when you're doing it right and
1:39:10 you and especially when you do something
1:39:11 that's relevant to you. So if you're a
1:39:13 writer you're like well it couldn't
1:39:14 possibly write this kind of thing and
1:39:17 then you prompt it and it does and it's
1:39:19 good
1:39:21 you're like holy [ __ ] Uh right that
1:39:25 moment like oh my god
1:39:28 when people have that first moment with
1:39:30 AI you kind of can't unhave it. And so
1:39:34 what I've experienced is people that are
1:39:36 really pissed off about it, even if they
1:39:38 reluctantly come into it and give it a
1:39:40 shot and play with it enough to have
1:39:43 their first Kevin Mallister moment,
1:39:46 then they're like, "Uh, okay, I think I
1:39:49 still hate it, but this is kind of
1:39:50 interesting." And then they get to, "Oh,
1:39:52 I think I kind of hate it, and actually
1:39:54 this is kind of inspiring." And then
1:39:55 they they get to this more sophisticated
1:39:58 place where they're both excited and
1:40:00 terrified at the same time. And then
1:40:03 sometimes like my co-founder of the AI
1:40:05 salon, Leah Faston, she's a professional
1:40:08 photographer,
1:40:10 she went through that cycle and she
1:40:12 actually is stepping away from the AI
1:40:13 salon because she realized for her
1:40:16 personally AI was not feeding her
1:40:19 creatively. She just didn't like it. And
1:40:22 like I have the deepest [ __ ] respect
1:40:24 for that because like she came into it
1:40:27 hard. like she co-founded a 30,000
1:40:30 person community with me of AI optimists
1:40:35 and then realized it wasn't for her.
1:40:38 Like that I'm totally down with. But
1:40:40 it's the people that are sitting on the
1:40:42 outside that are just using these tropes
1:40:44 like it's the world's greatest
1:40:45 plagiarism machine. It steals. It uses
1:40:47 too much water. It uses too much energy.
1:40:50 Some of those facts are true. Some of
1:40:52 them are not. But if you're using those
1:40:54 things as an excuse to sit out and not
1:40:57 try it and not understand what it makes
1:41:00 possible and not understand what it
1:41:01 might do for your ideas,
1:41:05 it just feels really [ __ ]
1:41:06 shortsighted to me. Like that's what
1:41:08 this channel's about. So
1:41:11 there you have it,
1:41:15 man. I'm on a [ __ ] tear tonight.
1:41:18 Tyrate Tuesdays is where it's at.
1:41:21 Hallelujah.
1:41:23 [Laughter]
1:41:28 Oh, good lord. Did that make any sense?
1:41:33 It'll be difficult to go on tour.
1:41:35 Actually, you know what's funny,
1:41:37 Professor Crystal? I think in-person
1:41:40 concerts. I think the value of in-person
1:41:42 things are about to go up in value
1:41:45 dramatically
1:41:48 because it's going to be so easy to
1:41:49 produce all this digital stuff. People
1:41:52 are going to have like hunger for human
1:41:56 connection.
1:41:58 Um I think in a weird way robots return
1:42:01 us to our humanity. I think they they
1:42:04 allow us to ascend
1:42:07 out of the drudgery of the systems that
1:42:11 we have to interact with today by auto
1:42:14 automating a lot of that. Let us ascend
1:42:15 out of that and then I I think the thing
1:42:18 left to do when you ascend out of the
1:42:20 drudgery is reconnect.
1:42:23 And so I think I think there will be
1:42:25 jobs like um I'll walk with you for an
1:42:29 hour for 50 bucks and people will pay 50
1:42:32 bucks to have someone come walk with
1:42:34 them. And I know that sounds weird right
1:42:36 now, but like we've got people making a
1:42:38 million dollars a year running YouTube
1:42:40 channels. I'm not one of them. I I need
1:42:43 to talk to one of those people and
1:42:44 figure out how they're doing it.
1:42:47 Um, but
1:42:51 like we're gonna we're going to want to
1:42:53 reconnect. So, I think I think tours and
1:42:55 concerts and you know, take take songs
1:42:58 that you've written in Sunno and start a
1:43:00 band. Make a cover band of the songs you
1:43:03 wrote using AI. Yeah. The Taylor Swift
1:43:07 four billion dollar economy.
1:43:10 All right. It's like when Fetch Fred
1:43:12 Savage had the Power Glove. Exactly.
1:43:14 It'll be difficult to go on tour.
1:43:18 Tony Williams, my cousin just went
1:43:20 through the same cycle of fearing it to
1:43:22 being inspired. Yeah, it's so I've got
1:43:25 Let me share one other thing and then
1:43:26 we'll we'll get on out of here. If you
1:43:28 go to the AI salon, so if you go to the
1:43:30 salon.ai and say join our community,
1:43:34 the very first little space is called
1:43:36 welcome to the AI salon
1:43:40 and
1:43:48 I've got in here what I call the we've
1:43:50 got the cycle of AI readiness. Play
1:43:53 first, create excellence, generously
1:43:55 lead. That's kind of the the the ethos
1:43:57 of the salon. Um the mission and I've
1:44:01 got the the five stages of AI adoption.
1:44:05 And I base these on the five stages of
1:44:07 grief because I because I think it's a
1:44:08 similar kind of thing. The first stage
1:44:10 of grief is is denial or dismissal. And
1:44:13 so I think a lot of people right now
1:44:15 they they've either demonized AI or
1:44:17 they've dismissed it. It's not very good
1:44:19 or or or they're just, you know, it's
1:44:21 bad, it's evil, right? It's going to
1:44:23 kill us, right? And so so step one is
1:44:26 dis dismissal. It's just another tech
1:44:28 fad. Nothing to see here. People
1:44:30 dismissing. Oh, did you use AI for that?
1:44:32 I can judge you now. I can judge you as
1:44:34 a talentless oaf. Yeah. Yeah. Yeah. AI.
1:44:38 I push a button, out comes a thing.
1:44:39 Yeah. Yeah. Whatever. Right. I get a lot
1:44:42 of that, right? That's dismissal right
1:44:44 there in stage one. Stage two is you've
1:44:46 had your first Kevin Mallister moment.
1:44:48 It's this moment of awe
1:44:51 where you actually for the first time
1:44:53 understand
1:44:55 you you internalize
1:44:58 kind of the power of what this thing is.
1:45:00 Computers have never acted this way in
1:45:03 our lifetimes.
1:45:06 Computers have always taken an input,
1:45:08 processed that data in some sort of
1:45:11 logic that was programmed by a human.
1:45:14 That was a deterministic output, right?
1:45:16 You put in an input, it processes it,
1:45:19 and it comes out in a predictable way.
1:45:21 That's not what computers are doing when
1:45:23 they're using generative AI. They are
1:45:25 generating new original stuff. Stage two
1:45:28 is where you experience that and you go,
1:45:31 "Oh,
1:45:32 I didn't know it could do that. I had no
1:45:34 idea that phrase, I had no idea. I hear
1:45:38 that over and over and over again where
1:45:40 I'm showing someone something. They're
1:45:42 like, "Oh yeah, I use chat GPT all the
1:45:44 time." Yep. Yeah. I use it for emails.
1:45:46 Yep. Yep. Yep. Yep. I use it a lot. Yep.
1:45:49 And then I'm like, "Oh, did you see it
1:45:50 could do this? Oh my god, it could do
1:45:52 that." And I'm like, "Well, yeah, if you
1:45:54 didn't know it could do that. Did you
1:45:55 know it could do?" Oh my. It can do I
1:45:58 had no idea. Like that I had no idea
1:46:01 moment. That's stage two. you realize
1:46:04 all of a sudden this thing is something
1:46:06 different, right? Stage three is you now
1:46:09 start playing with it and you you want
1:46:10 to discover it on your own and and you
1:46:12 go, "Oh my god, it can do this. It can
1:46:14 do that. It can do that." You get really
1:46:15 excited and then you realize, "Holy
1:46:17 [ __ ] I've spent 30 years trying to be
1:46:20 good at that and now it's better than
1:46:22 me." Fear, wonder, fear that coexists.
1:46:26 Stage three is this coexistence of
1:46:28 wonder and fear. Stage four is when you
1:46:32 kind of let go of the ego and say,
1:46:36 "Okay, it's smarter than me." Because it
1:46:39 is. If you don't think it is, it is. I'm
1:46:41 sorry. It is.
1:46:45 And you give into it and you say, "Wait
1:46:47 a minute. I could either be in
1:46:50 competition with this thing or I could
1:46:53 be in collaboration with this thing. I'm
1:46:55 going to choose to be in collaboration
1:46:56 with this thing. And what I'm going to
1:46:58 do is I'm gonna have the ideas and I'm
1:47:01 going to strap this AI on thing like a
1:47:03 jetpack and I'm going to have it amplify
1:47:06 my ideas. That's stage four. I find
1:47:08 myself right now going back and forth
1:47:10 between stage three and stage four. I
1:47:13 use it like a jetpack. I'm like, "Oo,
1:47:15 this is great." And then like some new
1:47:16 thing comes along and I'm like, "Holy
1:47:18 [ __ ] Oh my god, that's amazing. Oh my
1:47:19 god, that's terrifying." Right? Until I
1:47:22 kind of internalize it and put it back
1:47:23 in the jetpack. And then stage five is
1:47:26 is kind of pure self-expression or or uh
1:47:30 David Shapiro calls it cognitive
1:47:32 hyperabundance.
1:47:33 The the point at which it's just
1:47:35 literally anything you want you just ask
1:47:37 for and it will materialize.
1:47:40 We're not anywhere close to stage five
1:47:42 yet. The tech the tech isn't there. So
1:47:45 anyway
1:47:46 anyway all right. Yeah. Well, we're not
1:47:50 special.
1:47:53 Not special. The salon is great. So, do
1:47:56 me a favor. If you have not joined the
1:47:58 AI salon, if you have not joined the AI
1:48:00 salon, get your ass there. The salon.ai.
1:48:07 We have a brand new a shiny new website
1:48:12 that's amazing. Tells you all about it,
1:48:14 all the stuff we're up to. You'll see
1:48:16 events. You'll see posts from the
1:48:18 community.
1:48:21 And you say, "Join our community." and
1:48:22 that'll take you into the community
1:48:24 site.
1:48:26 And we're rearchitecting it right now.
1:48:28 We're kind of halfway through that. It's
1:48:29 really exciting. All right, so go do
1:48:31 that. Um, today's Tuesday, tomorrow's
1:48:34 Wednesday. So, tomorrow night it'll be
1:48:35 8:00 p.m. Same time, same place. Um,
1:48:39 I didn't even get to there were a couple
1:48:41 of things I wanted to show tonight.
1:48:42 There's some new features in Runway ML
1:48:45 that I thought were kind of exciting.
1:48:47 Um, I'm going to do So, when did I when
1:48:52 did we do back to basics, Brandon?
1:48:59 Three two months ago. September.
1:49:02 Okay. So in September I did this thing
1:49:04 called back to basics where I took five
1:49:06 days and day one was overview of AI, day
1:49:10 two was chat GPT, day three was images,
1:49:13 day four was sort of integrating it and
1:49:15 day five was questions.
1:49:18 I'm going to do a new back to basics.
1:49:21 I'm going to do a new five-day course.
1:49:23 I'll probably do it in the next three
1:49:24 weeks, four weeks, something like that.
1:49:27 I need to put it together. It's going to
1:49:29 be five days. So, five nights here
1:49:33 and all five nights are just going to be
1:49:35 chat GPT.
1:49:38 We're going to do five nights of chat
1:49:40 GPT
1:49:41 just figuring out what the [ __ ] is in
1:49:44 it. Because one of the things that I
1:49:47 confronted when I did my back to basics
1:49:50 course where chat GBT only had an hour
1:49:53 was in a in one hour I could not even
1:49:57 scratch the surface of what chat GPT now
1:50:00 does. So I'm going to do a full 5day
1:50:03 week coming up. So just know again just
1:50:06 keep coming here nightly. Become an
1:50:09 irregular. If you show up here nightly
1:50:11 you are an irregular. It's weird. You
1:50:13 shouldn't do that. You shouldn't come
1:50:15 here,
1:50:16 but people do,
1:50:19 but keep doing that. And then and then
1:50:21 we're going to do a deep a relatively
1:50:23 deep dive on chat GPT. We're going to
1:50:25 take a full week on it. Um, so anyway,
1:50:29 all right.
1:50:31 Beautiful. Beautiful. Tomorrow, Brandon
1:50:33 and Gwyn are at MCON, so they're there
1:50:36 as representatives of the AI salon. All
1:50:40 chat all the time. Exactly. How hard is
1:50:43 it to learn how to build an app? How
1:50:45 complicated of an app can I build by
1:50:47 myself? I'm has. So, first of all, what
1:50:50 you need to do is go watch the beginning
1:50:52 of this episode. So, when we stop when I
1:50:55 press stop tonight, this will be live on
1:50:59 the YouTube channel. So, the YouTube
1:51:01 channel is learning lab-ai
1:51:05 on YouTube.
1:51:07 Um, we vibecoded at the beginning of
1:51:09 this. We created an amazing app with a
1:51:11 single prompt. You can do incredibly
1:51:13 sophisticated apps with zero knowledge.
1:51:17 Um, they may not be scalable. They may
1:51:19 not be perfect, but if you've never
1:51:21 coded before, what you can do right now
1:51:24 with zero knowledge is insane.
1:51:28 And in fact, if you just want to blow
1:51:30 your own mind and and not watch me do
1:51:33 it, just go to lovable.dev.
1:51:36 dev lovab.dev.
1:51:40 Just go there right now, describe your
1:51:42 app, and hit the button and then sit
1:51:44 back and wait for five minutes and see
1:51:46 what it generates. It'll blow your mind.
1:51:49 It'll blow your mind.
1:51:52 Yeah, exactly. Try lovable.dev. There
1:51:55 you go. All right. Um, that's it. I hope
1:51:59 you all had fun tonight. Um,
1:52:03 enjoyed my little tirade. I was all
1:52:05 passionate tonight. Good lord. Um, I'm
1:52:09 gonna go rest and have fluids. I can
1:52:12 tell that I talked too much and didn't
1:52:13 drink enough. So, I will see you
1:52:16 tomorrow. Bye, everybody.