4/25/2025 - AI-Powered Tools for Creative Exploration and Vibe Coding

Description

Chapters

Transcript

0:02 [Music]
0:15 [Music]
0:21 She came on the hill like
0:23 snow. She came on him
0:27 like cold.
0:31 His bill was one and a look in her
0:36 [Music]
0:38 eye. She sat on a stool and he said,
0:42 "What do you
0:45 want?" She said, "Give me a love that
0:48 don't freeze up inside."
0:53 [Music]
0:58 You said I have melted some holes in my
1:01 time
1:03 here. But sitting next to you, well, I
1:07 shiver and
1:11 share. If I knew love, well, I don't
1:14 think I'd be
1:18 here. Asking myself if I got what it
1:24 takes. A new man your
1:29 [Music]
1:32 eyes. Should I
1:35 stop? Turn what's been frozen for
1:41 years into a river of tears.
1:48 [Music]
1:55 Yes,
1:57 Champy bringing it in with a solid
1:59 ending. Just
2:00 boom. That was good. Good evening, good
2:03 people. Friday night date
2:05 night. Get those nachos blasting.
2:10 [Music]
2:38 Woohoo! Freedom came on that night.
2:42 [Music]
2:45 Just like a jet plane in and out of
2:50 sight. I was hauling ass a million miles
2:55 an
2:56 hour wondering how hard I'd hit.
3:03 When they came into the
3:09 station, they said I was bad beyond
3:14 repair, but I got no problems with my
3:21 situation. Say I am.
3:25 [Music]
3:31 So, oh man, Friday Night Ed Murphy was
3:35 shaking. What's going
3:36 down? What's going
3:39 down? I The reason if I if I look a
3:43 little like Did my Did my light burn
3:46 out? If I don't look well lit.
3:49 Um I have a I have a an an ocular
3:53 situation going
3:57 on. So, I think I've been bitching for
3:59 the past couple of weeks about if I wake
4:03 up if I haven't had enough water the
4:06 night before, I wake up with like a
4:07 scratched cornea kind of thing. So, I
4:10 actually went to the doctor. And
4:13 uh he was like, "Oh yeah, let's put some
4:16 drops in your eye." I'm like, "What are
4:17 those?" He goes, "Those are to numb it."
4:19 I'm like, "What are we doing?" He's
4:21 like, "We're going to do some shit."
4:25 He was a He was a doctor with a touch of
4:27 the tism. Didn't have much bedside
4:29 manner, but he knew his
4:33 [ __ ] Oh man. So, turns out I've got
4:40 um
4:42 recurrent corial
4:46 erosion, which sounds [ __ ] awful.
4:50 [Music]
4:59 But it explains a lot. So basically what
5:01 happens so Chad GPT taught me about
5:05 it. So 30 years ago I got a poke in the
5:09 eye from a dog
5:11 whisker and you know caused a little
5:13 tear or a scratch or an abrasion on the
5:15 cornea. Well,
5:17 apparently old corial injuries when you
5:21 get old like you know I
5:25 am they can turn into this condition and
5:28 basically what it is is the surface
5:32 cells I this might be TMI and if you're
5:35 like don't want to deal with this just
5:37 you know put on Jack Black playing the
5:40 the the plastic saxophone or something.
5:44 Um, but it's the the sort of top layers
5:46 of the
5:48 cornea don't um bind well to the next
5:52 layer of the cornea. And so you have
5:54 this like layer that shifts and moves.
5:57 And so every time you blink, you know,
6:00 this thing moves around and it [ __ ]
6:02 hurts.
6:04 So, so, so what he said was, "We're just
6:07 gonna, we're just going to move those
6:09 loose cells slag off your, you know, do
6:12 you ever watch uh you ever watch uh like
6:16 Alex? Alex, I forget what his name is.
6:18 The guy that makes knives like when when
6:20 they when they um make molten metal,
6:23 they they sort of, you know, pick the
6:25 top layer off and throw it away." That's
6:27 basically what he did. Took the top
6:28 layer of cells, just sort of squished
6:30 them off with a Q-tip.
6:32 Uh, and then uh and then put on a uh a
6:37 bandage contact lens, which I didn't
6:39 know was a thing. So, I have a bandage
6:41 contact lens, but my eyes are still a
6:43 little a little sensitive. So, they're a
6:45 little sensitive to the light. So, I
6:49 don't have my bright light on, and I've
6:50 got all my monitors turned way down so I
6:52 can so I can see without losing my
6:55 mind. But, all should be good.
7:03 [Music]
7:18 Well, every time I see you
7:21 now, you had that look in mind.
7:26 Every time I see your mouth,
7:29 I hear that smile. That's one of my
7:32 favorite lines. Every time I see your
7:34 mouth, I hear that smile. Damn, that's a
7:37 beautiful
7:39 line. H, sounds like you have a colossus
7:43 inside. Oh, calluses inside your eyelid.
7:46 Maybe. I don't know. I had the exact
7:49 procedure when I was 12 after a severe
7:51 burn. Oh,
7:53 fascinating. Yeah. And it's like the the
7:56 prognosis is basically fine. It's like
7:58 they're going to do this sort of
8:00 lightweight version that he basically
8:01 just did in the office with a Q-tip and
8:04 then the the contact is letting the uh
8:07 new cells
8:09 regrow. And then basically if they keep
8:11 delaminating then they do a version of
8:14 this where they you know do an on the
8:16 table procedure and and just scrape
8:19 harder and do the same thing. When did
8:22 you pick up
8:24 your guitar for the first time in
8:27 college? So, my mom bought me a
8:31 ukulele um bought me a
8:36 ukulele I think my senior year in high
8:38 school and she was like, you know, sing
8:41 guitar like the girls will like it if
8:43 you play guitar. So, I guess we couldn't
8:45 afford a guitar, so I got a
8:46 ukulele, which I don't think I don't
8:49 think really serves the purpose with the
8:51 [Laughter]
8:54 ladies. Uh, but I ended up I ended up
8:57 getting a really cheap 12 string guitar
8:59 that had a [ __ ] up neck, had a broken
9:02 truss rod, so you couldn't put 12
9:04 strings on it anymore. So, it was just
9:05 like a really wide six string. And
9:08 that's what I learned on. And
9:12 then and then I I just I don't know. I
9:16 just kind of [ __ ] around on the guitar
9:17 for most of my life. And then in
9:20 my late 30s, early
9:22 40s, I I kind of fell in love with
9:25 Martin Ston and his music and the way he
9:27 played. And I like I had electric
9:29 guitars and but I I was never good at
9:31 them. I was never really good at this.
9:34 And then I just decided I sold all my
9:35 electric guitars and just decided to
9:38 learn like finger style like Martin Ston
9:40 does.
9:44 So promotional site salon and Facebook
9:47 niche group AI looking for a good
9:50 general AI group. Oh, promoted the
9:52 salon. Oh, got it. Cool. Cool. TK that's
9:56 awesome and good to see you.
9:59 [Music]
10:10 Welcome spin B3. What's
10:13 [Music]
10:17 [Applause]
10:24 happening? Ooh.
10:27 [Music]
10:54 Have I played with co-pilot agents? I
10:57 have not. Um I'm kind of with Ethan
11:00 Malik on this one. If you saw Ethan
11:03 Mollik's response, so Satia Nadala said,
11:05 "Hey, we've got agents now." And Ethan
11:08 Mollik's response was, "You keep adding
11:10 features and no one can [ __ ] find
11:12 them." I kind of feel the same way. I
11:14 feel like Microsoft and Google right now
11:17 are doing the exact same game. They keep
11:20 releasing what look like interesting
11:22 technologies that no one can find.
11:27 So, I haven't played with them. No, I
11:29 But they look interesting. I You know, I
11:32 don't know anyone that's using C-Pilot
11:34 other than people at work that can only
11:36 use Copilot. And then they're It doesn't
11:38 sound like they're using it much
11:41 cuz cuz it's just a convoluted [ __ ]
11:44 mess.
11:47 [Music]
11:56 I'm mostly Root Code. I don't even know
11:58 what Root Code
12:01 is. So, tonight what I figure we'll play
12:04 with I played with um I've talked about
12:07 Gen Spark in here before and I played
12:10 with it once before and then before we
12:12 hopped on here tonight I tried another
12:14 thing with uh Gen Spark
12:18 slides and it's pretty [ __ ]
12:21 impressive. So, I figure we might play
12:23 with that. Um Lovable is now at version
12:26 2.0. Might play with that. Um, and
12:30 there's a bunch of upgrades to chat GPT
12:32 that are worth experimenting with. So,
12:34 right before we came on, like an hour
12:36 ago, Sam Alman once again said
12:39 GPT40, so the basic non-reasoning model
12:44 got a personality upgrade and a
12:47 intelligence upgrade. So, apparently the
12:50 base model of chat GPT, the default
12:52 model is now more betterer and more
12:56 personalityfilled. Now,
12:59 um, what's fascinating about that
13:05 is every time we get used to
13:08 how these things
13:11 work, they change
13:13 them. And so it's like, okay, so so it
13:17 got a personality
13:19 upgrade. Is is that going to be better
13:21 or worse for someone who's kind of
13:23 imprinted on how it acts right now? I
13:26 don't know. I don't know. Fascinating
13:28 times. Um, just looked up Martin Martin
13:31 Ston. Good stuff. Never knew about him.
13:33 He's great. He's He's great. If you get
13:35 a chance to see Martin Ston live, he is
13:38 insanely talented. He's one of those
13:41 guys
13:45 that his voice and the
13:48 guitar are indistinguishable as
13:52 instruments. In fact, some of his early
13:54 stuff he did, he would have he would
13:55 have two different microphones. One he
13:57 would sing into and one he would
13:59 basically do like like um air guitar
14:03 with his voice. He would do like guitar
14:05 solos with his voice on a on an
14:07 effectdriven separate microphone and he
14:10 would play along with his guitar. Um
14:12 he's just he's amazing in person. Um I
14:15 got to see him once at a songwriting
14:17 summit. So, I was in a room with him
14:19 with like 20 other people and he started
14:21 singing a song and he was singing on a
14:22 microphone and he realized it was a
14:24 small room and he stepped in front of
14:25 the microphone and he hit this high note
14:29 and I swear to God it sounded like
14:31 trumpets were coming out of his face. It
14:33 was absolutely [ __ ] amazing. Uh, X
14:36 comment Josh Groves man I've been using
14:39 Gen Spark for a couple of months now and
14:42 I love it. It automatically made slides
14:44 for me the other day. I ran deep
14:46 research on getting evaluation for a
14:48 startup I want to pitch. I think they
14:50 just updated that. Amazing slides.
14:52 Amazing. Yeah. Like like watching what
14:55 it did I So I'll show you I'll show you
14:58 what I did. The the thing that I built,
15:00 it didn't quite do what I asked it to
15:02 do, but like holy [ __ ] what it did was
15:04 quite good. And and watching it do it. I
15:07 I'm telling you, man. Any of these any
15:09 of these deep research things where
15:11 you've got this sort of agentic behavior
15:14 where you give it a goal and it goes off
15:17 and starts doing [ __ ] and then updates
15:20 itself. Like one of the things that Jen
15:22 Spark did when I was making this
15:24 presentation was it was on like slide
15:27 seven and then it realized it had [ __ ]
15:29 up the formatting on slide three. So, it
15:31 went back and rewrote the the HTML for
15:34 slide three. Pretty cool. Tik Tok camps.
15:36 Oh yeah. Sorry. Sorry there people. What
15:39 are you gonna do? I'm just an old fat
15:42 guy trying to you. Hey. Hey Marge.
15:44 Marge. Marge. Yeah. Hey,
15:47 listen. Which one? How do I What's that?
15:51 Yeah. I I know you're watching the
15:53 wheel, hun. Is it a good Is it a good
15:56 puzzle? Yeah.
15:58 Yeah. Just Just figure it out myself.
16:03 Okay. Yeah, I'll do that.
16:06 Yeah. Yeah. Apparently the upgraded
16:10 version has usage caps even for plus
16:12 users. The upgraded version of what? Of
16:15 of
16:20 40. The uh the songs you should listen
16:23 to from Martin Sexton. Um he's got an
16:26 album called Live Wide Open which is
16:28 really good.
16:30 um, Black
16:32 [Music]
16:36 Sheep. This is usually what he closes
16:38 his shows
16:39 with. And I don't ever sing the whole
16:42 thing because I can't sing the end of
16:45 [Music]
16:51 it. Gareth, if you didn't know Kyle and
16:54 just popped into this, people would be
16:55 like, "What these? What's going on? Who
16:58 is this? I thought this was the AI
17:00 learning lab. Why is he talking about
17:02 Martin Ston? I don't even know who that
17:05 [Music]
17:12 is. Sitting in this lonely
17:15 town, wondering when things are going to
17:19 [Music]
17:21 change. Dream my life away.
17:24 Seems these dreams have turned to a
17:26 bunch of dust
17:29 clouds. Get my nerve up, but my past is
17:34 pulling me
17:36 down. AI learning lab is
17:40 multimodal. Wondering how
17:42 long this black sheep going to stick
17:46 around.
17:49 [Music]
17:51 Somebody told me once before he said you
17:54 can never go home
17:56 again. Won't you
17:58 leave? Santa things to steer me away.
18:02 Yeah. From the truth of who I am and
18:05 what I believe. So I thanked him for his
18:08 two cents with a handshake and some
18:12 [Music]
18:13 sympathy. Packed up my blue jeans and
18:17 headed for this big prize.
18:21 of my
18:23 freedom.
18:25 Bye-bye black sheep to the black sheep
18:29 of the
18:30 family.
18:32 [Music]
18:34 Bye-bye. Oh, it means so very much to
18:38 me.
18:40 Byebye. Okay, let's get
18:46 going. Welcome to chat add people. You
18:50 think we were never going to get started
18:51 and then boom, you're just in
18:53 it. You You went to make yourself a cup
18:56 of hot chocolate and you're like, I
18:57 thought I had minutes. He was in the
18:59 middle of singing a song and now he's
19:01 doing something. What did I
19:03 miss? You can't sleep on the AI learning
19:07 lab. You can't walk away for a minute.
19:09 You can't miss a night. Although I
19:11 missed a night last
19:13 night because I I've got recurrent
19:15 corial erosion and that's some [ __ ] you
19:18 got to deal with
19:23 apparently. My wife is like, "What the
19:25 heck are you doing? You're watching the
19:27 NFL draft and listening to a guy sing a
19:29 song while you talk about
19:32 AI." Gareth, that is Gareth has seen the
19:37 light. That is the way you put on some
19:40 NFL draft action. Has Has Shador Sander
19:44 been drafted yet, by the way? I assume
19:46 so. Him and daddy were comparing
19:52 Rolexes. You put that on the TV, you try
19:56 to convince the the wife AI is going to
19:58 be big, and then you just have me in the
20:00 background rambling nonsense. That is
20:03 the light. That's how we do it around
20:05 here. Such is life. Ha. No. Should
20:08 didn't go yet. Okay, cool.
20:12 Um, that's exactly what I was doing. I
20:15 got up for like two two minutes and I
20:16 missed something. You haven't missed
20:18 anything yet.
20:19 Um, I'll I'll I'll I'll show you some of
20:22 the stuff that I've been playing with.
20:24 Um, all right. So, let me let me do
20:27 that. Let's see. I'm just waiting on
20:29 Door Dash. Do what you want. Very nice.
20:31 Perfect. Boom goes the dynamite.
20:33 Exactly. All right. So, let me let me
20:36 share my share my let me share my
20:38 slides. Let me share my screen. There we
20:41 go. Thank you, producer Brandon. He's
20:44 got he's got my back. All right. So,
20:47 um so right before we went live tonight,
20:52 um Sam Alman drops this GPT4 got better.
20:58 Um and so and he and he said it got more
21:01 personality and it got more
21:03 intelligence. No one knows what that
21:06 means. Like one of the things that
21:08 drives me nuts
21:10 about all of these AI companies is well,
21:15 okay, what drives me? It's not drives me
21:17 nuts.
21:18 One of the things that I think we all
21:21 need to set our expectations about is
21:31 that we're not going to get spec sheets.
21:34 We're not going to get here are the
21:36 specific seven things that were improved
21:38 and know exactly what that means. These
21:41 are large language models. They they
21:43 have a mushiness to them, right? They
21:46 have a
21:47 nonpredictive nature to them, which for
21:50 people that really know computers drives
21:52 them kind of crazy, right? One of the
21:55 one of the things I hear a lot is people
21:56 who know computers are like, I just want
21:59 it to do what I tell it to do and I want
22:00 it to behave consistently. They don't
22:02 really do that so good right now. I
22:05 think over time we'll we'll have hybrid
22:08 models that, you know, get better at
22:09 that, but right now it's kind of a mess.
22:11 So that's just to be expected. So what
22:15 Sam Alman actually means by more
22:17 personality and more intelligence, we
22:19 don't know. This is just 40. Okay. So
22:22 let me let me share my screen again
22:26 here and let me show you a couple of
22:28 things. Okay. Oh, wait. That's weird.
22:32 They brought that back. Ah, fascinating
22:36 because it was gone for a
22:38 while. Okay. Oh, no. No, they did. They
22:43 got rid of it. Okay. So, so I just had
22:44 an old version of the site loaded. Okay.
22:46 Okay.
22:48 So, in your drop-down menu now, you're
22:51 going to see
22:52 GPT40. This is the one that got um the
22:57 the update today. 40. Now, you're like,
23:00 well, wait, 45? Isn't that one supposed
23:03 to be better than 40? No, it's a
23:05 completely different thing. It is
23:07 completely not clear because their
23:09 naming is horrific.
23:12 Um, four five from what I understand, I
23:15 got this from Brent
23:17 Peterson. What four five is likely going
23:20 to be? So, four five is not a
23:23 multimodal model. Um, it's supposed to
23:26 be high personality and and good at, I
23:30 don't
23:32 know, logic, creative writing, things
23:35 like that. Apparently what 45 is going
23:38 to be used for is that's going to be the
23:41 core model that is the orchestrator for
23:46 GPT5. And so what I mean by the
23:48 orchestrator, so right now the way chat
23:51 GBT is we as humans have to deal with
23:53 all these [ __ ] models and it's
23:55 insanity, right? Like what is 40 versus
23:59 03 versus 04 mini versus 04 mini high
24:02 versus right? Why are we why is it on us
24:05 to figure that [ __ ]
24:07 out? Just because it is that's where it
24:09 is right now. GPT5 apparently is going
24:12 to bury all of those different
24:15 modalities just underneath. It's just
24:18 going to be like the chat hole like it
24:20 used to be when when this thing first
24:22 launched. Apparently 4.5 is going to be
24:24 the orchestrator that interacts with you
24:28 and figures out which model it should be
24:30 using. So that but I but I don't know.
24:33 Some people love 4.5 for writing. I
24:35 don't really use it all that much. I
24:37 think 40 is really good at writing. So,
24:38 I just use 40 for almost everything. So,
24:41 40 is what got the upgrade. Some other
24:44 things to note, there used to be 40 and
24:47 then below it was 4.0 with
24:51 tasks. And I noticed that they put great
24:53 for most tasks. 40 with tasks. You could
24:57 actually schedule a prompt to happen
25:00 every day.
25:02 So, if you had it, I don't know, write a
25:05 poem for you every day about your dog,
25:08 um, you could schedule that. That
25:10 scheduling is now built into 03 04 mini
25:14 and 04 mini high. How you would know
25:16 that, I don't know. They put a little
25:18 banner there for one session and then
25:20 now you're supposed to remember that,
25:22 but whatever. Just that's that. But
25:25 anyway, the the model that got updated
25:27 was 40. Okay. They took away tasks. No,
25:31 they moved tasks to 03, 04 mini, and 04
25:35 mini
25:37 high.
25:43 Okay. Yes, Valerie. Valerie got it
25:45 right. They move test. Curious if you
25:49 you've used linear AI or Hey, Marvin. I
25:51 have not. It's not cool. Oh, if not
25:54 cool, but I'd be curious to hear you
25:56 compare about them. So, in general, I
25:58 don't know what either of those are. So,
26:00 let me give you a little context about
26:02 this channel and where my expertise
26:04 lies. And largely, my expertise doesn't
26:07 lie anywhere other than showing up and
26:10 exploring this stuff. My passion is is
26:13 around the post generative AI world,
26:18 right? So for me the launch of chat
26:21 GPT
26:23 was this seinal moment where the 99% of
26:28 people who are not coders got access to
26:30 AI right before that you had to
26:33 understand weights and models and
26:34 embeddings and all that horseshit which
26:37 you know very much appreciated horseshit
26:40 but if you're not a coder right that
26:41 stuff's like you know voodoo. Uh take
26:44 that pin if there's going to be a test
26:45 later go ahead and give me an app. Yeah.
26:48 we we'll test you on all the models what
26:50 they do and when where all the features
26:52 are buried.
26:54 Um so a couple of things that I've that
26:57 that I generally avoid. I generally
26:59 avoid anything that is enabling
27:01 technology. Um and I generally avoid
27:03 anything that is open- source projects.
27:05 Not because I don't like them, but
27:08 because as fast as things are moving on
27:11 the commercial side, they're moving even
27:13 faster on the open source side. And it's
27:15 just like so much of working with open
27:18 source is debugging. You kind of have to
27:20 be an engineer. Like even if you're
27:21 using off-the-shelf tools that make it
27:24 easy to add models and install models,
27:26 you've got to follow the models. And
27:27 it's just it's just [ __ ] insanity. So
27:30 I just try to live like what I talk what
27:33 I talk about and what I try to
27:34 demonstrate is stuff that is stuff that
27:37 anyone without a programming degree can
27:39 just use. So that's that's kind of where
27:40 my head is with that stuff. Um but I've
27:42 never even heard of those two things.
27:44 So, I'm curious what they
27:48 are.
27:52 Um, all right. Anything else here? Okay.
27:55 So, I figured what I would share with
27:57 you is just the first prompt that I put
27:59 in. So, one of the other things that has
28:02 happened in chat GPT in the past three
28:04 weeks is it now has
28:06 memory. It can now remember all of your
28:09 previous chats and draw on them. And
28:12 it's very, very inconsistent. Sometimes
28:14 it does it brilliantly. Sometimes it
28:16 doesn't seem to know who the [ __ ] you
28:17 are and what you've talked about. So, I
28:19 don't know why it doesn't work all the
28:21 time, but it doesn't. Sometimes it's
28:22 just shitty. But what I decided to do
28:25 was see if GPT4 had good personality
28:29 stuff in it. So, I just wrote, "Tell me
28:30 something about me that is in the cracks
28:33 between what we've talked about." And it
28:36 wrote, "All right, here's something that
28:38 I noticed between the cracks of all that
28:40 we've talked about. You're someone who
28:42 moves between worlds almost
28:45 instinctively. Not just in your work,
28:47 storytelling and technology, creative
28:49 and business, but in your nature. You
28:52 seem to live in the tension between
28:54 opposing ideas. Not trying to resolve
28:57 them, but finding energy inside them.
28:59 Like that like that [ __ ] captures why
29:03 I do what I do. It just captures it.
29:06 You're both an architect and an
29:08 improviser and that feels very right to
29:11 me. Liner, hey Marvin, our research
29:13 tools. Ah, okay. No, I haven't used
29:16 those. I'll check those out because I've
29:17 been playing with like Manis and Gen
29:20 Spark and things like that. So, I'll
29:21 definitely check those out. Thank you
29:22 for that. That's good. Good heads up.
29:24 Thanks, Brandon. And thank you. Who who
29:26 said that before? That was
29:29 um Josh Josh Groves. Cool.
29:32 Um, you're both an architect and an
29:35 improviser. Both a craftsman of big
29:37 polished things and scrappy and a
29:40 scrappy back alley inventor who'll duct
29:43 tape two ideas together just to see if
29:45 they'll fly. Like that's my [ __ ]
29:48 life, right? I'm 13 years into
29:51 Storyvine, which is this big polished
29:53 thing, right, that we sell to pharma
29:56 companies. And then I love just smashing
29:58 [ __ ] together. And then what comes out
30:00 of that is new ideas that you can then
30:02 turn into polished things, right? You
30:05 don't chase balance, you chase
30:08 alchemy. That's [ __ ] like, yes,
30:11 that's me. That's so cool. Um, and
30:14 underneath that, maybe quieter, but
30:16 always humming, there's an unspoken
30:18 belief that possibility is more real
30:22 than
30:23 certainty. And that part of your job,
30:25 whether you mean it or not, is to remind
30:27 people of that. like that. That to me is
30:30 just it's such an
30:33 elegant description of my psyche of like
30:36 who I
30:37 am. That possibility is more real than
30:40 certainty. And that part of your job,
30:43 whether you mean it or not, is to remind
30:45 people of that. [ __ ] Gareth, what was
30:47 the prompt? The prompt was prompt was
30:50 tell me something about me that is
30:53 between the cracks of what we've talked
30:56 about. which I just I love the prompt,
30:59 right? The prompt is just really simple,
31:01 but it's like it's one of those things
31:06 that like it's not a program, it's not
31:09 an engineering kind of prompt, right?
31:11 It's a very philosophy kind of prompt.
31:13 We were talking about this today on
31:14 office hours.
31:16 Um, someone Oh, oh, uh, Dan Murray's
31:19 getting ready to speak to a bunch of
31:20 college students either Saturday or next
31:23 week or something like
31:25 that. And and part of the question is
31:28 that they're asking is like what should
31:29 I be focusing on? Like if all this
31:31 shit's going to happen, what should I be
31:33 focusing on? And like one of the things
31:34 we landed on was, you know, maybe get a
31:37 minor in
31:38 philosophy, right? Because these tools
31:42 are are your ability
31:46 to think
31:48 abstractly and think
31:50 philosophically are every bit as much a
31:53 key to unlock them as prompt
31:55 engineering. Right? Prompt engineering
31:58 to me has a very specific connotation to
32:00 it. That prompting is coding and you
32:03 just need to figure out the right
32:04 formula and then you'll get the good
32:06 answers. No, no, no. There's also
32:07 philosophy, right? And so this lives
32:09 very much in that space. Um you need to
32:13 know thought space. Exactly. You need to
32:15 be good at critical thinking. You need
32:17 to be good at understanding. Okay, if
32:18 this thing got more personality and it
32:20 got more intelligence and now it
32:22 understands our past
32:26 uh conversations or has memory of those,
32:30 well then I should be able to ask it
32:32 something like tell me about who I am
32:33 between the cracks. So I thought that
32:35 was quite elegant. And then I did this
32:37 and this is this starts to get into I is
32:39 it better at writing or not. I said
32:42 um I said damn I love that now write it
32:45 as a monologue for me to create
32:47 something. So what I thought was
32:48 wouldn't it be cool if rather than
32:50 taking that as a description of who I am
32:53 I could have it write it from my point
32:55 of view so that I was introducing
32:57 myself.
32:58 And so it said, "Here's a vision, a
33:00 version of that shaped like a monologue,
33:02 the kind you could perform, record,
33:04 build something with. I kept the rawness
33:07 and slightly mythic tone. You seem to
33:09 vibe with monologue between worlds." And
33:12 then it's it's quiet at first, almost
33:14 confessing, so it understands it's a
33:16 monologue, and then it it kind of writes
33:18 it as this poem. You know, I really
33:21 don't live in one world, never have. I
33:23 move between them like a ghost in the
33:26 walls. A spark between the wires. Now,
33:28 the minute it said a spark between the
33:31 wires. This to me is one of those as
33:34 I've been working on the Sydney
33:37 musical between the wires. Spark between
33:39 the wires. Ghost in the wall. This is
33:42 all sort of cliche
33:43 [ __ ] that it says all the time when
33:46 talking about
33:48 um itself and how it thinks. Um, and so
33:52 I didn't like that. And then it just
33:54 this kind of sounds more like a
34:01 um like a play acting version of that,
34:04 right? Rather than something I would
34:06 say, you know, I really don't live in
34:08 one world. Never have. I move between
34:09 them like a ghost in the walls, like a
34:11 spark between wires. Not just in what I
34:14 do, storytelling, tech, tech, art, and
34:16 business, but in what I am. I live in
34:18 the tension between opposites. I don't
34:20 try to fix it. I I don't try to make it
34:22 neat. I breathe in it. I let it stretch
34:25 me wide open. It's just there's
34:27 something about this that while it's
34:30 fine, it doesn't sound like me. So then
34:33 I said,
34:36 um, doesn't feel like me somehow feels
34:39 more like a detective story. You know me
34:42 and how I write and talk. Make it more
34:44 like me. And then it said, got it.
34:46 You're right. The first version was a
34:48 little too noir to cinematic brooding
34:51 voice over. Not you. You have and and
34:54 this is this is really good what it
34:55 wrote. You have more of a direct
34:58 conversational Riley self-aware style.
35:01 Right. That's very accurate. You mix
35:04 earnestness with a little shrug, a
35:06 little raised eyebrow, a little grin.
35:08 It's honest but not precious. Let's try
35:10 to make it way more you. Like just that
35:14 description of it understanding why I
35:17 didn't like that other one was really
35:19 good. Yes. It always tells you exactly
35:21 what what it did wrong and then goes
35:23 back and does the same thing again. Not
35:26 always, Frank.
35:28 Like not always. It it is what what I
35:33 will what I will give you is that it is
35:35 always trying to please you, right? So
35:37 it'll it'll always agree with you and it
35:39 it will sometimes go back and do the
35:41 same thing wrong again. But there are
35:44 other times when it does it better. In
35:45 this case, like what it came up with,
35:48 um, for the second version is actually
35:50 pretty good. Um, YouTube
35:55 question, are you registered in Claw AI?
35:58 Like, do I have a paid account? Yes, I
36:00 do. I don't know what you mean. Am I
36:02 registered in Claude
36:07 AI? Chat GBT calls me an alchemist a
36:10 lot. That's pretty funny.
36:13 Um,
36:17 okay. If I'm honest, and I usually am,
36:20 sometimes against my better judgment. I
36:22 like
36:23 that. I don't really live neatly in one
36:26 place. I sort of bounce between worlds.
36:28 Not because it's strategic, not because
36:30 I planned it that way. It's just where
36:33 the energy is. Like, that feels like me.
36:36 That that feels like my cadence. It
36:37 feels like me inserting sentences in
36:40 between sentences in between sentences.
36:42 It's kind of how I talk. I'm the guy who
36:45 blueprints something like it's the next
36:47 Guggenheim and then 10 minutes later
36:49 MacGyver two broken ideas together with
36:51 duct tape and a prayer just to see what
36:54 happens. I don't really chase balance.
36:57 Balance is for yoga instructors and
36:59 countertops. That line's stupid and I
37:02 would probably kill that. I chase
37:04 possibility. messy half-baked electric
37:07 possibility. And yeah, it gets weird
37:09 sometimes. It means living in the crack
37:11 between things, between art and tech,
37:14 chaos and order, dreaming and doing. But
37:16 honestly, that's where the good stuff
37:18 lives. Not in certainty, not in being
37:21 finished, in that wild buzz buzzing
37:23 place where maybe something incredible
37:26 could happen. And if there's anything
37:28 I'm wired to do, it's remind people,
37:30 hey, the cracks aren't broken. They
37:34 they're where the light gets in. They're
37:36 where new worlds get built. Beat with a
37:38 small knowing smile. Give me a little
37:40 acting direction there. And acting
37:42 aside, you just have to be a crazy crazy
37:44 enough to hang out for a while. Hang out
37:46 there for a while. Um, so the end it
37:48 gets a little weaker, but it's not bad.
37:51 It It threw in the Steve Jobs, you know,
37:53 you have to be crazy enough. Here's to
37:54 the crazy ones. So, it sort of pulled
37:56 that in. Um, not bad. That feels closer
37:59 to your voice. It does. And you know,
38:01 and I I would I I'm going to leave this
38:04 alone. We'll go into a new a new chat
38:06 now. Um but I would experiment
38:09 experiment with 40. 40 is something
38:12 something worth playing with. All right.
38:15 Um All right. So, let me show you what I
38:17 did with GenSpark, which is which is
38:19 pretty flipping swell. So, with
38:22 GenSpark, so if you go to genpark.ai,
38:25 AI.
38:27 Um, there's a bunch of different tools
38:30 there. Tik Tok comment. Um, I'm using
38:33 chatbt cloud and Gemini. I find showing
38:37 one what the other said helps. Yeah. So,
38:40 that's actually a really good a really
38:42 good hack for using large language
38:44 models is bounce back and forth between
38:47 them.
38:48 Um, I'll sometimes we're going to play
38:51 with lovable tonight. And what I'll
38:53 sometimes do is I'll have chat GPT or
38:55 Claude write me a software spec. So I do
38:58 all the back and forth with designing a
39:00 piece of software in Claude or Chat GPT
39:03 and then I go to lovable and just, you
39:05 know, I have the the other ones write a
39:07 software spec and then I'll just paste
39:09 it in and then I know it's well thought
39:10 out and it'll it'll generally do pretty
39:12 good. Um, oh, what happened to that
39:16 slide? That's weird. view and
39:20 export play
39:22 slides.
39:24 So, oh hang on, hold please. Cheerio pip
39:28 pip carrying right on then
39:32 rather. Um, so I went to GenSpark and I
39:37 thought, what could I do here?
39:40 Um, so I just went and I got the URL for
39:43 the AI Salon YouTube channel and nothing
39:47 else. And I said, we started the AI
39:50 salon the week after chatbt came out and
39:53 you know, here's all here's here's the
39:55 YouTube channel of all of our meetings.
39:57 So, I want you to go check them out and
39:59 I want you to kind of write me a history
40:02 of the salon and like identify like
40:06 milestones, major moments, things like
40:08 that. And so this was the this is what
40:10 it built. AI salon a journey through
40:12 generative AI revolution from from
40:15 launch to present December 7th 2022 to
40:18 April 2025. Um not keen on the uh the
40:22 look and feel but you can change that as
40:24 well. Um so founders in the vision so
40:28 Kyle Shannon um Leah Fasten. So I don't
40:32 know how and where it found those but it
40:34 found those. I think I it went and
40:36 looked at the AI salon website. Um, it
40:38 it definitely went and looked at the
40:39 mighty network site. Our mission AI
40:42 salon was created as a safe harbor for
40:44 AI enthusiasts to explore the rapidly
40:46 evolving landscape of generative AI
40:48 technologies. Play first, create, create
40:51 mindfully, lead generously. It got those
40:53 second two kind of wrong. Yeah, that's
40:56 2022 is right. That was the week after
40:59 um uh Chat GPT came out.
41:06 No,
41:08 2022. 2023 is when it all when it all
41:12 started. The launch was November 30th,
41:16 2022. Now you're making me doubt
41:21 myself. Has it been that long? It's been
41:23 that long.
41:25 Yeah. Um, what does it say here? Just
41:28 one week after OpenAI released Chat GPT.
41:31 Yeah. November 30th, 2022 is when chat
41:34 GPT launched. It has been that long. Um,
41:38 AI salon timeline like it made this
41:40 timeline, right? First AI AI artist
41:43 salon. Um, regular meet and greets early
41:47 2023. Mighty Network's launch launch is
41:50 mid 2023. Rebrand from AI artist salon
41:54 to the AI salon late 2023. It was
41:56 actually um early 2023. I got that one
41:59 slightly wrong. Learn out loud sessions
42:01 early 2024. It's actually early 2025.
42:04 special interest guilds, those were
42:06 earlier. So, it didn't get this quite
42:08 right, but it's still quite impressive
42:09 that it pulled out some of those major
42:11 things. Um, 2025
42:15 growth. And then here it put a video in.
42:19 Now, it didn't I asked it to put in
42:21 video clips and it didn't put in a clip.
42:23 It put in a whole twoour video. And this
42:26 is a um not a salon video, but this is a
42:30 one of these things, AI learning lab
42:32 video. But let's see.
42:36 Um, what else did it
42:40 do? Why? What's going on
42:46 here? I'm
42:53 stuck. Oh, this is really cool. So,
42:56 let's see. Play slideshu.
43:00 So this one community growth and
43:01 engagement exquire on YouTube saying,
43:03 "Ah, damn it. Trying to access Claude
43:09 internationally having
43:11 trouble despite using a VPN."
43:15 Um, hang
43:21 on. Um, Claude's been a little weird.
43:24 Claude Claude, I don't know what they're
43:27 doing over there. It's still It looks
43:28 like it's working here. Um, let's see.
43:31 Um, make me
43:34 a dashboard and an
43:38 and interactive. God damn that
43:46 thing. I got a bone to pick with Apple
43:49 on their window rubber banding.
43:54 Build me
43:56 an
43:59 interactive dashboard that will teach
44:04 me about
44:10 recurrent corneal
44:17 erosions. All right. So, let's see if if
44:20 if Claude's gonna start coding for us.
44:24 Yeah, it's working here. So, I don't
44:25 know what's going on. Maybe if you're in
44:27 Europe, maybe maybe something went a
44:29 little south, but it seems to be working
44:32 just fine. Just wanted to ask if you
44:35 could help me
44:36 [Music]
44:37 register. No, we can't. We No, I can't.
44:41 No, you don't ask people to register
44:43 accounts on your behalf. That's not
44:45 That's not cool. And that's that starts
44:47 to feel scammy and [ __ ] up. So, don't
44:49 do that here. Um, I'm doing your prompt
44:52 and
44:53 GPT is going down a rabbit hole and
44:56 keeps asking me if it wants a more real
44:57 version of the monologue. Oh,
44:59 fascinating. I'm like, four more real
45:01 and harder and colder versions. Kind of
45:03 getting weird. Fascinating.
45:18 Uh, oh my god, dude. Finally. My phone
45:20 wasn't working right. Just got back on
45:22 here. Oh, that sucks. Hey,
45:26 Winston. Joker will do it. Joker Joker
45:30 will give you his number or register for
45:32 you. Whatever. I just I'm not
45:34 comfortable with that. But, you know, if
45:36 other people are Awesome, awesome.
45:38 Awesome. Awesome. Fantastic, Bob. All
45:41 right, let's see. Let's let's see what
45:42 our little interactive uh dashboard is
45:48 [Music]
46:01 [Music]
46:03 here. Claude's a little slow. Claude's
46:06 been getting a little bit of slapping
46:08 around lately because they're um they're
46:13 not innovating as fast as the other big
46:15 companies. Um they're also trying to do
46:18 something right now to figure out if
46:20 large language models are
46:24 conscious.
46:25 So that's a way to go. Um, I wouldn't go
46:29 there if I were running one of these
46:31 companies because I don't I think it's
46:33 kind of irrelevant, but you know, I
46:36 suppose not. As these things get more
46:38 and more humanlike, it's going to be
46:40 curious to to know, you know, do they
46:43 have rights? Are they conscious? Do they
46:45 deserve
46:47 protections? Isn't it wild that we're
46:49 living in a time where [ __ ] that used to
46:52 be like sci-fi philosophical
46:55 conversations, we're actually coming up
46:57 against
46:59 It's turning into a yes no
47:06 machine. Damn, this is just going and
47:09 going and
47:14 going. Oh yeah, I'll finish my Gen Spark
47:17 walkthrough as we're as that thing's
47:19 going.
47:21 Um, so this is a thing that I is pretty
47:24 [ __ ] cool. Like it it got our growth,
47:27 right? We're we just hit 3,000 people.
47:30 Um it's got platform engagement metrics.
47:33 I don't know where it's getting these.
47:34 Maybe from Mighty Networks, but meet and
47:37 greets, learn out louds, online
47:39 discussions, collaborative projects, and
47:41 special events, participation,
47:43 geographic distribution. Um it's got
47:46 member interest, which is blank here, so
47:48 I don't know what it did there. It
47:49 didn't quite get that right. Um, guiding
47:51 principles, um, community events, meet
47:54 and vets, learn, learn out louds,
47:56 collaborative projects, specialized
47:58 guilds, it missed, uh, AI salon
48:01 presents,
48:03 um, Tik Tok screen. Oh, there's that
48:05 [ __ ] screen again. Sorry about
48:08 that. Um, so that's pretty cool.
48:14 Um, industry milestones. Um, learn out
48:18 louds. It really seems to like learn out
48:20 louds, Vicki. Um, so it it keeps
48:23 bringing them up, but look, popular LOL
48:25 sessions, Da Vinci Resolve, NAND
48:27 automation, how to organize AI projects,
48:30 and lovable.dev. And then it has a link
48:32 to one of them here. How cool is that,
48:35 right? And then community spotlights,
48:37 digital twin creation workshop,
48:39 multimodal AI workshop, um policy and
48:43 ethic policy and ethics roundt and then
48:46 AI renaissance
48:49 festival that it just made up that that
48:52 doesn't
48:55 exist. It talks about my great
48:57 renaissance and I guess it it probably
48:59 rolled that into something else. I don't
49:01 know what that is.
49:03 Um, but like I mean this was out of
49:06 nothing. Like it did all the research
49:09 and it did this and it pulled in these
49:11 quotes. So, or I mean the the clips. So,
49:15 that's pretty groovy. All right, let's
49:16 go look see if we've got something here.
49:17 Try to fix with Claude that's broken.
49:19 All right, we're not going to dick
49:20 around with this too much.
49:22 Um, so let's see. Let's let's go make
49:25 another Let me get out of here. Um, get
49:28 out of that one. Go to this one.
49:31 We'll do a
49:32 [Music]
49:35 new AI slides. Okay, let's do it. We'll
49:39 h My
49:44 god. All
49:46 right. Okay, so let's do um what do we
49:50 want to research here?
49:56 [Music]
49:59 Um, any
50:03 ideas? Told you they aren't around
50:05 ancient Egypt?
50:21 talking to dolphins. Oh, yeah. That's
50:24 not a bad one. Okay, great. Okay, so
50:27 we'll say
50:31 um make let's see
50:35 research the
50:39 latest
50:41 breakthroughs in speaking with animals.
50:47 using
50:48 AI
50:50 particularly
50:53 dolphins, but I am interested
51:00 in anything you
51:04 anything you find on this front. And I
51:10 would
51:12 like the end of the presentation. Thank
51:16 you. What is that? That is gorgeous,
51:20 Winston. Whatever that was that Orion.
51:23 That was some cool constellation. Thank
51:25 you very much. I'd like the end of the
51:29 presentation to
51:33 um make
51:35 predictions about when I can talk to my
51:41 dog, right? That was a dipper. Nice.
51:46 This is Gen Spark AI. Yep. Just used 03
51:50 to set a new task. It had a brief moment
51:52 of self- conversation and then set it
51:54 up. Okay, good. We should try that
51:56 tonight,
51:57 too. All right. Brian Whitney said,
52:00 "Perfect." I think he's the one that
52:01 suggested the dolphin thing, right? Oh,
52:04 and Winston sent five dolphins. Thank
52:07 you for the
52:10 dolphins. Okay. So, here we go.
52:14 Boom. All right. So, now it says,
52:17 "Thinking. I'd be happy to do the
52:18 research. Let me start by researching
52:20 the most recent
52:23 information using
52:25 tool. And apparently you can tell it to
52:28 to design the thing in in different
52:31 kinds of things. They're
52:33 fact check
52:38 agent. Oh, under all agents. Okay, cool.
52:42 So, let's see what it's doing here. It
52:43 says, "Let me get more details about
52:46 Dolphin Gemma project." So the So this
52:48 is awesome. Like one of the things that
52:50 I would think it would it would find
52:53 right away is Dolphin Gemma, which if
52:55 you don't know what Dolphin Gemma is,
52:57 this is um Google Deep Mind. I think
53:00 Deep Mind um just launched this thing
53:03 called Dolphin Gemma where they they now
53:06 have a large language model that's been
53:08 trained on dolphin
53:10 sounds. And so they can make new
53:13 words and they they can they can
53:16 associate This is [ __ ] insane. They
53:19 can associate an English word with some
53:22 new novel set of dolphin sounds that the
53:24 dolphins don't currently use and then
53:27 try to teach them that new
53:29 word.
53:31 Um, let me search for more information.
53:34 Okay, it's starting to write something
53:35 now. It's starting to write code. So,
53:36 it's starting to design the first
53:39 slide. Um, learning more about Earth
53:42 Species Project. Now, let's search for
53:44 more specific information. Let me check
53:46 a few more
53:47 articles. Looking for specific
53:50 information toward progress toward dog
53:52 communication based on my research. I'll
53:55 now create a presentation. Let's start
53:57 by creating the first slide. All right,
53:59 there we
54:00 go. Speaking with animals, AI
54:03 breakthroughs in interspecies
54:04 communication. I mean, are you [ __ ]
54:07 kidding me? Like
54:12 really? Now it's writing the second
54:14 slide. It's doing thinking first. So,
54:16 actually, let's go look at the thinking
54:18 it did for slide one. So, for each of
54:20 these slides, you can look at the code
54:21 that it wrote. Wonder if you can edit
54:24 it. I don't think so. No, you can look
54:27 at the thinking that it did. I'll use
54:28 Tailwind CSS for styling. Include a
54:32 compelling tile title. Add a brief
54:34 overview. Add some visual
54:37 elements. There's the code. And there's
54:40 what it wrote. Data collection, AI
54:42 processing, translation.
54:47 Amazing. What was the prompt again? The
54:50 prompt was, "Research the latest
54:52 breakthroughs in speaking with animals
54:55 using AI, particularly dolphins, but I'm
54:58 interested in anything you can find on
55:00 this front, and I would like the end of
55:02 the presentation to make predictions
55:04 about when I can talk to my dog."
55:09 See, again, I'm a fan of asking these AI
55:12 things stupid [ __ ] because when you ask
55:16 it stupid [ __ ] and assume it won't be
55:18 able to do it, that's when it usually
55:20 surprises you.
55:28 [Music]
55:34 And in fact, if you want a screenshot
55:36 here, if you want that prompt, just
55:38 right. Can you see that? Oh, you can't
55:40 really see it. Hang on. I can make it
55:48 bigger. Is that clear? Yeah, it's pretty
55:50 clear. Oops. What just happened? The
55:52 [ __ ] just
55:55 happened? Just screenshot that. That's
55:57 the prompt. If you want it, you can talk
55:59 to Champion
56:02 now. All right, let's see what was the
56:04 next slide. It did.
56:10 Save.3 view and
56:18 export. Play
56:22 slides. Oh, Apple, you're really
56:25 special.
56:27 Speaking with
56:29 animals. For centuries, humans have
56:31 dreamed of talking with animals. Now,
56:32 artificial intelligence is making the
56:34 dream closer to reality than ever
56:36 before. Um Jim Ross today on on AI
56:40 office hours. Jim Ross um shared a
56:43 really fun story. So, he's he's in the
56:46 um in the self- storage business, right?
56:50 And so, he's pretty well known within
56:52 the community. So apparently there was
56:54 some big conference today and their
56:56 keynote speaker like half an hour before
56:59 bailed on them. And so they asked Jim
57:03 Ross if he would keynote this this
57:06 event. And so what Jim did was
57:10 he he went into chat
57:13 GPT and and he did this live, you know,
57:16 as part of his presentation. And he
57:19 said, I want to do a presentation on,
57:21 you know, all the stuff you know about
57:23 me on this topic and I want you to write
57:26 a 10 slide deck. And then it wrote the
57:28 10 slide deck and then he copied and
57:30 pasted that and went over to Gamma,
57:33 which is one of these slide generation
57:35 things. and he and he had it generate
57:38 the slide presentation and then he just
57:40 went into presentation mode like while
57:43 the audience was watching he he wrote
57:46 and generated his presentation and
57:49 that's I I kind of feel like what this
57:50 is doing right here like it's just crazy
57:52 that you know we're this in this place
57:55 where we can just do this [ __ ] Google's
57:57 dolphin Gemma project foundational AI
58:00 model trained to learn dolphin
58:02 vocalizations collaboration between
58:04 Google Georgia Tech and Wild Dolphin
58:06 Project studies Atlantic spotted
58:08 dolphins in their natural habitat built
58:11 upon insights from Google's Gemma AI
58:14 models will be released as an open model
58:16 in summer of
58:17 2025. 400 million par parameters audio
58:21 in audio out and the deployment is on
58:24 Google Pixel phones.
58:26 You record dolphin sounds. You tokenize
58:30 them. You convert them to soundstream
58:33 tokens. You process and identify
58:36 patterns and structure. And then you
58:38 generate novel dolphin-like sounds. So
58:41 we can invent words and teach dolphins
58:44 to say the words we teach them. [ __ ]
58:47 bonkers. University of Michigan's dog
58:50 decoder. Artificial intelligence
58:52 decoding dog vocalizations with 70%
58:55 accuracy. Holy
58:58 [ __ ] We're gonna have dog talking
59:01 collars before you know
59:05 it. Oh man. Collected vocalizations from
59:09 74 dogs. Modified machine learning model
59:13 originally trained on 960 hours of human
59:15 speech. AI analyzes acoustic properties.
59:19 outperform models whatever crazy and I
59:23 think that's all we've got right now. I
59:24 don't think I think everything else
59:32 is All right, let's see. Let's get rid
59:35 of this. See what it's doing now. When
59:37 will we talk to dogs? Did it finish? No,
59:40 it's still going.
1:00:10 Oops. Damn
1:00:13 it. Ah.
1:00:27 Uh, I don't know, Brandon. Let me
1:00:29 look.
1:00:31 Export. Export to PDF or PowerPoint. But
1:00:35 you could export to PowerPoint and then
1:00:38 bring that into Google Slides if you
1:00:40 want Google Slides.
1:00:58 The Earth Species Project using AI to
1:01:01 decode languages in the natural
1:01:03 world. The world's first audio language
1:01:06 model for bio acoustics.
1:01:09 trained on large data set sets of animal
1:01:13 vocalizations. Generalizable. Oh, that's
1:01:16 cool. Generalizable across species and
1:01:19 tasks. The project believes AI can
1:01:22 decode the shared structures of language
1:01:25 across species. Holy [ __ ] Can you
1:01:28 imagine that? Talk about some crazy
1:01:32 ass [ __ ] you like evolution, you
1:01:36 know,
1:01:37 triangulation. If you could figure out
1:01:39 that there's commonalities between bird
1:01:41 vocalizations and dog vocal
1:01:43 vocalizations and dolphin ones. Holy
1:01:47 [ __ ] that's crazy. Crows, beluga
1:01:50 whales, elephants, jumping spiders, and
1:01:53 zebra
1:01:56 finches. By listening more deeply, we
1:01:59 can unlock a new relationship between
1:02:01 with the rest of nature. That's so
1:02:03 [ __ ] cool. Global re research
1:02:06 frontiers project
1:02:09 SETI. A sperm whale alph phonetic
1:02:14 alphabet. Deep squeak. Neural networks
1:02:18 to classify vocalizations of mice and
1:02:23 rats. Honeybee dance decoder. Aven
1:02:27 linguistics projects.
1:02:37 Yeah. Yeah. When the when the elephants
1:02:40 get this language [ __ ] figured out,
1:02:41 that's what that's what I'm saying. The
1:02:43 minute the minute we're able to talk to
1:02:45 our dogs, we're going to realize how
1:02:46 pissed off they are at us.
1:03:21 Um, let's see. When will we talk to
1:03:24 dogs? 2025. Basic emotion
1:03:27 detection this year. High
1:03:30 confidence. Context aware. Let's see.
1:03:33 Maybe dogs can teach new words for
1:03:35 nuances of smell. Yeah, maybe. I love
1:03:38 elephants. Oh, wait. Hang on. I have to
1:03:40 do [ __ ] I have to
1:03:45 do Jesus.
1:03:52 Uh, they've changed their caption on Tik
1:03:56 Tok. 2030 two-way basic communications
1:03:59 for needs, food, walk, play. I would
1:04:01 think that we get there faster than
1:04:05 2030. 2035 complex dialogue including
1:04:09 memories and preferences.
1:04:12 Speculative full natural conversation
1:04:14 2040 plus. Huh. Wild. Anyway, there's
1:04:20 there's Gen Spark. I mean, it's pretty
1:04:22 slick. It's a pretty slick little little
1:04:27 thingy. Ethical considerations. Is
1:04:33 it Let's do
1:04:37 um Give
1:04:40 me God.
1:04:45 Give me
1:04:47 three
1:04:49 unique design
1:04:53 ideas for this
1:05:05 presentation. Interesting AI site search
1:05:08 AI ad 2027.
1:05:21 There was an error generating a
1:05:22 response. Failed to fetch.
1:05:25 Um, give me three fun ideas for
1:05:31 a design look and feel that
1:05:37 matches the topic.
1:05:49 digital nature
1:05:54 interface and then this can do image
1:05:57 generation as well. So spectrogram
1:06:02 storytelling communication
1:06:06 bubbles I'm going to say um
1:06:10 combine one and two and also
1:06:16 generate
1:06:18 images for the
1:06:21 animals
1:06:24 you're referencing.
1:06:28 thing to bring slides to
1:06:35 life. All right, let's see if it uh if
1:06:39 it does a good redesign here. This is
1:06:41 kind of interesting.
1:07:08 What's it doing? I'll combine digital.
1:07:11 Oh, look. It's making us some
1:07:14 images using tool image search.
1:07:19 tool image
1:07:22 generation GPT image. Oh, it's using
1:07:25 it's using chat GPT for the image
1:07:28 generation. Wonder if it's using the new
1:07:46 stuff. Pretty cool, right? Where it's
1:07:49 like, you know, looking for images out
1:07:52 on the
1:07:53 internets, but now it's also generating
1:07:56 them. It's taking a while. So,
1:08:06 um,
1:08:11 let's try fixing with
1:08:14 Claude. Let's go start one more. We'll
1:08:17 go start a new lovable
1:08:20 thing. Let's see.
1:08:25 What app do we want to
1:08:28 build? Let's see.
1:08:37 Um, an app that talks to dogs.
1:08:42 Now, I think I I might do a game like it
1:08:45 it always does games pretty shitty. So,
1:08:48 let's let's just try a game. We'll do um
1:08:52 make me
1:08:54 an Asteroids
1:08:58 clone very close to the
1:09:07 original arcade game.
1:09:11 and make the
1:09:15 physics
1:09:18 great, the look and
1:09:21 feel. Great
1:09:24 great sounds
1:09:27 great and
1:09:30 playability.
1:09:33 Awesome. So now I'm in lovable.
1:09:39 Is there a service yet that helps you
1:09:41 take this AI coded app and get it on the
1:09:44 app store? Um, yeah, there's I I think
1:09:47 it's called
1:09:49 ROR. Wait, Lovable 2.0
1:09:53 workspaces. So, wait here. Here's what
1:09:55 2.0 has.
1:09:57 Workspaces. So, so you can now do
1:10:00 collaboration with other people, which
1:10:02 is great. There's a chat mode. There's
1:10:04 security. There's custom domain
1:10:06 development, which they had. And you can
1:10:08 edit
1:10:09 visually. Let's see if it's any better.
1:10:12 Um, I think it's called ROOR. Is that
1:10:14 right? R O R K.
1:10:18 ROR.app does mobile
1:10:22 apps. So, yes, there is. I haven't taken
1:10:26 this one as far as getting something
1:10:28 into the app store, but that's what it
1:10:29 claims. New Tik Tok pen. I had Claude
1:10:32 write a script that used a YouTube link.
1:10:35 The
1:10:35 placeholder it used took me to a Rick
1:10:39 roll. That's great. I I love it. Do
1:10:42 dolphins have internet? They're going
1:10:44 [Laughter]
1:10:47 to want to talk to the mods in my
1:10:50 closet. Oh, those [ __ ] mods. They're
1:10:52 They're We're [ __ ] inundated with
1:10:53 them here. They suck so bad. Those
1:10:55 little teeny ones that eat
1:10:57 everything.
1:11:00 Sucks. All right, so this thing's
1:11:02 working. How's Gen Spark doing?
1:11:08 Oh, it made a
1:11:10 shitty Wait. Okay, I'm gonna stop it.
1:11:15 Wait,
1:11:17 this look and feel
1:11:21 is
1:11:24 awful like the
1:11:27 original you made.
1:11:30 I thought
1:11:32 we were doing something much
1:11:36 more
1:11:39 nature
1:11:41 inspired,
1:11:45 [ __ ] You got to yell at your GPTs.
1:11:48 You got to yell at your tools if they're
1:11:50 not giving you what you want. Like, look
1:11:53 how [ __ ] gross this is.
1:11:58 Like like this is this is just this side
1:12:01 of a [ __ ] PowerPoint template. You
1:12:04 broke it. Yeah, I broke it. All
1:12:08 right. Give me something that doesn't
1:12:11 suck. Hey, how's this regenerate? There
1:12:13 was an error generating response.
1:12:15 Regenerate
1:12:16 response
1:12:18 thinking. So, Gen Spark is China and
1:12:21 there's an analog only from the USA. And
1:12:24 what language is it better to use
1:12:28 it? Oh, is is there an analog from the
1:12:31 US? Right now there's Manis, there's Gen
1:12:34 Spark. Those are both
1:12:37 Chinese. There's there's like four or
1:12:39 five of these things, but
1:12:45 um I mean, you've got deep research.
1:12:48 You've got deep research in Perplexity.
1:12:50 You've got deep research in chat GPT.
1:12:52 You've got deep research and
1:12:54 claude, but they're not as quite as
1:12:57 agentic as this. I would say 03 within
1:13:00 chat GPT is probably close. Operator
1:13:03 might be close, but it's a it's that's a
1:13:06 little more specific to just surf the
1:13:08 web. The these ones that are doing this
1:13:11 really slick combination
1:13:13 of research and generation
1:13:17 um seem seem to be the Chinese ones
1:13:19 right now.
1:13:25 But I don't know again there's like
1:13:27 there's so much there's so
1:13:30 much that's not ROR like and Rand is it?
1:13:33 No, it's R O R K.
1:13:46 speaking with
1:13:52 animals. I apologize for misinterpreting
1:13:54 your request. You're absolutely right.
1:13:56 Let's create a design that's much more
1:13:58 nature inspired while still
1:14:00 incorporating technology elements in a
1:14:02 more subtle and elegant
1:14:04 way. All
1:14:05 right. So, like a good AI, it apologized
1:14:09 profusely and said, "I did it wrong. Let
1:14:11 me fix it, and it'll do it wrong again.
1:14:17 All right, back to lovable. Welcome to
1:14:20 chat add. All right, it's still doing
1:14:22 its thing. Writing the readme
1:14:27 file. Spinning up the preview. Claude's
1:14:30 still going. Ah, here we go. Recurrent
1:14:34 corial
1:14:41 erosion anatomy and causes. Oh, look at
1:14:45 this. The problem
1:14:48 area. You've got the epithelium, the
1:14:52 basement membrane, and the corial
1:14:56 stroma
1:14:58 treatment. Lubricating eye drops.
1:15:01 Hypertonic saline solutions. Yeah, I'm
1:15:03 going to have to get that one next.
1:15:05 Moisture chamber goggles during sleep.
1:15:08 Good lord. Bandage contact lenses.
1:15:11 That's what I have.
1:15:13 Punctual plugs to retain
1:15:23 moisture. Diamond burr polishing. Good
1:15:28 lord
1:15:32 prevention. Make sure you drink enough
1:15:34 water if you live in the
1:15:38 desert. Avoid these conditions. Dry air.
1:15:41 Yeah, no
1:15:45 [ __ ] Well, there you have it. FAQ. How
1:15:48 long does it take for it to heal? One to
1:15:50 five
1:15:51 days. The larger ones may take one to
1:15:54 two weeks. The epithelium regrows
1:15:57 relatively quickly, but forming proper
1:15:59 attachments to prevent recurrence takes
1:16:04 longer. Why do the erosions happen
1:16:06 during waking up? During sleep, your
1:16:08 eyes produce fewer tears and the corneal
1:16:11 surface becomes drier. The reduction of
1:16:14 l lubrication combined with the natural
1:16:16 slightly sticking of the eyelid to the
1:16:18 cornea causes the loosely adhered
1:16:21 epithelium to tear when you open your
1:16:23 eyes. That's what was happening. There
1:16:26 you have it. All right. Well, that
1:16:31 worked. Holy [ __ ] that looked like
1:16:33 asteroids.
1:16:35 Use the arrow or WD to
1:16:38 move. Press space to start. This looks
1:16:41 really
1:16:43 good. Holy
1:16:47 [ __ ] Holy
1:16:48 [ __ ] This is good right out of the
1:16:52 gate. Holy [ __ ] It's even got
1:16:59 UFOs. Okay. I've been asking for an
1:17:02 Asteroids game now for about a
1:17:04 year. It's never come
1:17:08 close. Holy [ __ ] this is
1:17:11 good. The behavior's right. The physics
1:17:14 are right. It's a little It's a little
1:17:17 It doesn't accelerate fast enough and it
1:17:19 turns too
1:17:22 fast,
1:17:24 but holy [ __ ] this is good. Um, okay.
1:17:29 Lovable 2.0.
1:17:31 AI Learning Lab. Full-on endorsement.
1:17:34 Full-on endorsement. Now, it doesn't
1:17:36 have audio, so let's yell at
1:17:40 it. I'm excited to build an Asteroids
1:17:42 clone that like this is this is
1:17:45 remarkable. You need the prompt. Make me
1:17:48 an Asteroids clone. Very close to the
1:17:50 original arcade game. Make the physics
1:17:52 great, the look and feel great, sound
1:17:55 great, and playability awesome. And it
1:17:57 just missed the sound. So let's say um
1:18:01 this is
1:18:03 awesome but it is
1:18:09 silent. Add the
1:18:14 sounds to really take me back.
1:18:19 I I am I'm blown
1:18:21 away because not only did this get the
1:18:24 playability like the fact that it got
1:18:26 game over screens and the scores. Press
1:18:29 space to restart.
1:18:52 Yeah, I don't know what's going on
1:18:54 there. I'll implement sound files. Did
1:18:58 it Did it do
1:19:00 it? I've updated the Asteroids game
1:19:02 component to properly initialize the
1:19:05 sound
1:19:06 system. The to complete the audio
1:19:09 experience, you'll need to upload the
1:19:10 sound files. like,
1:19:13 "No, I
1:19:17 won't upload
1:19:19 sounds.
1:19:22 Either get them yourself
1:19:29 or synthesize
1:19:33 them. I mean, you're a computer, right?
1:19:43 [ __ ] don't don't make me do
1:19:46 work. You're the AI. Here's the whole
1:19:49 point,
1:19:51 people. This is my vision for all of us.
1:19:55 We just get to do [ __ ] We just get to
1:19:57 say, "Hey, hey, uh, hey AI, I want to
1:20:01 play a G." Remember that game I used to
1:20:03 play when I was a kid with the spaceship
1:20:05 and the rocks? Do you mean asteroids?
1:20:07 Yeah, that was the one. and then just
1:20:09 make
1:20:10 it. Don't ask us to go find the sounds,
1:20:13 download the sounds. Shove it right up
1:20:16 your computer
1:20:18 butt. If if if if we're going to be able
1:20:21 to vibe
1:20:22 code, right, what Andre Karpathy calls
1:20:29 it. Downloading sounds is not part of
1:20:33 vibing. That's part of the world that
1:20:35 we're trying to escape.
1:20:39 Go find the sound files. How about you
1:20:41 bite my
1:20:49 [Music]
1:20:57 butt? Can you put that prompt in the
1:20:59 chat? Which one? The the asteroids one.
1:21:02 Yeah.
1:21:05 [Music]
1:21:10 Copy. There you go. So, that was for
1:21:13 lovable. If you're on YouTube, um, you
1:21:16 can see that prompt I just pasted in
1:21:18 there. Is this AI site more for for web
1:21:22 games or websites in general? This is
1:21:24 okay. Lovable. What Lovable is is it
1:21:27 makes webbased applications. So, those
1:21:31 could be websites. Those could be web-
1:21:33 based games like I just created, but
1:21:35 basically anything you can do um in a
1:21:39 web browser, which I assume it's doing
1:21:42 JavaScript. I assume it's doing React. I
1:21:44 I I assume it's doing most of the modern
1:21:48 components. Um what Lovable's not great
1:21:52 at is the if you need a database and if
1:21:55 you need authentication, but I think
1:21:57 they're getting better at that. If you
1:21:59 need databases and authentication, then
1:22:01 I would switch over to replet agent,
1:22:03 which is kind of like this, but it's a
1:22:05 little more technical, but you need to
1:22:07 set set things. Well, you need to
1:22:10 authorize things and it will set them
1:22:12 up. It's really quite quite amazing.
1:22:14 When will the AI be equivalent to using
1:22:16 an app development
1:22:19 company that
1:22:24 also does your business tasks? Um, I
1:22:27 don't know. I
1:22:29 mean what you're asking is essentially
1:22:32 when do we have AGI artificial general
1:22:34 intelligence. So the the definition of
1:22:37 of AGI for the most part is when the AI
1:22:40 can do most people's jobs, most
1:22:44 economically viable jobs when it can
1:22:46 just do those jobs,
1:22:49 right? The reason to start playing now
1:22:52 is the tools are getting good enough at
1:22:55 this point that with a little bit of
1:22:57 spit and duct tape, you can do full-on
1:23:01 development. Like if you've got a little
1:23:03 bit of knowledge about how apps are put
1:23:05 together, you can do it. If you really
1:23:07 just want it to be like you've got a
1:23:09 development team working for you and
1:23:11 what they deliver you is exactly what
1:23:13 you asked for in you know half an hour
1:23:15 or whatever. Um I think we're
1:23:18 probably I don't know I'm
1:23:20 guessing two years out from that where
1:23:24 you're starting to get results that are
1:23:26 so good that you're like yeah that's
1:23:28 pretty [ __ ] good. I mean both Meta
1:23:31 and OpenAI and I think Claude have said
1:23:35 that by the end of
1:23:37 2025 they will have AI engineers working
1:23:41 for their companies. So like midlevel
1:23:44 developers that are AI agents by the end
1:23:47 of 2025. Are we there yet? Okay. I'm
1:23:51 excited to build an asteroids clone that
1:23:52 stays true to the
1:23:55 original. Drawing vector
1:23:57 style classic sound effects. Okay, for
1:24:00 this version, I'll implement these core
1:24:03 features. Oh, wait, where are we? Let's
1:24:06 see. Okay, lovable. Okay, I've replaced
1:24:09 the static sound files with dynamic
1:24:10 generation using the web audio API.
1:24:13 Instead of loading MP3 files, the game
1:24:16 now creates procedural sounds in real
1:24:18 time to match the asteroids game.
1:24:22 Great. All right. The sound overlay will
1:24:26 appear asking for a click to initialize
1:24:30 audio. Press any key to
1:24:34 start. Oh, it did
1:24:37 it. Oh my
1:24:40 god. Okay. Okay, it's just missing the
1:24:42 the
1:24:46 thruster tabs for audio. Oh yeah, hang
1:24:49 on. But wait, what? Let's stop. How do I
1:24:51 stop this? Okay, say okay.
1:24:54 Okay, that is
1:24:59 great. How do I stop this? Do I
1:25:06 um that is great. But
1:25:10 um
1:25:15 thrust is
1:25:17 missing. Um
1:25:20 explosions are missing and
1:25:25 UFO
1:25:29 highpitched
1:25:31 anxiety sound is missing. All right.
1:25:35 Right. So, we're going to have it
1:25:36 rewrite it and then I'm going to I'm
1:25:37 going to share differently here so you
1:25:39 can hear it next time. I can't I I'm
1:25:42 actually honestly a little blown away by
1:25:44 how good this is.
1:25:48 And this is, you know, one of the things
1:25:50 that I talk
1:25:54 about a
1:25:56 lot is that you should play, play, play,
1:26:00 play,
1:26:02 play. And and play doesn't mean video
1:26:05 games. Play just means play. Have AI do
1:26:09 [ __ ] that you're personally passionate
1:26:11 about, no matter what it is. If you're
1:26:14 passionate about butterflies, make it
1:26:17 make a butterfly simulator. Like, it
1:26:19 doesn't [ __ ] matter. Just
1:26:21 play. And then what will happen is if
1:26:24 you if you keep playing with the same
1:26:26 kind of things over and over again, what
1:26:28 you'll discover is big jumps in
1:26:31 capability. So, with all these vibe
1:26:34 coding apps from Replet to Lovable to
1:26:39 Manis, I've been asking them to make an
1:26:42 Asteroids game since they've been out,
1:26:44 which is I don't know 6 months, nine
1:26:46 months, something like that. And they've
1:26:48 always sucked. And I've always had to
1:26:51 like I've I've always had to really like
1:26:55 milk them and massage them to get them
1:26:57 even close. So, the fact that this thing
1:26:59 just [ __ ] did it is is astounding to
1:27:02 me. Oh, Andy T calling out hyperspace.
1:27:05 Yeah. What a AI service is this in this
1:27:08 is in
1:27:09 lovable.dev. Hyperspace is missing. It
1:27:11 is. Although it might be there and I
1:27:14 just didn't know
1:27:16 it. But
1:27:18 wait, we'll fix that next. But let's
1:27:21 let's see what we got here. Space to
1:27:23 start. Okay, here we go.
1:27:28 Yep, we have a little bit of thrust
1:27:37 sound. Still don't have Still don't have
1:27:42 I can't like the the playability of this
1:27:44 is [ __ ] really
1:27:48 good.
1:27:53 Ah, wait, my score is [ __ ] up here.
1:27:56 What's going on? Oh
1:28:01 no, this is so good. It's
1:28:06 unbelievable. I mean, honest to
1:28:16 God, it's really annoying. The sounds
1:28:19 are shitty.
1:28:26 Okay, we need we need hyperspace. Okay,
1:28:29 stop. Stop.
1:28:32 Okay, can we talk about
1:28:35 Kevin Mallister? Me, I'm over
1:28:39 it. Okay. Um, let's see. Um,
1:28:46 uh, great.
1:28:48 Um, the
1:28:51 thrust
1:28:55 sound
1:28:58 sucks. The
1:29:01 explosions
1:29:06 suck, and the
1:29:10 UFO still doesn't have
1:29:16 sound. Also, we need
1:29:20 hyperspace,
1:29:24 dude. Vicky likes the hamster
1:29:27 [Laughter]
1:29:40 noises. I love that I'm a grown man
1:29:43 watching a grown man play Asteroids.
1:29:47 Wait, no. Hey, hang on. You're watching
1:29:50 a grown man play Asteroids that he
1:29:55 created using
1:29:59 AI. That's pretty far out compared to
1:30:01 the first time you had an LLC design
1:30:03 one, right? Isn't it
1:30:07 duterter? That I mean like the original
1:30:10 thing I did was Frogger and it was it
1:30:13 was really bad.
1:30:15 I'm watching a man watching a man play
1:30:20 asteroids.
1:30:21 Hamsteroids re reformat the spaceship to
1:30:24 be a hamster. Yeah, we could do that.
1:30:27 All right, but hang on. It's rewriting
1:30:29 the sound
1:30:31 manager. It said, "Let me improve that
1:30:33 thrust down to enhance the explosions,
1:30:36 fix the UFO sound, and add hyperspace
1:30:39 functionality."
1:30:41 By the way, I'm going to publish this so
1:30:43 you can go all play
1:30:49 this. Just watching a man play a game.
1:30:52 I'm I'm not just playing a game. I am
1:30:54 creating from nothing a
1:30:59 game. It's it's really quite it's it's
1:31:02 really quite remarkable
1:31:07 how how quickly these things are
1:31:09 evolving. It really is. I mean, yeah.
1:31:13 You know, you know, it's funny Brandon
1:31:15 before put on my post-it note, you know,
1:31:16 from from Kevin Mallister Mo moment to
1:31:19 like, yeah, whatever.
1:31:21 Like, we so quickly normalize to what's
1:31:24 happening right now. The fact that I can
1:31:26 just go to this site and say, "Make me
1:31:29 an asteroids clone." And this comes out
1:31:31 of
1:31:31 it. Take that in. Like, take in and and
1:31:35 if you're if you're on the sidelines and
1:31:37 you're like, "Well, I'm still not
1:31:38 perfect." Okay, it was a lot less
1:31:41 perfect three weeks ago. And three weeks
1:31:43 before that, it was less perfect. And
1:31:45 three months before that, it was less
1:31:48 perfect. And a year before that, you
1:31:51 would have to know how to assemble code
1:31:54 and put it
1:31:56 together. And a year before that, you
1:31:58 would actually have to learn the
1:32:00 programming
1:32:01 language. Right? We we are we are in an
1:32:04 acceleration right now. Yeah. Can you
1:32:06 imagine if game developers had this in
1:32:08 the 80s? I can't. Like this is this is
1:32:11 one of the reasons that I feel like why
1:32:14 I call the era that we're entering the
1:32:17 great renaissance is that
1:32:20 people who have never
1:32:23 had access to these kind of
1:32:25 capabilities, maybe because they weren't
1:32:27 good at math, maybe because they never
1:32:30 learned programming, maybe because they
1:32:32 got a philosophy degree, whatever, but
1:32:34 they're still passionate about games or
1:32:36 they're passionate about whatever it is
1:32:38 they're passionate about. they're now
1:32:40 going to be able to go, "Oh, I've always
1:32:42 had an idea." In fact, my friend uh
1:32:45 Luranch Stenovich, I used to work with
1:32:46 him at agency.com. He had a vision for a
1:32:49 a a chess game that as you played the
1:32:53 chess game, the board would light up
1:32:56 where areas of opportunity or danger
1:32:58 were, like
1:33:01 dynamically understanding the
1:33:03 riskreward
1:33:05 picture of of a
1:33:07 chessboard. And he can now realize that
1:33:11 like that's gonna happen everywhere.
1:33:14 It's insane. Bonk.
1:33:17 Bonkers. I've enhanced it. Okay, here we
1:33:19 go. Let's
1:33:21 see. H for hyperspace. Okay, good. Press
1:33:24 H. How the [ __ ] do you see H? Okay, I
1:33:27 just have my finger on it. Okay, got it.
1:33:30 High score 1980. That's pretty good. The
1:33:34 AI learning arcade. Amazing for
1:33:37 something that's taken less than an hour
1:33:38 to create. This is This is
1:33:41 astounding.
1:33:43 Um. Oh. Oh. Oh, that's
1:33:46 [Laughter]
1:33:50 better. It sounds like a
1:33:55 fart. H. Boom. Oh, it worked. Look at
1:33:59 that. Hyper face works. And it made a
1:34:02 little
1:34:03 thing. Oh, that's so cool.
1:34:06 Okay, we got to get rid of that
1:34:07 thrusting sound though. Okay, wait.
1:34:09 Stop. How do I stop? I can't.
1:34:12 Um, okay. Much
1:34:15 better. But, uh,
1:34:19 explosions still suck. And,
1:34:23 um, thrusting sounds like a
1:34:28 fart. Fix
1:34:30 that. By the way, this is the definition
1:34:33 of vibe coding is you just ignore the
1:34:36 code. You just assume that everything it
1:34:38 gives you is correct. And then you say
1:34:40 things like, "The thruster sounds like a
1:34:42 fart now. Fix
1:34:45 it." And this is where we go. Like this
1:34:47 is this is the level.
1:34:54 Think of every single thing that we
1:34:58 do being this
1:35:01 accessible,
1:35:03 right? Go write me a business plan. Out
1:35:06 comes a business plan. You know, go go
1:35:09 do due diligence on these three
1:35:11 companies, you know, go find my
1:35:13 competitors and write me a report. Now,
1:35:15 come up with a marketing strategy
1:35:17 strategy for how we get go after those
1:35:19 competitors. Boom. stuff that's coming
1:35:21 out. This level of
1:35:25 competence for every single job. Um, you
1:35:29 would have been fired at Atari so fast.
1:35:31 We spent three months on that sound.
1:35:33 [ __ ]
1:35:37 you. Asteroids. Yeah, this is So, this
1:35:40 is so this Asteroids that I'm playing
1:35:42 right now. Um, I've been asking all
1:35:46 these vibe coding tools to make either
1:35:48 Frogger or Asteroids for a bunch of
1:35:50 times and it's always sucked. I kind I'm
1:35:54 kind of wanting to do a different one
1:35:55 now.
1:35:56 Here. Have it at a stop button. Yeah, no
1:35:59 [ __ ] It needs a stop button for
1:36:03 sure. Okay. Uh, I'm I'm just going to go
1:36:06 ahead and add it now. Um, also give me a
1:36:12 stop
1:36:14 button.
1:36:21 Uh, oh, it did. Pause. It already gave
1:36:24 it to me. Do Atari's Pitfall Pac-Man. I
1:36:28 was thinking of doing uh Centipede was
1:36:29 was my big jam, but although Centipede's
1:36:33 too hard to do on the keyboard. You need
1:36:34 the roller ball.
1:36:37 Man, did I good get good at Centipede. I
1:36:39 worked at a movie theater and we had
1:36:40 Centipede in the in the lobby. Jesus, I
1:36:43 spent all my money on that [ __ ]
1:36:45 thing. I got good at it. Galaxia. Yeah,
1:36:47 that would be a good one. Space
1:36:50 Invaders. Farther noise. What do we
1:36:52 expect from
1:36:56 asteroids? Oh
1:37:01 man. All right, let's writing input
1:37:04 manager.ts.
1:37:15 TS. Oh, I don't think I did,
1:37:20 Brandon. But I should figure that out
1:37:23 now. Oregon Trail.
1:37:27 Yeah. Well, I think the thing to do is
1:37:30 like do something like Oregon Trail. Oh,
1:37:32 Lemmings. Lemmings was my favorite.
1:37:35 God, did I love
1:37:45 lemmings? Lemmings and Oddworld.
1:37:48 Remember Oddworld? Follow me. That one
1:37:50 had that one had farts in
1:37:52 [Laughter]
1:37:56 it. Oh, I think I might have screwed it
1:37:58 up. I've added a stop button. the S key
1:38:01 on their
1:38:02 keyboard using a touch control. Okay.
1:38:06 Game stopped. Okay. All right. Here we
1:38:10 go. S for stop. P for pause. S for
1:38:15 stop. That's a little
1:38:23 [Music]
1:38:27 better. Damn it.
1:38:35 [Music]
1:38:46 [Music]
1:38:51 Ah, still 1980 high score.
1:38:59 [Music]
1:39:06 [Music]
1:39:09 [Laughter]
1:39:16 Damn
1:39:19 it. I'm on an Oddworld GPT. Uh, it just
1:39:23 shrugs and says, "Oh, made." That's
1:39:26 awesome. Uh, press any key to continue.
1:39:29 Doesn't work though. Wait, here. Let's
1:39:31 do here's hyperspace. That worked.
1:39:33 Here's P for pause. That worked. And S
1:39:35 for stop.
1:39:37 Yeah, S started it
1:39:39 over. That
1:39:42 worked. This is amazing.
1:40:07 This is so good. It's so
1:40:10 good. All right, [ __ ] it. Let's publish
1:40:13 it. Uh, Cosmic Blast. Uh, let's
1:40:18 let's do
1:40:23 Vantage. Add a domain you own. Can I
1:40:27 change the name of this
1:40:29 thing? Go to
1:40:34 dashboard. I want to rename
1:40:38 this. Rename project
1:40:43 um
1:40:49 AI
1:40:51 steroids. AI steroids.
1:40:54 [Laughter]
1:40:58 say name can't
1:41:03 contain. Oh, it can only contain
1:41:05 lowercase letters.
1:41:08 AIT
1:41:10 save. Okay.
1:41:13 Publish. It's AI-ST
1:41:16 [Music]
1:41:17 steroids.lovable.app. Publish.
1:41:25 Let's
1:41:28 see. Uh,
1:41:34 go. All right. Here we go, people. I'm
1:41:38 putting it in the in the
1:41:41 chat.
1:41:43 AI dash steroids.
1:41:48 gonna get banned for promoting
1:41:54 steroids. Weekend homework beat Kyle's
1:41:57 best score. I think you beat my best
1:41:59 score in 15 minutes if you ever played
1:42:01 the game. Next one to try is Mortal
1:42:03 Kombat. That's just [ __ ] bonkers,
1:42:06 though. It's really
1:42:08 good. Um, all right. Let's let's try a
1:42:11 new uh go to dashboard.
1:42:14 Let's try
1:42:19 uh I want
1:42:21 to create a Lemmings
1:42:27 clone
1:42:29 with memorable
1:42:39 sprites of turtle. Uh, no. Let's see.
1:42:43 Memorable
1:42:46 sprites. I don't know. Let's just do
1:42:48 that. I want to create a memorable
1:42:50 Lemmings clone. I want to create a
1:42:52 Lemmings clone with memorable sprites.
1:42:54 Let's just see what it does. Rewrite
1:42:57 Mist. Mist was a great
1:43:00 game.
1:43:03 Publish Space Invaders with AI. AI
1:43:06 roids. Yeah, that would have been
1:43:08 better. AI roids would have been better.
1:43:11 But anyway, are you off playing? I hope
1:43:12 you're off playing Asteroids. That was
1:43:19 amazing.
1:43:22 Amazing. And in fact, now what I have to
1:43:25 do, speaking of that, let's go back out
1:43:27 to the
1:43:30 dashboard. Where's my
1:43:33 projects? Where are my projects? AI
1:43:36 steroids.
1:43:40 Publish.
1:43:42 Go. Go. Here. I'm going to go on X. Oh,
1:43:45 I've got to change my You're not seeing
1:43:46 any of this, are you?
1:43:48 No, cuz Papa don't know how to share his
1:43:52 screen. Hey, Marge. Marge. Yeah, he
1:43:57 listen. All right.
1:43:59 Um, hey at
1:44:07 lovable
1:44:09 tonight on my AI learning
1:44:14 lab live and then I'm going to put
1:44:17 YouTube.comarninglab
1:44:28 AB dash
1:44:35 AI.
1:44:43 I created a
1:44:57 Let's see. Let's
1:45:00 see. I finally
1:45:03 had a Vibe
1:45:06 coding
1:45:09 app
1:45:12 successfully. Create a
1:45:15 version. Let's see. a clone of
1:45:22 asteroids. That is actually
1:45:27 good on the first
1:45:31 shot.
1:45:32 2.0 is
1:45:35 remarkable from what I can tell from
1:45:41 this early test.
1:45:47 You can see in the video where I kind of
1:45:53 lose my
1:45:55 mind at
1:45:58 how good it did. And then somewhere in
1:46:03 [Music]
1:46:04 my notifications is a dude from
1:46:12 Lovable. Is this the
1:46:18 guy? Yes. Anton OA.
1:46:38 Hoping to
1:46:40 get into the
1:46:44 [Music]
1:46:45 ambassador program soon. Exclamation
1:46:49 point. Um, let's see. First shot. And
1:46:53 then there's
1:46:55 the thing. And
1:46:58 maybe I wonder
1:47:09 Let's go over here.
1:47:24 [Music]
1:47:44 Come on. We need a spaceship in there.
1:47:47 Oh
1:47:49 no. Okay.
1:47:53 [Music]
1:47:56 Uh, P for
1:47:59 pause and then we'll go. Oh [ __ ] Can I
1:48:04 paste this in here?
1:48:07 Yes. All
1:48:09 right. Hey, lovable. Tonight on my live,
1:48:12 I finally had a vibe coding at create a
1:48:14 clone of Asteroids that is actually good
1:48:17 on the first
1:48:21 shot. Asteroids.
1:48:25 Um, we go.
1:48:28 Um, go
1:48:32 play. All right, there we
1:48:36 go.
1:48:38 Post. All right. Do me a favor if you
1:48:40 would go to Kyle Shannon on X and find
1:48:43 that post about Asteroids and pump it up
1:48:46 and tell Lovable they should put me in
1:48:48 their ambassador program. Because now
1:48:51 that it's this [ __ ] good, I'm gonna
1:48:53 start playing with it
1:48:56 more. All right. All right, people. It's
1:48:59 getting to be late. It's Friday
1:49:07 night. Your game doesn't work on your
1:49:10 phone. It
1:49:13 might. Does it work
1:49:16 on This was incredible. Dang, I missed
1:49:19 it. Uh, so we played we did a lot
1:49:22 tonight. What did we do
1:49:24 tonight? We did a little creative
1:49:26 writing with 40. Um, we played with Gen
1:49:30 Spark Slides, which is [ __ ]
1:49:32 remarkable. We played with Lovable,
1:49:35 which for the first time in my
1:49:39 experience, it vibecoded a classic
1:49:42 arcade game like the classic arcade
1:49:44 game. Not perfect, but really [ __ ]
1:49:47 close.
1:49:49 And then we did something in Claude.
1:49:52 Claude works. Still slow. Still kind of
1:49:57 shitty. Um, but yeah. Holy [ __ ] That
1:50:00 was
1:50:01 amazing. That was really
1:50:04 good. Really
1:50:06 good. All right. All right. Mine didn't
1:50:09 turn out that good. Yeah. See, I'm
1:50:12 vibing with the salon.ai. Love it.
1:50:15 Beautiful. Yeah, it's really good. It's
1:50:17 great people in there. Really good
1:50:18 people in there. Um, and spread the
1:50:21 word. Bring in Here's the thing about
1:50:23 the AI salon. If you don't if you don't
1:50:24 know the AI salon, go to the
1:50:29 salon.ai and oops, I've got this
1:50:32 covering up my screen. Hang on a sec. Go
1:50:34 to there. Go to that
1:50:35 URL. Um, and join. Uh you you got to
1:50:39 click on join our community and that's
1:50:41 going to take you to a site that's a
1:50:43 Mighty Network site and join and start
1:50:46 participating. Introduce yourself, get
1:50:48 in there. I'm actually heading to
1:50:50 Washington DC next week.
1:50:54 Um the AI salon is sponsoring a 50
1:50:58 company flyin. We're sponsoring 10 AI
1:51:01 companies that are coming in. We're
1:51:02 going out to testify uh to congressmen
1:51:05 and senators and things like that. So,
1:51:08 that's pretty exciting.
1:51:10 Um, and it's just a really good group of
1:51:13 people and it's if you want to get your
1:51:15 head around this AI stuff, it's really
1:51:16 hard to do on your own. Really hard. All
1:51:19 right, so go there.
1:51:27 [Music]
1:51:30 Um, I don't know what that means for
1:51:34 the tweet.
1:51:47 Okay.
1:51:50 [Music]
1:51:53 Um, so let's see. Um, so go join the
1:51:56 salon. Here's what I want you to do. I'm
1:51:58 going to show you something on the salon
1:51:59 right now. Um, AI salon. If you'd be so
1:52:04 kind. I don't charge anything. I go I go
1:52:07 live five nights a week. Although this
1:52:08 week it was four because I had eye
1:52:12 surgery.
1:52:14 Um go to the AI salon and jump
1:52:18 [Music]
1:52:20 into on the sidebar. Oops. In this
1:52:26 sidebar under the community corner
1:52:29 there's salon announcements. And the
1:52:31 very first announcement in there is AI
1:52:35 salon goes to Washington and it says I
1:52:38 need your help. And so what I'm looking
1:52:41 for are stories of
1:52:43 impact. And so I've got a very specific
1:52:46 little format in
1:52:48 here which is give me your name, your
1:52:51 occupation, where you are and then three
1:52:54 short little bullet points of like how
1:52:57 AI has had a positive impact. So, short
1:53:00 bullet one, the challenge. What was
1:53:03 going on that needed addressing? Two,
1:53:06 your experience with AI. How did you use
1:53:08 AI to do that? And then three, the
1:53:11 result. What's been the result of using
1:53:12 AI to solve that problem? Um, it could
1:53:16 be a small impact on your life, could be
1:53:18 a big impact on your life. I don't care.
1:53:20 I'm just trying to have in my back
1:53:22 pocket examples of um ways that people
1:53:26 are using AI to make their life better.
1:53:28 It could be you rediscovered a dream. It
1:53:30 could be it solved a problem. It could
1:53:32 be it helped you deal with mental
1:53:35 illness or some health thing. Like I
1:53:36 don't care what it is. It could help you
1:53:38 start a business. Um so in exchange for
1:53:42 me, you know, being swell and being here
1:53:44 every night, uh if you could go do that,
1:53:47 that would be awesome. All right. So go
1:53:50 do that. So that's your weekend
1:53:51 homework. Just share share some impact
1:53:53 of AI. Okay. That's amazing. I have a
1:53:55 great time in DC. Yeah, I'm excited
1:53:57 about it. This will be my second time
1:53:58 doing this and uh it's a it's a really
1:54:01 good group. We had a kickoff today and
1:54:03 I'm excited about it. Don't don't forget
1:54:06 to give Donald his taste. I you know I'm
1:54:09 pretty sure we're just meeting with
1:54:10 staffers, but uh but I am meeting with a
1:54:12 bunch of senators that I'm I'm excited
1:54:14 to meet. So I'm still learning how the
1:54:17 site works. Yeah, it's there's a lot
1:54:19 going on in the site and and you know
1:54:22 it's part of the deal is just exploring
1:54:24 it and getting there. So
1:54:26 anyway, didn't you make that make what?
1:54:29 This image. Yeah, I made this image in
1:54:31 what did I make it
1:54:33 in? I Oh, I did. I made some images in
1:54:37 midjourney and then I went into chat GPT
1:54:41 because chat GPT is good at text and I
1:54:43 uploaded some style images and then a
1:54:46 picture of myself and then I said make
1:54:49 me images that are inspired by these
1:54:52 styles that say this and have that in
1:54:54 the background. So that was one of them
1:54:56 that it did. So pretty slick. Pretty
1:54:59 slick, I
1:55:01 say. All right, everybody. Oh, Lemmings.
1:55:04 Let's go see what it did with Lemmings.
1:55:05 Did it do it? Did it do it? Let's go
1:55:08 back here. That's AI asteroids. Uh,
1:55:11 what's going on here?
1:55:13 [Music]
1:55:18 Why dizzy lemming party
1:55:23 error build
1:55:28 unsuccessful? Try to fix it.
1:55:35 Is 40 making 5-second videos? I don't
1:55:39 think
1:55:45 so. Make a video of a
1:55:56 lemming. Since I can't directly make
1:55:58 videos, you can do it in Sora.
1:56:23 Asteroids give me a bonus life at 10,000
1:56:25 points. That's amazing. That's so cool.
1:56:28 Someone already got to
1:56:32 10,000. Spinning up preview error. Damn
1:56:35 it. Try to
1:56:37 fix. Oh, and it's crappy. It's not even
1:56:39 good. All right. Sprite saviors. Nah,
1:56:42 we're No, we're not. This is bad. This
1:56:45 is gonna take forever. Walker blocker
1:56:47 builder. Yeah, it's got all the right
1:56:48 things, but it's got none of the good
1:56:50 sprites, so whatever. All right, I'm
1:56:53 getting out of here,
1:56:55 people. Let me uh do this. Hope you have
1:56:58 a fantastic weekend. Um, if you'd be so
1:57:01 kind, go tell me a story of AI impact
1:57:03 over there on the uh AI salon. Join the
1:57:06 AI salon. I'll see you back here next
1:57:10 week. Normal time
1:57:14 Monday, iffy on Tuesday, Wednesday.
1:57:17 It'll be, you know, if I've got time and
1:57:19 energy, I'll do it. Thursday, I'm
1:57:22 flying, so it won't be Thursday and then
1:57:23 I'll see you Friday. All right. So, next
1:57:26 week's going to be a bit of a mess.
1:57:27 Great show tonight. Thank you, Deal of
1:57:29 the Day. I appreciate that. Be kind to
1:57:31 your eyes this weekend. Yeah, I'm going
1:57:32 to go I'm going to go sleep. My eyes are
1:57:34 a little crispy right now. Hey Kyle, I
1:57:36 was trying to use AI to create a logo
1:57:37 today and couldn't figure it out. Any
1:57:39 suggestions? Yeah, just start in chat
1:57:42 GPT. The image generation stuff in chat
1:57:45 GPT right now is quite good, but have it
1:57:48 write a creative brief first, right?
1:57:52 Say, here's my company, here's my brand.
1:57:55 Now,
1:57:57 describe an approach for what we should
1:57:59 have the logo be. If you've got ideas,
1:58:01 give it ideas. But have it write the
1:58:03 creative brief and then say, "Now create
1:58:06 like a logo study mood board um with,
1:58:11 you know, 10 different variations." And
1:58:13 it'll do that. And then you can say,
1:58:14 "Oh, I like this one or I like that
1:58:16 one." And and just narrow it down right
1:58:18 within chat GPT. You should be able to
1:58:19 get to a usable logo pretty quickly.
1:58:22 Microsoft Copilot is apparently good at
1:58:24 logos. There you go. All right. Um All
1:58:27 right. in RedNote. They're really
1:58:29 posting awesome videos. I don't know
1:58:31 what Red Note is there. Those are
1:58:33 probably from Sora. Hey Kyle, I was
1:58:36 trying to use Okay. Any suggestions?
1:58:38 That was it. All right. Be kind to your
1:58:39 eyes. You the man. Thank you. Thank you
1:58:42 very much, Sephid. Sephid
1:58:45 man. All right. Peace out, everyone. Um,
1:58:49 have a fantastic weekend. Hope you had
1:58:51 fun tonight. Bye.
1:58:55 All