4/17/2026 - Mapping the History of Generative AI Through Three Years of the AI Learning Lab

Description

Chapters

Transcript

0:00 some football.
0:09 Don't
0:31 Hello.
0:36 Hello.
0:57 Hey,
1:21 Hi Champy. What you doing? Good boy.
1:23 Friday night date night everybody.
1:25 Friday night date night. What is
1:27 happening? We're going to be doing some
1:29 fun stuff tonight. We are going to be
1:31 playing with the We're going to be
1:33 playing with history.
1:36 So, some of you may know there's this
1:39 fella on the planet.
1:45 His name is Digital Gods.
1:50 And at some point in his life,
1:56 he was lorded.
1:58 So he's lorded digital gods,
2:05 which I think is a demotion, quite
2:06 frankly, right? You're a god and
2:09 someone's like, "Now you're a lord."
2:12 Probably kind of a letown, I would
2:14 think. Anyway
2:26 Oh, cool. So, if you go to community
2:29 corner
2:30 in uh
2:39 in the AI salon, there's actually an
2:41 interview with Lord Digital Gods.
2:43 Anyway,
2:45 here's the deal.
2:47 Mr. Gods,
2:51 he does a lot of work for this channel.
2:54 And one of the things he does is he
2:56 makes the he makes the thumbnails and he
2:59 makes the descriptions and he's been
3:01 geeking out on how to automate that and
3:02 do all that stuff, but it's still a lot
3:04 of work and a lot of it's manual because
3:07 AI still sucks. And so, he just does
3:10 this tirelessly. and I can't thank him
3:13 enough. And what he's created is
3:17 documents of like the history of all of
3:20 the the videos on the channel. So like
3:23 all the descriptions, the chapters,
3:25 things like that.
3:27 So I asked him to provide that for me
3:32 and I put it into Notebook LM. And so
3:35 I've been notebook lmming like the
3:38 history of of generative AI
3:42 um as seen through the lens of you know
3:46 me rambling for for three years. Um and
3:50 it's wild. It it it's it's sort of
3:52 broken it down into five eras. And so I
3:56 figured we'd go play with that. So
3:57 that'll be fun.
4:01 And it was funny. I was asking um uh
4:05 Notebook LM to to do some some clipping
4:09 for me. It had actually Notebook LM
4:11 actually recommended GenSpark Clip Genie
4:15 said go why don't you go have some other
4:17 tool do that. What you're asking me is a
4:19 little much.
4:27 So, so we'll we'll play with with uh the
4:29 Lord Digital Gods document, which is
4:32 really kind of cool.
4:45 Um, and then we're going to go play with
4:48 um Claude Designer. So, Claude came out
4:52 with a design tool.
4:58 which if I were
5:02 someone like Figma,
5:05 I'd be like, "Ah, awesome.
5:21 Um
5:31 uh and it looks pretty slick. You you
5:33 you can upload you can point it to a
5:34 GitHub repository. You can point it to a
5:36 Figma board. You can upload um brands
5:40 brand guidelines. Um you can upload
5:42 logos, things like that. And then you
5:44 just start asking it for [ __ ] and it
5:46 starts building it. And then you can
5:48 download those things as files that you
5:50 can import into other applications that
5:54 are editable. So I think it's a
5:57 relatively big deal. But
6:00 as we discover on this channel,
6:01 sometimes these things look good in
6:04 theory, but are janky and unusable.
6:08 I have a feeling this one's not going to
6:10 be janky and unusable. But you know me,
6:12 I am willing to I'll call it like it is.
6:16 If it is janky and unusable.
6:21 Sky Timbo. Hey, it's been a while.
6:22 What's shaking?
6:26 You and I here all alone.
6:31 Sunday morning here at Ying Yang. What's
6:36 happening? If someone new to AI were
6:39 only going to buy one subscription, what
6:41 would you tell them to buy
6:43 today? Claude.
6:47 Well, no, no, no, no, no. I wouldn't
6:57 one subscription. Well, so here's the
6:59 thing. What are they trying to do?
7:02 Right? If what they're trying to do is
7:04 code and engineer, I would I would buy
7:07 Claude. If you want something more
7:09 rounded that can be really good at
7:11 coding, I would go chat GPT because
7:14 there you've got codeex, but you've also
7:16 got image generation.
7:18 So,
7:21 I don't know. Hope you've been well.
7:24 Likewise. Just general use. Imagine a
7:26 50-year-old mom. Ah, okay. A 50-year-old
7:29 mom. I'm going to turn I'm about to turn
7:31 61. That's That's rough on me,
7:36 Yilong. That's That's rough on That's
7:38 That's rough on Daddy Kyle.
7:45 Uh Chachi PT,
7:48 it's okay. You didn't make me this old.
7:59 Like
8:01 Oh, no. Wait.
8:14 She came on here like slow moving cold
8:18 front
8:28 and a look in her eyes.
8:33 She sat on a stool and he said, "What do
8:36 you want?"
8:40 She said, "Give me a love that don't
8:43 freeze up inside."
8:52 Said, "I have melted some hearts in my
8:54 time, dear."
8:58 But to sit next to you, well, I shiver
9:01 and shake.
9:05 And if I knew love, well, I don't think
9:08 I'd be here.
9:12 Asking myself if I've got what it takes.
9:19 I do your icy blue heart.
9:24 Should
9:26 I stop?
9:29 Turn what's been frozen for you
9:35 into a river of tears.
9:50 Um,
9:52 chat GPT is what she should spend her 20
9:56 hardearned dollars on or whatever
9:59 denomination. I don't know. Are you in
10:00 the States?
10:02 Um, chat GBT.
10:23 Actually, you know what else, Elon? Is
10:26 uh
10:28 if she wants to spend an extra little
10:30 bit, I would say chatt plus suno.
10:37 It's 10 bucks a month and you can just
10:40 make amazing music. And like music just
10:43 it just it just [ __ ] makes you happy.
10:47 Music's the best.
10:49 So, I would go I would go two
10:50 subscriptions. Chat PT 20 bucks. Sunno
10:55 10 bucks.
11:00 Simmies in the house.
11:06 So, by the way, happy Friday night date
11:08 night. I hope you have your nachos and
11:10 your dirty hot pocket.
11:19 Giving props to Jim Gaffigan.
11:22 Your hot pocket
11:24 should be molten lava hot in the center
11:28 but frozen around it. So it should be
11:31 hard to bite into,
11:34 risk breaking a tooth and burn your
11:36 tongue. Okay.
11:43 I like suno. I use it some when I write.
11:45 Yeah. Which I don't do much anymore.
11:54 Sunno's helped me get through some some
11:56 major stuff.
11:58 Yep. My own strapedic
12:02 soundtrack.
12:04 That's a cool word. Strapedic. Pedic.
12:13 Well, welcome back for everyone. I I
12:15 love that we're having a little bit of
12:16 reunion night tonight. People People
12:18 coming back
12:28 in a westerly
12:30 direction.
12:34 Oh, therapeutic. I thought I thought you
12:37 gave me some some new ass word I'd never
12:40 heard of.
12:42 Scare therapeutic was pretty good.
12:46 It was had all sorts of hes in it, some
12:48 silent consonants. I liked it. But
12:52 therapeutic too. Yeah, sure.
12:59 Um, yeah.
13:02 So, yeah, just chat is it's more than
13:05 you need, but but the paid version of it
13:07 is what you should be using
13:09 and you should explore it. You should be
13:11 making images with it. The images are
13:13 really good. You can make signs and
13:15 cards and all sorts of crap like that.
13:18 And then, yeah, make some flipping
13:20 music.
13:41 So, we're going to be playing with
13:42 Notebook LM tonight.
13:44 We're going to be mining the history of
13:46 the AI learning lab and really
13:50 the history of of uh
13:53 generative AI. Oh.
13:56 >> Um,
13:59 one of the things that I'm really
14:01 excited about,
14:05 so a couple of things. Okay,
14:08 when I
14:10 graduated high school.
14:13 >> Hey, Kyle, if you ignore the sticky note
14:15 long enough, I'm going to pop up on
14:16 stage.
14:17 >> I already talked about it. I already
14:18 talked about that. I already read that
14:20 one.
14:21 >> Oh, all right. Well, I need to pay more
14:23 attention then.
14:32 That was pretty good. I like Claude.
14:34 Also, the political thing uh made me
14:36 like them more. Um but they're always
14:39 down. It feels Here's the thing about
14:41 Claude being down. Why they were down is
14:43 is likely why when when OpenAI goes
14:46 down, I would say Claude is a is a
14:48 strong second. Claude Claude's really
14:51 strong at writing. It's incredibly
14:53 strong at coding. I I just think if your
14:55 mom's going to use it for um for just
14:58 more general stuff, just chat chat GBT
15:00 is a little more accessible. Um but but
15:02 Claude's not bad. And yeah, if from from
15:04 a political standpoint, if you don't
15:06 like, you know, the fact that Open AI is
15:08 going with the targeting of humans
15:10 without human supervision, uh more power
15:13 to you. Um recommend Claude. Either one
15:16 of those two. Um you could also do
15:18 Gemini. The problem with Gemini is it's
15:21 still a little
15:24 it assumes you've got a level of
15:26 technical prowess that um it's just not
15:29 as it's just not as user friendly as the
15:31 other two.
15:33 Um yeah, so yeah, I get that. Uh but
15:37 we'll playing with Notebook LM and then
15:40 we'll be playing with Cloud Designer.
15:44 Oh. Oh, yeah. High school. Okay. What
15:46 was I going to talk about? Oh. Oh, I'm
15:49 just talking about Okay. So,
15:53 so when when when chat GPT first came
15:56 out, I had these visions of like, oh my
15:58 god, I think this this might be
16:01 happening. And here's what I think might
16:03 be happening. So, when I graduated high
16:05 school, I I was doing lots of plays and
16:08 theater, and I was like, I love the
16:09 theater, and I want to study theater. I
16:11 want to study how to be an actor.
16:13 And but I also had this voice in my head
16:16 like, you know, you're never going to
16:17 work. You're never going to get a job.
16:20 You like computers, go do the computer
16:22 thing. I didn't get much pressure from
16:23 my parents, from my mom. She didn't give
16:25 a [ __ ] She was happy with me to do
16:27 whatever. But this was like an in
16:29 internal survival mechanism.
16:32 I got to have skills.
16:34 So I started out as a computer science
16:36 major. and I walked into my first
16:38 computer science class at Penn State
16:41 University, York campus,
16:44 which is not Penn State University.
16:52 And uh I walked into my first computer
16:54 science class and the room is full of
16:56 mainframe IBM mainframe computers and
16:59 the professor walks out with a stack of
17:01 punch cards and he goes, "This is a
17:03 computer program
17:06 and you're going to spend your life
17:07 doing large reports for large
17:09 corporations on these giant machines and
17:11 you have to type them on these pieces of
17:13 cardboard." And my only experience with
17:16 computers was like an Apple 2
17:19 and uh and and the original Mac
17:24 and uh I looked at those punch cards and
17:27 I was like I will not be on this planet
17:30 if that's my job.
17:32 So I changed my major to theater and so
17:35 I have a degree in acting and I and I
17:37 just I decided at a young age I would
17:39 much rather be happy than have money.
17:42 And so I moved to New York out of school
17:44 and I pursued a a life in the arts and
17:49 and so for years
17:52 like people didn't even have to make fun
17:54 of me for my acting degree. I would make
17:56 fun of myself for my acting degree,
17:58 right? You know, as I'm eating my boxed
18:00 macaroni and cheese.
18:03 Um,
18:11 and now we're living in a world where
18:14 one of the most
18:17 technically advanced companies in the
18:19 world, Google, and one of their their
18:22 most
18:24 um technically advanced
18:26 um divisions, Google DeepMind,
18:30 they just hired a philosopher
18:34 So, so the the liberal arts major thing
18:36 seems to be coming around. One of the
18:38 other things that I've done all my life
18:41 is digital hoarding.
18:43 Like for all of my life, I'm like, I
18:46 should save these files because someday
18:48 they might be valuable.
18:51 And like even when I do these lives, I'm
18:53 like, let's get them on YouTube. Someday
18:56 they're going to be valuable. We live in
18:58 that world
19:01 where because what I would have never
19:03 done
19:04 is gone back and looked at them all,
19:07 gone back and looked at three years of
19:08 video and tried to figure out what's in
19:10 there. But we now live in a world where
19:13 digital hoarders and liberal arts majors
19:16 are rewarded.
19:21 What a [ __ ] time to be alive. Are you
19:24 kidding me?
19:29 Super important. Now that Hot Pockets
19:32 have eliminated the crisper sleeve.
19:35 Do they still have GenX cred? They
19:37 eliminated the crisper sleeve.
19:42 No.
19:43 Boycott Hot Pockets. What? How? How are
19:47 you supposed to make them crispy? Are
19:48 you just Are they just soggy and tough?
19:52 No.
19:53 No.
19:55 It's about time rewarded.
19:59 Yep. All gone. No, you should listen.
20:02 We're making an official change here.
20:04 Friday night date night. Nachos are
20:06 still in
20:08 unless the International Nacho Society
20:11 decided that corn chips are no longer
20:13 acceptable for nachos.
20:15 What do you mean there's no crisper
20:17 sleeve for the hot pocket? And what? I
20:19 suppose you're supposed to cook them in
20:20 a clean microwave?
20:23 That's not acceptable. No longer the
20:25 official food of the AI learning lab
20:28 Friday night date night.
20:30 Two words, air fryer. Oh, that's so
20:33 [ __ ] sad.
20:36 Wow. The crisper sleeve has fallen the
20:40 way of the air fryer. We are such a
20:42 shitty society.
20:57 Okay, let's go. Let's get started.
21:01 Let's get going. Let's get going. Let's
21:03 get going. Yeah. Yeah. Yeah. Hey, Panda
21:05 was shaking. I can accept the nacho
21:08 surrogate. Yeah, nachos. Nacho is still
21:10 super acceptable. By the way, I'm
21:12 drinking a peach monster tonight.
21:17 Why am I drinking Monster Energy drink
21:19 at 8:22 p.m. Mountain time?
21:23 because I know how to live.
21:29 Who kn Who knew the downfall of society
21:31 would come down to Hot Pockets? I listen
21:34 I as an executive at Here's the deal.
21:40 You're an executive in the Hot Pocket
21:41 franchise, right? And the CFO is like,
21:44 "Listen, listen. Highly refined flower
21:47 prices are through the roof. We we've
21:50 gone from a nickel a ton to 6 cents a
21:52 ton. We got to cut costs. Uh hey, just
21:56 quick question. Uh about this sleeve
22:00 thing. Uh it it's it's mighty expensive.
22:03 That costs us uh 34 cents per uh per
22:07 packet.
22:09 And uh you know that's up from 0.2. So
22:14 you know what I'm saying? It's kind of
22:16 cutting into the margins. So, uh, why
22:19 don't you think about if there's some
22:20 way we could eliminate the, uh, crisper
22:24 sleeve?
22:25 >> But the crisper sleeve is is is everyone
22:27 loves the crisper sleeve. It's it's all
22:29 Jim Gaffigan talks about.
22:31 >> Yeah. Yeah. No, it's it's legacy.
22:38 They [ __ ] caved. They [ __ ] caved.
22:43 [ __ ] spineless corporate weenies. Uh,
22:46 hey, listen team. I hate to break it to
22:48 you. This, and listen, this is not
22:50 coming from me. This is coming from from
22:53 the top.
22:55 We're going to have to lose the crisper
22:57 sleeve. What do you mean? Hot pockets
23:00 are going to be soggy and tough. I I
23:02 know. I know. I What? I They said
23:06 there's an alternative heating method
23:07 now. What do you mean? It's It's sturdy
23:10 microwaves. Dirty microwaves is what
23:12 It's what we've always cooked them in. I
23:14 know. I know. I know that's the brand. I
23:16 get it. Apparently air fryers are taking
23:20 over. Air fryers. It's just hot air. I
23:23 know. I know. I know.
23:28 Wow.
23:30 Staggering.
23:33 I'm a little shook.
23:35 I'm a little shook. Clearly, I don't
23:37 need to have hot pockets to know that
23:39 this change happened probably five years
23:41 ago.
23:46 Just a reminder, it's Gareth. Hi,
23:48 Gareth.
23:50 I just got an awesome email. I've been
23:52 accepted into the Claude Partner
23:54 Program. That's actually really [ __ ]
23:55 cool.
24:01 Am I the one who's not a fan of Hot
24:03 Pockets? Nobody's a fan of Hot Pockets.
24:05 Just white trash people occasionally eat
24:07 them and pretend like they don't.
24:13 I need more ice. I'll be back. Hold,
24:15 please.
25:01 All
25:07 right, I'm back.
25:10 Let's see.
25:18 I'm going to share my screen.
25:22 I'm gonna share. Let's see.
25:26 Is this it here? Yeah.
25:32 Okay, so we're in notebook LM.
25:37 Let me get black bar going.
25:50 You eat the cardboard with lava because
25:52 you have to. Exactly.
25:55 Apparently turtles dislike crisper
25:57 sleeves as much as they hate straws.
26:05 You know, listen, I hate to break it to
26:08 you, but there's some guy somewhere that
26:10 invented that silver coating on the
26:13 inside of the crisper sleeve. It's got
26:15 just enough metal to crisp the hot
26:18 pocket, but not enough metal to cause it
26:20 to spark and burn your house down. Like
26:24 someone did that. That's someone's
26:26 grandpa.
26:28 And he's got to be like, "Yeah, I
26:30 invented the crisper sleeve." And
26:31 everyone's like, "What are you talking
26:33 about?"
26:35 He's like, "The crisper sleeve in the
26:36 hot pockets. I I I made the the silver
26:39 lining. That's that's me." Like there's
26:42 no What are you talking? There's no
26:44 crisper sleeve. What is a crisper
26:45 sleeve? You put it in the air fryer.
26:48 These goddamn air fryers. You know,
26:51 there's a guy that's really sad out
26:52 there right now.
26:56 I thought you were talking about
26:58 Chrisper, the DNA DNA editor.
27:02 It didn't arc in the microwave genius.
27:05 Exactly. He probably made millions on it
27:07 and doesn't care how they use it. No,
27:09 it's all the people that invented that
27:11 corporate thing always take great pride
27:14 in that thing.
27:16 You you know the little dimple on the I
27:19 my it was my idea for the dimple.
27:23 Corporate people are weird man. They
27:26 spend 35 years in obscurity. They got to
27:29 take credit for the thing they made. So
27:31 like the the liner of the of the crisper
27:34 sleeve. That guy, you know, probably
27:37 worked at 3M or one of those kind of
27:40 companies.
27:42 My liver is still trying to detox the
27:44 crisper sleeve metal.
27:49 Oh my god. Okay.
27:51 So, so what I did here,
27:55 there's this really cool document. Can I
27:57 look at this document from within here?
28:01 Uh, what's going on? Why can I not uh
28:04 Oh. Oh.
28:06 Source guide.
28:15 Yeah. So, all of this all of this
28:17 document
28:19 So, this is a PDF document that's got
28:21 every
28:23 um like up through March, every
28:27 description of YouTube
28:30 videos with chapters and all that sort
28:33 of stuff.
28:36 Um,
28:38 so that's the core source document. The
28:40 other document I put in there was a
28:41 YouTube channel itself, but that seems
28:43 to be pretty useless. Um, and then I and
28:46 then I just went in and I had uh
28:50 I had
28:53 notebook LM like tell me what was in
28:56 there.
28:59 And I had it come up with five phases.
29:07 So here's what it came up with. Phase
29:09 one, the primordial soup, the and the
29:12 chatbt awakening early to mid 2023.
29:16 The initial phase was characterized by
29:18 explosive growth and mix of awe, fear,
29:20 and skepticism. As Chat GPT reached 100
29:24 million users, it's up to 900 million
29:26 now. Host Kyle Shannon frequently
29:28 compared the technological leap to the
29:30 early days of the worldwide web. I still
29:32 do that because I'm insufferable. I'm
29:34 like the guy that invented the hot
29:36 pocket crisper. You know, I was there
29:38 for the early days of the worldwide web.
29:42 People are like, "Shut up, old man.
29:46 Did you invent the iPhone? Then we don't
29:48 care.
29:56 Conversations focused on overcoming fear
29:59 of job displacement still and
30:01 encouraging users to experience Kevin
30:03 Mallister moments.
30:05 personal breakthroughs where individuals
30:08 realized the technology's profound
30:10 potential. Early exploration
30:13 centered on basic prompting, debunking
30:15 AI myths, and democratizing creativity
30:17 through first generation im image tools
30:20 like Dolly 2 and Midjourney.
30:23 That's pretty cool. Um,
30:26 let's see. I tried to make a slide deck
30:29 with Notebook LM today and it wasn't
30:30 very good and it couldn't edit. Well,
30:33 we're gonna go play with uh we're gonna
30:35 go play with uh what you call it.
30:39 Um
30:42 Claude Designer. We'll we'll see if
30:44 there's something interesting there. I'm
30:45 also trying to make a uh
30:48 I'm trying to make a video overview of
30:51 of the channel. It's taking a really
30:53 long time, but it said it might.
30:59 Oh, there's a chronicle. Oh, cool.
31:07 Okay. Anyway,
31:09 wait. Let's look at this chronicle.
31:12 So, I had to do a
31:16 How do I Where do I Hang on. Hold,
31:18 please.
31:20 Show prompt.
31:24 Okay.
31:26 And can I make the font bigger? Yeah.
31:28 Oh, here we go.
31:35 All right. Chronicle of the Great
31:37 Repurpose.
31:40 No. Well, okay. So, it got that wrong.
31:42 So, I don't know where it got the Great
31:43 Repurpose [ __ ] That's that's much
31:45 later. Historical Foundation. The dawn
31:47 of generative of the generative era,
31:49 April to late 2024.
31:52 In retrospect, the 2023 era was merely
31:55 the first tremor of the tectonic shift.
31:56 It was a period defined by initial
31:58 magic. A brief shimmering window of a
32:01 simple act of prompting chatpt felt like
32:03 a brush with the divine. Archaeological
32:06 evidence from the AI salon archive
32:08 suggests the frenetic energy of
32:10 discovery. By mid 2024, the landscape
32:13 was more cynical. The vending machine
32:16 mindset. This is wrong. This is not
32:18 good. Actually, that's a bad report. All
32:22 right, we'll go back to that central
32:25 thing because that was actually pretty
32:26 good.
32:35 Okay, here we go. Primordial soup. So,
32:38 phase 2 was late 2023 to early 2024.
32:43 So, that was late 2023 to early 2024.
32:46 That was when we did GPT for good. Oh,
32:48 yeah. Look at this. Customization,
32:50 democratization, and multimodality.
32:53 As the hype began to settle into
32:54 practical application, OpenAI introduced
32:57 the ability to create custom GPTs and
32:59 launched the GPT store. Remember the GPT
33:02 store? We were all going to be [ __ ]
33:03 rich making custom GPTs.
33:07 And then at some point, I think it was
33:09 Source Camp was like, uh, hey, could we
33:12 charge money for these? And Sam Alton
33:15 was like, yeah, yeah, it's going to be a
33:17 store. It's going to be awesome.
33:18 everybody's going to be rich. And then
33:20 he promptly forgot about it.
33:23 And Source Camp still doesn't have her
33:25 answer.
33:36 AI summary. Debating if the hot pocket
33:39 crisper sleeve is a good invention.
33:41 Share your opinions. I'm glad you're
33:43 showing this. No answer and 199 GPTs
33:47 later. Exactly. Um, this era marked the
33:50 shift from users simply interacting with
33:52 the chatbot to programming personalized
33:55 AI assistance. Remember that when we
33:58 were making those custom GPTs, the
34:00 miraculous part of that was you were
34:03 doing sophisticated programming without
34:06 having to program. It was just English
34:07 language.
34:09 Um, and the results were these these
34:12 pretty remarkable sophisticated tools.
34:16 Um Cindy [ __ ] this week thought she was
34:20 teaching advanced um agentic workflows
34:23 and it turned out the people in the room
34:26 didn't even know what a custom GPT was.
34:28 So she spent the whole time teaching
34:30 them custom GPTs. That was this week. So
34:33 the the gap between where we are if
34:37 you've been paying attention to this
34:38 stuff for three years and where people
34:40 are that are just starting is pretty
34:43 vast at this point.
34:45 that the world is starting to shatter in
34:47 into these distinct modes.
34:51 A major highlight of this period was the
34:53 GPT for good 24-hour hackathon where the
34:56 AI salon community collaborated to build
34:58 hundreds of custom GPT bots for
35:00 nonprofits. Simultaneously, AI became
35:03 multimodal. That was November of 24 or
35:06 23.
35:08 No, it was early 2024.
35:11 G gaining the ability to see GPT4 vision
35:14 speak advanced voice models and perform
35:16 complex data analysis code interpreter.
35:18 This is really good. This is actually
35:20 what's cool about this. This notebook LM
35:22 actually is capturing kind of the
35:24 highlight technologies of the day.
35:26 That's really cool. Phase three mid 2024
35:29 the creative explosion in video and
35:31 audio. Yeah, that makes sense. The
35:33 landscape shifted dramatically toward
35:35 rich media generation. The announcement
35:38 of OpenAI's Sora. Remember that?
35:40 Remember when the that the Sora video
35:42 came out? We're like, "Oh my god, that's
35:44 amazing."
35:46 And then like six months later they they
35:48 introduced like the shitty version of
35:49 Sora that sucked.
35:52 And then all the other video competitors
35:54 caught up and surpassed them and they
35:56 finally introduced Sora as like the
35:58 social media app which they've now
36:01 killed.
36:03 Wild.
36:05 Alongside tools like Runway, Gen 3,
36:08 Lumalab's Dream Machine signaled the
36:09 arrival of highquality texttovideo
36:11 capabilities. In parallel, AI music
36:14 generators like Sunno and Udo
36:16 democratized music production and
36:18 allowed users to generate full songs
36:20 with vocals, instrumentation, and text
36:22 prompts. This phase empowered creators
36:24 to become directors, blending tools to
36:26 create complex projects such as
36:29 Shannon's AI assisted Broadway musical
36:31 concept Sydney and artificial love
36:34 story, which is still kicking around.
36:37 We're around two years ahead of the rest
36:39 of the world. I know it's crazy, right?
36:42 Phase four, Agentic AI, deep research
36:45 and vibe coding, late 2024 to 2025.
36:49 This phase saw a pivot pivot from
36:51 passive commercial chat bots to active
36:55 autonomous agents capable of executing
36:59 multi-step tasks across the internet
37:01 such as OpenAI's operator Manis and
37:04 GenSpark. The concept of vibe coding
37:06 became a central theme as platforms like
37:08 lovable and replet enabled non-engineers
37:10 to build and launch functional web
37:12 applications simply by describing them
37:14 in plain English. Additionally, models
37:17 equipped with reasoning engines,
37:18 OpenAI's 01 and 03 and deep research
37:22 capabilities transformed how users
37:24 gathered this. This just this little
37:27 document alone is absolutely
37:29 fascinating. What are these links to?
37:33 Oh, it's it's it's links to chapters
37:36 within within the document.
37:39 This is all thanks to Lord Digital Gods,
37:41 by the way. Like none of this would be
37:42 possible without what he put together.
37:44 So, I don't know if you're here, Glenn,
37:46 but if you are, you you [ __ ] rock.
37:49 Okay. Um,
37:52 additionally, models equipped with
37:54 reasoning engines. Okay. Did uh
37:57 transformed how users gathered,
37:59 analyzed, and synthesized vast amounts
38:01 of complex data. Phase five, the great
38:04 repurpose and identity shift 2026. By
38:06 early 2026, the AI agents became highly
38:09 capable
38:11 of automating technical tasks, even
38:13 replacing the need for traditional
38:15 coding. The conversation shifted from
38:18 tool mastery to an existential reckoning
38:21 dubbed the great repurpose, which
38:23 happened yeah this year. With expertise
38:25 increasingly commoditized, professionals
38:28 faced an involuntary job redefinition
38:30 and identity crisis. The community
38:33 recognized that the future success would
38:34 rely less on technical execution and
38:36 more on soft skills, human connection,
38:39 taste, and intentionality. Creators
38:41 began managing custom autonomous agents
38:43 like Shannon's Adambot. He's he's an
38:46 embarrassment to the family name uh to
38:48 orchestrate complex digital workshops,
38:51 freeing humans to focus on meaning,
38:53 authenticity, and exploring the
38:54 question, what do I truly want to do?
38:58 Really good. And then it it provided
39:01 like clips
39:04 and so
39:06 Tik Tok pin. Thank you. That's why I
39:09 only do one show a night.
39:12 You're the best. You're the best. Law
39:15 digital gods. You really are. Um this is
39:18 so cool. Okay, so so what I'm going to
39:21 do, so I had to I had to pull out clips
39:24 that I could go watch, which was kind of
39:26 fun.
39:27 Let's see. Midjourney's end goal. Let's
39:30 see. Spider Jank spell.
39:38 The superhero named Spider Jank
39:48 casts a spell
39:52 with
39:54 his spider magic wand.
40:08 5 seconds. Generate.
40:11 All right. So, this does it pretty
40:13 quick, but it's going to be like 30
40:16 seconds here. We'll go back over to
40:17 cling. Clang cling.
40:21 That's so funny. Um Oh, I just closed my
40:25 sharing window. Hang on.
40:32 So that's kind of cool. So I've got a
40:34 list of
40:37 Yeah, these are pretty good.
40:40 Phase one, except I think what this did
40:42 was it it included
40:45 videos from the wrong years.
40:53 That's 23. Anyway,
40:56 but the like the thing I was saying
40:58 before about
41:00 um we live in this crazy time where
41:03 digital hoarding is being rewarded. So,
41:07 what we're going to do is this. I had
41:08 this thing write a um
41:14 a a prompt for Gen Spark. So, when I was
41:17 trying to get it to do good stuff
41:19 earlier,
41:22 system prompt. Okay, here's the complete
41:24 all-in-one prompt you can copy into clip
41:26 clip genie. Okay, so here's what I'm
41:29 going to do. Copy this. So, this
41:32 describes the sections and then it it
41:34 points out what clips to go make from
41:37 which videos.
41:40 Okay. So, I'm going to go now to
41:46 GenSpark.
41:55 Ultra mode now powered by Opus 4.7.
41:59 Stronger reasoning.
42:03 I guess I could turn that on. It's going
42:05 to blow through my credits though, isn't
42:07 it?
42:09 A new version of this website is
42:11 available. Refresh now.
42:16 Genpark claw desktop app.
42:20 Genpark claw is now free cuz it was like
42:23 $69 a month. Black bar. Um,
42:29 okay. Okay. So, I could go straight to
42:31 Clip Genie, but I think what I'm going
42:33 to do is I'm going to say, "Here's a
42:36 prompt
42:39 that
42:41 definitely requires
42:44 Clip Genie,
42:48 but this is a historical
42:53 look at Gen AI through the lens
43:00 of three years of
43:04 AI learning lab
43:08 video.
43:09 Wait, let's see. Live stream
43:14 video archives.
43:18 This could also be an interesting
43:22 presentation
43:24 with video clips.
43:28 I will let you decide
43:34 how to express
43:37 this important
43:41 historical
43:43 recounting
43:45 of the early evolution
43:49 of
43:53 Genai.
43:55 And then I'm going to paste in that
43:56 whole giant [ __ ] thing. And
43:59 we're going to let it do ultra. And what
44:01 we're probably gonna do here
44:04 uses more credits. I'm going to turn off
44:05 ultra. We don't need ultra for this. Um,
44:08 what I'm going to do here
44:11 is just turn this bad boy loose.
44:16 Oh, apparently I'm not. What? What
44:18 happened?
44:22 Your GenSpark Plus plan auto renews on
44:25 May 13th.
44:27 Oh, did I blow through all my Oh, you've
44:29 used all your credits.
44:32 Oh, damn it.
44:39 Well, so so there you have it. There you
44:42 have it, kids. Apparently,
44:47 I guess I'm not I guess I'm not making
44:49 some fancy ass historical document till
44:52 my credits renew. That's all right.
44:55 That's fine. Whatever. It's not a
44:56 problem. Not a problem. All right.
45:01 It's just tragic.
45:03 But I think what I will do, maybe I'll
45:06 do this over the weekend. I think what I
45:07 will do is I'll take this stuff from
45:10 from Notebook LM,
45:13 you know, this stuff where it's got
45:14 these like all these clips and these
45:18 descriptions of the five phases.
45:21 I you know, I'm going to I'm going to go
45:23 out on No, no, we'll just leave it like
45:25 that. I'll turn that into some sort of
45:27 LinkedIn post or something like that.
45:28 That could be kind of fun.
45:30 Um, I turned this into a cinematic
45:34 podcast
45:36 and so let's watch we'll we'll watch it
45:40 for a bit. It's 5 minutes. If it doesn't
45:44 horrifically suck,
45:46 then we will um
45:54 we'll keep watching it.
45:57 George Carlin didn't narrate it, right?
46:00 Yeah. No, we got booted last night from
46:02 YouTube. Oh, what I found out from
46:04 YouTube, for those of you that got were
46:05 watching on YouTube and it just stopped,
46:08 it was it wasn't because he was saying
46:10 cuss words, it's because that clip was
46:12 copywritten and so I was streaming a
46:15 copywritten work live. And so the way
46:18 YouTube works is if they sense that
46:19 you're streaming copywritten work,
46:22 they'll pause the live until you stop
46:24 streaming it. So, if I had just stopped
46:26 the George Carlin clip, it probably
46:28 would have come back, but I didn't know
46:29 that until just now. Okay. Is that a
46:32 silicon graphics machine? Oh, it's a
46:33 Next cube.
46:35 >> That's interesting.
46:36 >> Kyle Shannon helped build the early
46:38 worldwide web on systems like this next
46:41 computer.
46:42 >> No.
46:42 >> Decades later, as a new technological
46:44 wave began to swell in 2023, he started
46:48 recording. He launched a nightly live
46:50 stream, a three-year video archive
46:53 capturing the modern artificial
46:55 intelligence revolution in real time.
46:58 This archive isn't just a tech log. It's
47:00 a map of human psychology. It tracks the
47:03 specific predictable stages our
47:05 collective mind goes through when a
47:07 technology begins to outpace our
47:10 >> By the way, this is horrifically bad
47:12 right now.
47:13 >> In early 2023, we were in stage one
47:16 dismissal. Most people saw chat bots as
47:19 janky novelties. They laughed at the
47:22 broken responses and the bizarre
47:23 hallucinations, treating the technology
47:26 as a flawed parlor trick that couldn't
47:28 handle basic facts. Because the errors
47:31 were so visible, we ignored the
47:33 trajectory. We focused on the failures
47:35 of the moment, creating a cultural blind
47:38 spot that hid the sheer speed at which
47:40 the underlying neural networks were
47:42 compounding in power. By late 2023, the
47:45 tone shifted into stage two, awe tools
47:49 like midjourney and Dolly reached a
47:52 >> which by the way it I didn't tell it to
47:55 use these stages of dismissal dismissal
47:58 law. This is the five stages of AI
48:00 adoption. I didn't tell it to do this.
48:02 It found that in that document
48:04 >> threshold where a digital painting could
48:06 be assembled out of static in seconds,
48:08 producing images that looked
48:10 indistinguishable from human artistry.
48:12 Users discovered they could type a messy
48:15 paragraph into a text box and watch as a
48:17 help me write this button instantly
48:19 converted it into a coherent business
48:21 plan or a working block of code. This
48:24 created a vacuum effect. The technology
48:27 was suddenly capable of high-level
48:29 cognitive tasks, but the average user
48:31 had no framework to understand how a
48:33 machine was suddenly thinking or
48:35 creating. Ae is rarely a permanent
48:37 state.
48:38 >> This is weird cuz this is
48:43 a this is shitty,
48:46 but it's also um
48:49 it's not it's not using the the
48:53 document.
48:54 It's just making [ __ ] up.
48:57 Um,
49:01 explainer brief. Let's see.
49:06 I'm going to make an audio podcast
49:08 because, well, maybe I'll do a video
49:09 one.
49:11 An explainer.
49:15 All right, let's do an explainer. Choose
49:17 the visual style.
49:19 Classic custom anime watercolor
49:23 heritage paper craft retro print. We'll
49:27 go retro print
49:30 career focus professional persona.
49:35 All right. Let's say um
49:38 use the transcript. Wait, use the the
49:43 content of the docu the content
49:48 of the PDF with all the YouTube, all the
49:54 AI learning lab
49:59 um video descriptions
50:03 and chapters
50:07 to
50:10 Chronicle
50:16 the
50:20 evolution
50:22 of
50:26 Gen AI
50:29 from its birth
50:33 in 2022.
50:36 And we'll put even though
50:40 These
50:42 transcripts
50:44 start
50:46 in 2023
50:51 until March of 2026.
50:56 Um,
50:58 create your own logical
51:02 eras
51:03 based on the contents of the shows.
51:09 All right, let's see how that does.
51:12 So, we'll come back to that. Okay, let's
51:14 go play with Jen uh Claude Designer. Um,
51:20 so if you go to if you want to play
51:22 along, claude.ai
51:25 slashdesigner
51:28 and I haven't really play Oh,
51:31 slashdesigner slashdesign might be
51:34 slashdesign.
51:37 Yes. Slashdesign. Okay.
51:41 Meet Claude design. Oh, apparently I
51:43 need to share this for you to see it.
51:47 Did you input custom instructions to
51:49 help improve form and I did Kuno, but
51:52 apparently I put in like you know what?
51:54 You know what it almost did, Kuno? It it
51:57 seemed to make that entire video based
51:59 on my description rather than actually
52:01 looking at the source document. that
52:03 seemed to be just a screw-up of of uh
52:06 either bad prompting on my part or just
52:08 bad
52:10 bad uh I don't know something management
52:13 on Notebook LM's part, but we'll see.
52:15 We'll see what it does with the with
52:17 that other one.
52:25 Can I go back? Is there audio for this?
52:27 You know what? Hang on a sec.
52:32 Let's go to Let's go to X. Maybe people
52:34 have done videos on this already.
52:40 Search
52:42 Claude design.
52:47 And then we're going to play with it.
52:58 Why?
53:00 I'm confused here.
53:09 There. That's better for Tik Tok,
53:13 right?
53:15 A little bit more better.
53:24 Here you go. Okay. Um, cloud design.
53:43 Oh, that's interesting. So, was the
53:45 official Figma integration with Claude
53:48 just to grab data? Claude officially
53:51 released design. From now on, without
53:53 integrating state-of-the-art large
53:56 models, there's no traffic growth. But
53:57 if you integrate, you might get replaced
53:59 by distillation.
54:01 Killing the goose that lays the golden
54:02 egg. Okay, so here's a video.
54:32 I love
54:44 it. Heat. Heat.
55:24 All All right, cool.
55:27 So, that's Tik Tok pin. Why does it
55:30 sound like an 80s video game? Oh,
55:32 everyone's just trying to get nostalgic.
55:34 Everyone Everyone's all excited about
55:36 Gen Xers. No, they're not.
55:42 Cloud design is the new Canva. Why pay
55:44 for Canva when you can hand it off to
55:45 Cloud Code? RIP lovable.
55:49 Cloud design is honestly pretty [ __ ]
55:51 great. They're really leaning into what
55:54 is still the most creative model.
55:58 I asked it to give me three directions
55:59 for a website I got and the results are
56:02 stunning.
56:04 Introducing Claude design by Anthropic.
56:07 Make prototype slides, one pages by
56:09 talking to Claude.
56:11 Enthropic released code today. This is
56:14 18 minutes.
56:19 I spent the last hour giving it a first
56:21 look and shared my thoughts in the video
56:23 below. This is a big drop. This is a new
56:26 design surface from Anthropic.
56:29 Holy moly, Cloud Design is amazing. It
56:31 straight up read my codebase. Ask me
56:33 some questions and I gave it some vague
56:36 make it better, please.
56:39 It's so much better than the crap I
56:40 designed. Loving it.
56:43 So, I can already tell you Cloud Design
56:45 is unlike anything else out there right
56:46 now. I just finished comprehensive
56:49 design philosophy components library
56:51 fonts directory animation gallery last
56:53 week and with design I can conveniently
56:56 link okay
57:14 the designer's position is very stable.
57:18 Even if so many companies take turns
57:20 launching design agents, they still
57:22 can't replace full-time designers.
57:27 Designers of the future will only be
57:29 greater.
57:31 Well, yeah, but like it it this is
57:34 changing the definition of who gets to
57:35 be a designer. Like it used to be you
57:38 had to have tactical skills to be a
57:40 designer. What these tools are going to
57:42 allow is that anyone who's got taste
57:46 Right? Anyone who's got intuition and
57:48 taste
57:50 and has, you know, enough curiosity to
57:53 learn these tools can actually do design
57:55 now.
58:02 Designers, stop using Claude.
58:06 All right, let's go play with it.
58:20 Okay.
58:22 So, what we've got here, company name
58:25 and blurb. Let me go back. Can I change
58:27 the name of this thing?
58:31 Any other notes?
58:34 Why can I I not change the name of this?
58:40 Link code from your computer. This
58:42 doesn't upload the whole codebase. Cloud
58:43 will copy selected files.
58:49 Upload a Figma file here. Add fonts,
58:53 logos, and assets. All right. So, let me
58:56 do I'll do some story stuff.
59:06 Hang on a sec. I'm just I'm just
59:08 navigating through
59:16 Here's brand assets. So, we've got, you
59:20 know, I'm going to go change my sharing
59:22 so you can see what I'm doing on screen.
59:26 So, basically what I'm looking to do
59:28 right now is go find documents.
59:33 Is this our brand sheet?
59:38 Yeah. Yeah. So, there's our brand sheet.
59:40 Okay.
59:42 Butus.
59:54 So, that's our brand sheet.
59:59 Let's see
1:00:02 what else do we have here.
1:00:14 or business cards.
1:00:18 And I don't need that.
1:00:20 Fonts, logos, assets. What's this?
1:00:25 Oh, that's our PopSocket thing. Well,
1:00:27 we'll throw that up there. Why not?
1:00:36 logos.
1:00:39 Let's see.
1:00:50 There's our little dude.
1:00:53 So, there's the PNG of him. Yeah, that's
1:00:56 good.
1:00:58 Drag that in there.
1:01:02 Oops.
1:01:08 And then we'll go reversed.
1:01:11 We'll grab that PNG.
1:01:14 Beautiful.
1:01:19 And then we'll go. Ah, Damn it.
1:01:23 Let's see. RGB.
1:01:32 logo
1:01:35 no tag
1:01:37 RGB
1:01:39 two color
1:01:42 we'll go PNG
1:01:47 that one
1:01:51 and then we'll go
1:01:55 reverse alt white background ground.
1:02:00 Yep, that's good. That goes there.
1:02:20 And then there's that one, which I like
1:02:22 too. Put that there. Okay. Okay, so we
1:02:26 have our stylesheet. We have logos.
1:02:27 Black bar, please. Sorry about that.
1:02:30 Sorry about that. Carrying on. Redo chip
1:02:34 pip.
1:02:35 I can ask it to search for things on the
1:02:38 web
1:02:40 and do tasks for me on the computer. It
1:02:42 has memory. That's cool. Then the other
1:02:44 thing I can do here. Well, wait. So,
1:02:49 hang on a sec. Um, I should have a brand
1:02:53 strategy document. Let me let me see if
1:02:55 I can go find that for a second.
1:02:59 Um,
1:03:09 logos carry the guides.
1:03:14 Sharable logos. Oh, that's logos. Okay,
1:03:16 hang on.
1:03:21 App sheet graphics copy for story mind
1:03:24 images
1:03:26 head style branding. Oh, here's here's
1:03:30 photography.
1:03:35 No
1:03:36 oneliners business cards.
1:03:49 Ah, storyline brand strategy. Here we
1:03:51 go.
1:03:56 This is a PowerPoint file.
1:04:06 I wonder if it can read this. That'll be
1:04:08 interesting to see.
1:04:53 Oh, bless you. Thank you.
1:04:56 I can ask it to do search for things on
1:04:58 the web. Okay. Do I want to do I want to
1:05:01 put this document in here?
1:05:05 I think that was the old one. This is
1:05:13 Lab summary final deck for creatives and
1:05:16 contractors. All right, let's throw that
1:05:18 in there. Let's see if it'll eat that.
1:05:20 It seemed to. Okay, now the other thing
1:05:23 I can do
1:05:25 is I can grab
1:05:36 I don't feel like doing that. You can
1:05:39 grab a GitHub
1:05:41 folder. I don't think I want to do that
1:05:42 right now because I'm just
1:05:43 experimenting. But you can throw a
1:05:45 GitHub folder on here for like your your
1:05:48 front end for your for your web app
1:05:52 um or for I mean for your applications
1:05:55 like you know your your interface stuff
1:05:57 for how you design files.
1:05:59 So you can either link code from your
1:06:01 computer
1:06:04 adds font add fonts logos assets. I
1:06:06 wonder why I can't add a website
1:06:10 company name and blurb. So, it's story
1:06:13 vine
1:06:15 and then the blurb is I wrote this
1:06:19 earlier and it was pretty good
1:06:24 story overview.
1:06:27 Here it is.
1:06:29 This did pretty good. It's It's actually
1:06:32 nice when you have enough history in
1:06:35 chat GBT that you can just go, "Hey,
1:06:37 give me a blurb about this and it
1:06:39 actually knows what the [ __ ] it's
1:06:40 talking about." It actually wrote this
1:06:43 really good line at the end that I don't
1:06:44 think we've ever used, but I like it. At
1:06:46 its core, StoryVine is about something
1:06:48 simple but hard to scale. Real people
1:06:51 telling real stories delivered in a way
1:06:54 that actually connects. It's pretty
1:06:56 [ __ ] good.
1:06:58 Okay, any other notes? Um, let's see.
1:07:06 The StoryVine brand.
1:07:11 The StoryVine brand is
1:07:15 professional
1:07:17 but has an informal humorous tone
1:07:21 to it.
1:07:28 The design should be clean
1:07:35 and the copy
1:07:38 always accessible.
1:07:41 Period.
1:07:43 We refer to the little character that is
1:07:46 part of our logo as the gnome.
1:07:51 And we talk about the gnomes up in the
1:07:53 cloud that automagically make videos.
1:07:58 Period. So that's a good example of the
1:08:01 kind of
1:08:03 fun yet professional informality that
1:08:07 I'm talking about. Period.
1:08:10 Okay. I don't know if that's any good,
1:08:12 but
1:08:13 that at least give us at least it has
1:08:15 our logos. It has our brand stylesheet.
1:08:18 It's got some description of the company
1:08:20 and and our brand. So, continue to
1:08:23 generation. It will take about five
1:08:25 minutes to generate your design system.
1:08:28 You can step away. Keep the tab in the
1:08:30 background. Okay, cool. So, we'll let
1:08:32 that do that. That's doing its thing.
1:08:35 Let's go back to
1:08:38 um Notebook LM.
1:08:41 All right. So, our our explainer video
1:08:43 is done. Let's go look at this and see
1:08:45 if this is sucks less. Hang on a sec.
1:08:49 Wasn't there like a funny image thing
1:08:50 that we did a year ago when we were all
1:08:52 animals? Todd Waller, I'm late. Oh my
1:08:56 god, Todd, we've got to start over.
1:09:00 All right, hang on. Let me get my
1:09:01 guitar. Champ, you got to get up. Todd's
1:09:03 late.
1:09:05 Champ,
1:09:07 he's not getting up. Union. Union dog,
1:09:11 he's not getting up for you, Todd. Sorry
1:09:13 about that. Okay. Uh, I need to change
1:09:16 my sharing tabs
1:09:23 turtles.
1:09:26 Okay,
1:09:28 let's see if this video sucks less.
1:09:32 Why is it called the great repurpose?
1:09:33 Okay, we got to we got to fix this. What
1:09:36 the [ __ ]
1:09:37 AI learning lab descriptions.
1:09:41 Why is this called the great repurpose?
1:09:49 Um, what's this? Click
1:09:52 configure chat. Well, whatever.
1:09:55 >> All right, let's talk about the
1:09:57 whirlwind evolution of generative AI.
1:09:59 Seriously, it has been a wild ride.
1:10:02 Today, we're going to break down this
1:10:03 entire journey into four distinct eras
1:10:06 using some incredible source material to
1:10:08 map out how we went from just playing
1:10:10 around with these tools to well to a
1:10:13 full-blown identity crisis in just a
1:10:15 couple of years. And this quote right
1:10:17 here, this is what sets the stage for
1:10:19 everything we're about to cover.
1:10:20 >> I remember such a powerful line because
1:10:22 it perfectly captures that feeling, you
1:10:24 know, that feeling of standing on the
1:10:27 edge of some massive irreversible
1:10:29 change. So, let's take a trip back to
1:10:31 where this all kicked off. Okay, so
1:10:33 let's dive right into era 1. This was a
1:10:36 time of just pure chaotic discovery. It
1:10:39 was messy. It was exciting. And
1:10:41 honestly, it felt a whole lot like the
1:10:43 early days of the internet. Remember
1:10:45 that? The dialup sounds, the clunky
1:10:47 websites. It was a total wild west where
1:10:50 nobody really knew the rules yet. And it
1:10:52 all happens so unbelievably fast. I
1:10:54 mean, look at this timeline. Chat GPT
1:10:57 drops in late 2022 and boom, by the
1:11:00 middle of 2023, the entire world is
1:11:02 talking about it. And the call to action
1:11:04 back then wasn't about becoming an
1:11:06 expert or building a business. It was
1:11:08 just get curious, go play, see what
1:11:10 happens. And this is what's so
1:11:12 fascinating.
1:11:13 >> Which, by the way, I still think that
1:11:16 that's, you know, the community's call
1:11:18 to action in mid2023 is about getting
1:11:21 curious and start playing. I still think
1:11:22 that's what it's about.
1:11:24 >> Go play. See what happens.
1:11:26 And this is what's so fascinating. That
1:11:28 feeling was so familiar. If you were
1:11:30 around in the 90s, you know what I'm
1:11:31 talking about. The AI tools were
1:11:33 described as janky and unpredictable.
1:11:35 Just like those early web pages. But
1:11:37 just like the web, they were absolutely
1:11:39 overflowing with this raw untapped
1:11:42 potential. The web democratized
1:11:44 information. This was democratizing
1:11:46 creation itself. And that was the whole
1:11:48 philosophy. That was it. It wasn't about
1:11:51 making perfect polished products. It was
1:11:54 all about experimenting, about pushing
1:11:56 these new kind of broken tools to their
1:11:58 absolute limits, just to see what they
1:12:00 could do. The best way to learn was just
1:12:02 to play. But things started to speed up
1:12:06 dramatically. This brings us to the next
1:12:08 era where the entire game changed. AI
1:12:11 moved way beyond just turnurning out
1:12:12 text. It started to see, to hear, and to
1:12:16 speak. It was basically getting senses.
1:12:19 And wow, what a leap this was. All of a
1:12:22 sudden, AI wasn't just a writer anymore.
1:12:24 With tools like Dolly E3, it was an
1:12:26 artist. With Suno, it was a musician.
1:12:28 With real- time voice, it became a
1:12:30 conversational.
1:12:33 The style of this is kind of generic and
1:12:35 and whatever, but like this is pretty
1:12:38 cool that out of just me rambling for
1:12:41 three years, it's pulling this history
1:12:42 out. It's pretty cool.
1:12:44 >> And you could already see the early
1:12:46 hints of text to video, which was just
1:12:47 mindblowing at the time. And you have to
1:12:50 remember this wasn't a one-horse race.
1:12:52 Not at all. When major players like
1:12:54 Google dropped their Gemini models, it
1:12:56 was like pouring gasoline on an already
1:12:58 raging fire. The competition just kicked
1:13:01 into hyperdrive and it felt like these
1:13:02 AIs were getting smarter. It actually
1:13:05 somehow seemed to miss that Gemini's
1:13:06 initial entries into this sucked so bad
1:13:09 we made fun of them for a year and a
1:13:11 half. They've caught up
1:13:13 >> every single week. And here's the most
1:13:15 crucial part. All this incredible power
1:13:18 was being handed directly to us, to
1:13:21 everyone. You didn't need to be a top
1:13:23 tier coder anymore to build your own
1:13:25 specialized AI. You could literally just
1:13:27 talk to it in plain English, give it
1:13:29 some files, and create your own custom
1:13:31 GPT for well, pretty much anything you
1:13:34 could dream up. Okay. Lord Digital Gods,
1:13:38 play is number one. It's the It's the
1:13:40 way mammals learn at the early stages.
1:13:43 Play opens your mind and you are more
1:13:45 receptive. Yep. So cool.
1:13:47 >> So, this brings us to 2025. By this
1:13:50 point, some of the initial novelty
1:13:52 starts to fade a little and the focus of
1:13:54 the whole community begins to shift.
1:13:56 It's not just about collecting cool new
1:13:58 AI toys anymore. It's about figuring out
1:14:00 how to actually put them to work. So,
1:14:02 that initial phase, that wideeyed
1:14:05 excitement of just discovering what was
1:14:06 out there, it started to mature into
1:14:08 something way more practical. People
1:14:10 began asking, "Okay, this stuff is
1:14:13 amazing, but how do I make it a real
1:14:15 habit? How do I consistently weave this
1:14:17 into my daily life to actually improve
1:14:19 my work?" And that line of thinking led
1:14:22 to a total reinvention of how work gets
1:14:24 done. It was no longer about using one
1:14:26 little tool for one little task. No, it
1:14:28 was about building entirely new AI
1:14:30 powered workflows from the ground up for
1:14:33 everything from a complex marketing
1:14:34 campaign to seriously a pet selfie app.
1:14:38 And right here, this is where you start
1:14:40 to feel a little bit of a chill. This is
1:14:42 where we see the first real hints of a
1:14:44 much deeper identity crisis. Because as
1:14:47 AI got better and better at filling the
1:14:48 gaps in our own skills, coding, writing,
1:14:51 design, it started to encroach on the
1:14:53 very tasks that used to define our
1:14:55 professional careers. And that leads us
1:14:57 straight into the era we're in now. An
1:14:59 era where the most important questions
1:15:01 have become less about the technology
1:15:03 and a lot more about us. Welcome to the
1:15:06 human frontier. This is it. This is the
1:15:08 big unavoidable question of 2026. When
1:15:12 we have these hyper capable AI agents
1:15:14 that can manage all the digital
1:15:15 complexity that used to fill up our
1:15:16 workdays, what's actually left for us to
1:15:19 do? What is our purpose? The source
1:15:22 material calls this period the great
1:15:24 repurpose. And honestly, it's the
1:15:25 perfect name. This isn't just about
1:15:27 >> the source material.
1:15:31 This is so generic and shitty.
1:15:33 >> A new tech tool. It's a massive cultural
1:15:35 movement. a moment where we're all being
1:15:37 forced to confront our own value and our
1:15:39 own identity in a world where our tasks
1:15:41 aren't uniquely ours anymore. Now,
1:15:43 listen to that phrase, a mass ego death.
1:15:47 It's startling, right? But it cuts right
1:15:49 to the heart of it. It's this profound
1:15:51 challenge to our sense of selfworth when
1:15:54 the things we do, the skills we spent
1:15:56 our whole lives building, can now be
1:15:58 done by a machine. And this just shows
1:16:00 how much the focus has shifted. It's
1:16:02 absolutely wild. We've gone from just
1:16:05 using these tools to actively designing
1:16:07 them as collaborators. We're talking
1:16:09 about giving them a soul, a core
1:16:11 identity, values, and memory. The goal
1:16:13 is no longer just to make it work, but
1:16:15 to give it a personality. And so, our
1:16:17 whole journey comes full circle. We
1:16:19 started out just tinkering with these
1:16:21 clunky, janky tools. And now, we've
1:16:24 arrived at a place where the most
1:16:25 crucial skill you can have isn't prompt
1:16:28 engineering or coding. It's the deeply
1:16:30 human skill of self-awareness, of
1:16:32 knowing your own purpose so you can
1:16:34 direct these unbelievably powerful tools
1:16:36 with real intention. Which leaves us
1:16:39 with this final really provocative
1:16:41 thought to chew on as we keep building
1:16:43 this future. Are we really just pushing
1:16:45 the limits of technology? Or are we
1:16:47 finally being forced to explore the true
1:16:49 frontier, which is what it actually
1:16:51 means to be human? Something to ponder.
1:16:53 Thanks for watching. All right, that was
1:16:56 crappy and generic but accurate. So, I
1:17:00 think a next like like how to how to
1:17:04 turn that into something usable would be
1:17:07 to write a better prompt that basically
1:17:08 gives the context of what the AI
1:17:10 learning lab's about and tell it to like
1:17:13 pull the personality out of the out of
1:17:15 the [ __ ] document that like that
1:17:17 thing so lacked personality. It was
1:17:19 awful.
1:17:22 Did you not choose cinematic? I did not
1:17:24 choose cinematic because the cinematic
1:17:26 one was was really [ __ ] awful. So,
1:17:29 I'll do cinematic next time.
1:17:32 Um,
1:17:35 I could do cinematic again here. Hang
1:17:37 on.
1:17:39 We'll do cinematic. New rich immersive
1:17:41 experience. Let's see.
1:17:45 Um,
1:17:49 I want you to create a
1:17:55 historical
1:17:59 What's going on with this video? What is
1:18:03 Hang on?
1:18:06 Let me reload this
1:18:10 video.
1:18:16 I want you to create a historical
1:18:20 overview. What? What is going on? Why is
1:18:22 it doing that?
1:18:31 you to create
1:18:37 a historical overview
1:18:41 of the last three years of generative AI
1:18:45 based on the transcript
1:18:48 which are the description and chapters
1:18:55 of the YouTube channel of the AI
1:18:57 learning lab hosted by Kyle Shannon.
1:19:00 Period.
1:19:02 Make sure you look at the
1:19:06 personality and humor of
1:19:09 the descriptions
1:19:19 and
1:19:22 break the history down into
1:19:26 five distinct eras
1:19:30 based on what you see in the document.
1:19:32 Period.
1:19:36 The video should be titled
1:19:48 the history of generative AI
1:19:51 through the eyes of the AI learning lab.
1:19:56 Comma
1:19:59 and irregular perspective.
1:20:02 Period.
1:20:08 Should be called
1:20:11 should be titled
1:20:14 the history of generative AI through the
1:20:18 eyes of the AI
1:20:20 learning lab.
1:20:28 and
1:20:30 irregular perspective. That's pretty
1:20:33 good, right?
1:20:36 Okay, let's see if that makes a less
1:20:38 shitty thing. Now, let's jump back over
1:20:40 to Claude. I got to share this tab now.
1:20:51 Did it do it? No assets registered yet.
1:20:54 Design system. What happened here?
1:20:58 Running script. Running script. Chat
1:21:00 upstream. Error 17. Agent upstream
1:21:03 returned. Invalid request.
1:21:07 Describe what you want to create. Design
1:21:09 files.
1:21:12 Assets.
1:21:23 Oh, no. It It's using that deck. That
1:21:26 deck was a conceptual deck. All right.
1:21:29 That's bad.
1:21:32 Let's see.
1:21:36 Uploads. Can I get rid of brand deck?
1:21:39 Story vine brand deck.
1:21:41 Can I get rid of this? Can I get rid of
1:21:43 something? Delete.
1:21:47 Delete.
1:21:49 Delete.
1:21:51 Delete.
1:21:53 Okay.
1:21:57 Do I need to like redo something?
1:22:01 Design system. No assets registered yet.
1:22:06 Um,
1:22:08 make me a
1:22:11 promo card
1:22:13 for
1:22:17 a
1:22:24 conference.
1:22:25 Let me just give it a really shitty
1:22:27 prompt and see if it does anything
1:22:29 remotely interesting.
1:22:31 Caramelizing onions.
1:22:34 Click comment in the toolbar
1:22:38 and then click any element
1:22:41 to annotate it.
1:22:51 You cannot unsee bad font pairing.
1:22:53 Choose carefully.
1:22:56 Copied Storyline logo guided wrote
1:22:59 conference promo card writing styles
1:23:05 design system creating your design
1:23:07 system
1:23:15 I'll start by exploring the provided
1:23:16 brand assets to understand um extract PP
1:23:21 write the content fundamentals in the
1:23:23 read me
1:23:25 Right. Visual foundations.
1:23:32 Huh.
1:23:37 Writing styles.
1:23:44 That's still generating.
1:23:47 All right. Anybody got any questions,
1:23:49 thoughts?
1:23:54 Weaver. Weaver's here. It's like it's
1:23:57 like old old reunion night.
1:24:00 Good to see everybody. Missing fonts.
1:24:16 Huh?
1:24:20 Okay. Okay, so these are real editable
1:24:21 things except
1:24:25 Wait.
1:24:27 Oh, it's still working.
1:24:31 Powering your story, honoring your
1:24:32 voice. That looks like it got that off
1:24:34 our website.
1:24:36 Stop by booth 1422. Film a story in 90
1:24:39 seconds.
1:24:41 Free PopSockets. That's pretty funny.
1:24:44 They got our colors right.
1:24:50 So wait, how do I upload fonts?
1:25:03 Oh, I see these are scalable and
1:25:06 resizable. I see what's happening here.
1:25:19 So the important thing about the
1:25:21 scalable and resizable is these are not
1:25:24 so right now you can do this kind of
1:25:26 design stuff in
1:25:30 Gemini Nano Banana and in chat GPT but
1:25:34 it's just generating a flat image. This
1:25:38 is actually creating designs and it's
1:25:40 it's using the wrong fonts right now
1:25:43 because I didn't upload the right fonts.
1:25:45 But we can that that's fixable.
1:25:48 That is fixable.
1:25:51 Fixed V3.
1:25:54 Good catch. The gnome's orange body
1:25:56 dissolves into the Let me fix. Oh, so it
1:25:58 caught its own errors. V3 now uses an
1:26:01 all-white reverse gnome that read reads
1:26:04 cleanly against the blue background. I'm
1:26:06 going to say um use the white slash
1:26:11 orange logo
1:26:14 against the blue
1:26:18 rather than the all white.
1:26:23 Let's see if it fixes that. Let's see if
1:26:25 it understands.
1:26:27 But this is kind of cool. So now what
1:26:29 are we going to do? We're going to
1:26:30 export these as
1:26:34 I guess let's export them as a PDF and
1:26:38 see what we get.
1:26:46 Why did it not export? Export as PDF.
1:26:49 Maybe because it's rendering.
1:26:54 Share.
1:26:58 Can I
1:27:02 Can I zoom? How do I
1:27:11 meet the gnomes?
1:27:13 They do the editing so you don't have
1:27:15 to. That's pretty cool. That's pretty
1:27:17 good actually.
1:27:25 Missing brand font. Upload fonts. Okay,
1:27:27 cool. So, let me go download Open Sands
1:27:30 and Monserat. Um, Google Fonts, Open
1:27:34 Sands.
1:27:37 Download
1:27:40 Open Sands.
1:27:44 Let's see
1:27:50 Styles. Get font.
1:27:55 Download all.
1:27:58 Okay, there's open sands. Beautiful.
1:28:01 Now, let's see if we can get Monserat. I
1:28:03 don't I think Monserat is not a Google
1:28:05 font.
1:28:08 Um,
1:28:11 let's see
1:28:14 one font family. Okay. Uh, Mont Serat
1:28:25 couldn't find any fonts. Oh, here we go.
1:28:28 Monzerat. Get font. Download all. Cool.
1:28:33 Okay.
1:28:35 Beautiful. I bet I wonder if I can
1:28:37 upload the zip files. I bet I can.
1:28:39 Upload fonts.
1:28:43 Nope. Okay. So, we'll go.
1:28:48 Oh, you're not seeing this, are you? No.
1:28:51 Hang on.
1:28:53 Everybody just calm down.
1:29:07 Okay, there's our fonts. So, let's
1:29:09 upload fonts.
1:29:17 Open.
1:29:21 Upload fonts.
1:29:32 Oh, that's cool. There's static. Static
1:29:35 and variable. Let's just upload the
1:29:37 variable ones and see if it can figure
1:29:38 it out.
1:29:44 font uploaded.
1:29:46 So, let me stop it here. I'm going to
1:29:48 say I uploaded
1:29:52 Open Sands and Mont Surat
1:29:58 um variable fonts. Please update designs
1:30:07 accordingly.
1:30:17 Shang
1:30:18 dang dang.
1:30:29 Let me check what got uploaded. So, we
1:30:31 might need to upload the static weights
1:30:35 fonts instead of the variable ones.
1:30:38 I shoed for Mac Minis this morning, but
1:30:40 they're all sold out. Uh, hey Todd, I
1:30:44 got a
1:30:46 I got the MacBook Pro, the M5.
1:30:49 Um, I got it in like a week
1:30:52 and um I got the 14inch one and it is
1:30:58 really impressive. It's very small.
1:31:01 What's fascinating about the 14-inch one
1:31:03 is I thought it would be like a total
1:31:04 compromise. like I had in my head the
1:31:07 12-in one which is like this little baby
1:31:09 screen. It's not. It's decent, but it's
1:31:12 not as like, you know, overbearing as
1:31:15 the uh
1:31:18 as the 16 in. And you can get a pretty
1:31:22 baller machine with a screen on it. So,
1:31:24 you can put it anywhere. Like, I know
1:31:26 you don't need it. you you don't need to
1:31:28 pay for the screen, but if you get the
1:31:29 14 inch, it's
1:31:33 it's not that much more to have the
1:31:35 screen. Um, okay. Let's see.
1:31:39 Caveats whenever you're ready. Still to
1:31:41 build preview cards for the design
1:31:42 system tab, product surfaces. Okay. Are
1:31:47 those
1:31:51 design files?
1:31:54 Oh, design system. Where's Oh, here we
1:31:58 go.
1:32:02 Yeah, I think these are the right font.
1:32:03 They are. Wow. Very cool.
1:32:09 Free PopSockets at the booth.
1:32:16 Okay, now let's export as PDF.
1:32:21 Save as PDF.
1:32:31 Did it do it?
1:32:34 No.
1:32:36 Writing styles.
1:32:48 Real people. Real stories scaled.
1:32:55 See how pharma and health systems
1:32:57 capture compliant onbrand patient
1:32:59 videos. I wonder where I got all that
1:33:01 [ __ ] Oh.
1:33:04 Huh.
1:33:11 Oh, cool.
1:33:13 Look at that.
1:33:21 Save as PDF.
1:33:35 Huh.
1:33:42 So, these are definitely designs.
1:33:44 They're not great, but you know,
1:33:52 All right.
1:33:54 Kind of cool.
1:33:59 Like this one's got a bunch of problems
1:34:00 with it.
1:34:05 All right, let's have it. Um,
1:34:10 let's say, um,
1:34:13 I want you to
1:34:18 generate screens for a
1:34:28 teleprompter app.
1:34:33 Let's just see what it does because I
1:34:34 didn't give it any any interface stuff.
1:34:53 Here's the thing about the Todd the the
1:34:56 M5 versus the M4 M4.
1:35:02 The architecture of the M4 is that
1:35:05 there's one AI accelerator in front of
1:35:07 the GPUs.
1:35:10 And in the M5 there's a there's an AI
1:35:13 accelerator per GPU.
1:35:16 Um
1:35:19 I don't know the difference that makes.
1:35:21 I assume it make makes a difference.
1:35:23 Just in general with Max, my philosophy
1:35:25 has always been buy the one the farthest
1:35:28 the farthest to the right you can
1:35:33 because you know it'll last you a year
1:35:37 longer than than you know if you get the
1:35:40 M4 because because what Apple's going to
1:35:43 do is later this year they're going to
1:35:44 they're going to come out with a bunch
1:35:45 of machines that are just inference
1:35:47 beasts. So, anything we buy right now,
1:35:50 we're going to look at what comes out in
1:35:52 the fall. We're be like, "Oh, [ __ ] I
1:35:53 should have waited."
1:35:57 So, might as well have an M5 because I'm
1:35:59 sure those will be some sort of crazy
1:36:01 ass M5 Ultra or something like that.
1:36:05 Um, teleprompter. Oh, wait. Who's this
1:36:08 for? The Story Minds mobile capture app.
1:36:12 platform, iPhone,
1:36:18 iPad landscape.
1:36:40 How does the script get in there?
1:36:47 How does the prompt scroll?
1:36:56 All right.
1:37:00 Reviewing the draft design system. I
1:37:03 wonder if this is going to go look at
1:37:05 storyvine.com and get screenshots of our
1:37:08 app. That would be interesting.
1:37:11 I doubt it, but we'll see.
1:37:15 I bought a Neo for casual use and the
1:37:18 13-inch screen is tolerable.
1:37:20 Oh, that's interesting. Sea slug. Yeah,
1:37:22 I like the the 14 I'm really I'm really
1:37:25 happy with the 14-inch MacBook Pro.
1:37:29 Like the screen is super clean. super
1:37:32 clear, super fast.
1:37:36 Um, the keyboard feels nice. It's funny,
1:37:39 it's got really teeny tiny speaker um
1:37:42 channels,
1:37:44 but the the sound is really good on it.
1:37:47 Like the engineering on the the MacBook
1:37:50 Pro right now is really tight. And I
1:37:52 know they're coming up with a new design
1:37:54 in the fall, but this this form factor
1:37:58 is is really impressive.
1:38:01 I saw the Neo at Costco. Yeah, the Neo
1:38:04 is basically just on a mobile on a on a
1:38:07 mobile chip, right?
1:38:09 The Neo has side speakers that are quite
1:38:11 surprising. Yeah, like when I when I
1:38:14 play audio on like these little teeny
1:38:16 tiny things should not sound as good as
1:38:18 it sounds. Like I hear stereo
1:38:20 separation, I hear bass.
1:38:23 It's pretty wild. It's pretty wild what
1:38:25 they've done. Apple, I got to say, like
1:38:28 I've been making fun of Apple now for
1:38:30 what, three years straight. That just
1:38:32 how did they miss the [ __ ] boat on
1:38:34 this? And Apple's going to sneak in
1:38:36 through the back door
1:38:39 by building the hardware cuz what's
1:38:42 going to happen is this this we we
1:38:43 talked about this today in office hours.
1:38:46 Um
1:38:50 it it's it it seems pretty clear that
1:38:53 starting with Mythos
1:38:55 um the Frontier model companies are
1:38:58 going to stop releasing everything they
1:39:00 build and you're going to need special
1:39:02 access to get access to the to the most
1:39:05 powerful models from the frontier model
1:39:07 companies. So us plebs
1:39:10 are not going to have access to all the
1:39:12 cool toys anymore, right? So, the cool
1:39:14 toys are going to go to the government
1:39:15 and to big business
1:39:18 and they're all going to go play their
1:39:19 [ __ ] game.
1:39:21 And what's going to happen is China will
1:39:24 keep providing open- source models and
1:39:26 then some of these guys will take you
1:39:29 know powerful pieces of what they do and
1:39:31 drop that down to open source and we'll
1:39:35 pull all these models increasingly to
1:39:39 our local machines and we'll be doing
1:39:41 all the inference locally on our local
1:39:42 machines. So having a badass like Mac
1:39:45 Mini or MacBook Pro that can run these
1:39:48 machines like Quen 3.6 6 came out today
1:39:52 and Brent Peterson on the on the office
1:39:55 hours said that Quen 3.6 is the first
1:40:00 open-source model that runs OpenClaw
1:40:03 well 3.5 didn't. He said like literally
1:40:07 a month ago Brent Peterson said do not
1:40:10 run open claw locally. Don't buy
1:40:12 yourself a good machine. The the local
1:40:14 models just aren't aren't there. They
1:40:16 suck. And today he said 3.6 6 came out
1:40:20 and it works and it's good. So, so we're
1:40:24 now at a point where the hardware is
1:40:26 powerful enough and the models are
1:40:29 efficient enough that we can run local
1:40:31 models that are good enough to do some
1:40:33 of this powerful work locally without
1:40:36 having to go to open AI. I think that's
1:40:38 where things start to move over the next
1:40:40 two to three years because the big
1:40:42 frontier model companies, they're going
1:40:44 to be like, "Well, this is this AI is so
1:40:46 powerful that that you normal people
1:40:48 can't have it. Only only governments can
1:40:51 have it." Kyle, is your is your Mac M
1:40:55 laptop have a touchcreen? It does not.
1:40:58 So, the touchscreen is coming in the new
1:41:02 MacBook Pro that's going to launch in
1:41:03 the fall.
1:41:05 Is going to have a a dual stack OLED
1:41:08 screen
1:41:10 um that's just basically like super
1:41:12 bright and super contrasty and it's
1:41:15 going to be touchscreen. They're
1:41:16 completely redesigning the architecture.
1:41:18 So the MacBook that I just bought is the
1:41:21 last of the current architecture
1:41:24 which has been it's been this
1:41:26 architecture basically the chassis has
1:41:28 been the same since 2022.
1:41:33 But what I like about that is it is like
1:41:39 there's no fluff on it. Like it is just
1:41:42 a solid machine. The new one's probably
1:41:45 going to be janky because it's a whole
1:41:46 new architecture, right? So, they'll
1:41:48 make mistakes on that one. It'll be the
1:41:50 second or third generation that that
1:41:52 gets really tight. But the in the fall,
1:41:55 the MacBook Pros will have touchscreens,
1:41:58 which Steve Jobs is flipping in his
1:41:59 grave. He said we should never have a
1:42:01 touch screen. You're in different modes.
1:42:04 Uh, let's see. Question for you. Um,
1:42:06 wait, wait, wait. What's this? What's
1:42:08 this done?
1:42:10 Welcome.
1:42:12 Where's my screens?
1:42:22 Your design is ready. Your feedback will
1:42:24 improvement. New design using this
1:42:27 system.
1:42:29 What
1:42:36 should the script source be? Um, you
1:42:40 decide.
1:42:42 I just want to I just [ __ ] I just want
1:42:47 to
1:42:49 There's a good typo. I just want to see
1:42:53 your UX chops
1:42:57 right now.
1:42:59 Send
1:43:22 Okay, that's
1:43:24 that's doing its thing. Um, let's go
1:43:27 back to notebook LM. Does this this is
1:43:30 still generating a video. That takes a
1:43:32 while.
1:43:36 D.
1:43:40 So, here's the deal with claw design.
1:43:43 Um,
1:43:45 this is a very, okay, it looks like what
1:43:49 what is going to be an incredibly
1:43:51 powerful tool that's going to take all
1:43:55 of us some time to understand the quirks
1:43:58 of.
1:44:01 So, I would encourage you cla.ai/design.
1:44:04 And if you have a Claude account, and
1:44:06 probably even if you don't, just go
1:44:08 start playing with it. Upload your
1:44:10 logos, upload your brand standards,
1:44:12 point it to your GitHub repositories if
1:44:14 you've got front-end stuff in there. Let
1:44:17 it learn your design
1:44:20 stuff
1:44:24 and then
1:44:26 we will we will learn together on this
1:44:28 one. But this seems like a new paradigm.
1:44:33 And when you export,
1:44:35 yeah, I can export to Canva. Send to
1:44:38 Canva.
1:44:40 So if it gives me a shitty layout, I can
1:44:43 export it as a as a PowerPoint.
1:44:47 I can export it as standalone HTML. I
1:44:50 can hand it off to Claude Code. Oh,
1:44:52 that's really cool. Okay, so if you come
1:44:54 over here and you do screen designs for
1:44:56 an app that you really like, you just
1:44:58 hand them to Claude Code and say, "Write
1:45:00 the back end."
1:45:02 It's pretty [ __ ] slick. This is good.
1:45:04 This is good, people.
1:45:07 This is good.
1:45:10 All right, that's still working. This is
1:45:12 still working. Everyone's still working.
1:45:18 Thoughts, questions? Anyone still awake?
1:45:29 I'm sorry, Kyle. I fell asleep.
1:45:33 Hey Kyle, last night I was watching you
1:45:35 and I dropped the phone on my face. Now
1:45:38 I have a band-aid on it. Could I get a
1:45:41 refund? Give us your craziest prediction
1:45:44 for what these tools will do in the
1:45:45 future.
1:45:47 Todd Waller's here. Yeah, but you're
1:45:50 shopping for new computers. That's
1:45:51 exciting. Cam is just listening.
1:45:55 Job said vertical screens cause arm
1:45:57 fatigue over time. your arm will want to
1:45:59 fall off.
1:46:01 Yeah. Um,
1:46:04 craziest predictions for the future. Um,
1:46:12 here you go, Kelly. This is the big one.
1:46:18 I think that we're in a quaint era
1:46:21 where
1:46:23 where human software engineers are
1:46:27 designing software for human beings to
1:46:30 pretend that they still need to be
1:46:35 in the loop.
1:46:38 And I think that where these tools
1:46:41 ultimately go
1:46:43 is that the humans are going to train
1:46:48 the humans are going to train the AI
1:46:52 what we want and how we design
1:46:56 and then at some point they'll just
1:46:57 remove the front end the human front end
1:47:00 from these things like maybe there's a
1:47:02 dashboard where you log in and approve
1:47:04 things
1:47:06 but you're probably not going to
1:47:08 actually be doing all this
1:47:11 right agents are probably going to be
1:47:13 using all these different tools
1:47:16 time All
1:47:17 right, let me jump over to look what I
1:47:19 made
1:47:21 before we go. Um,
1:47:27 I know Silverf Fox has a song in here.
1:47:29 We So, we should go listen to that, but
1:47:31 I got to share different.
1:47:39 So, I'm over on over on the AIO.
1:47:50 Oh yeah. So, so here's a cool thing. Um,
1:47:55 let let me let me dig through here and
1:47:57 then I'll do that. Brandon, give me five
1:47:59 minutes.
1:48:01 Okay. So, Silver Fox has a cool song in
1:48:04 here. Oh, those are cool. Dr. J. Oh,
1:48:06 that's a Sunno song, too. All right. So,
1:48:08 we'll listen to a couple of Sunno songs.
1:48:11 Have you read the AI news today?
1:48:15 Share this tab instead.
1:48:32 I woke up this morning to a glowing
1:48:34 screen. The headline said the smartest
1:48:36 thing we'd ever seen had solved the
1:48:39 proof that stumped us for a hundred
1:48:41 years. And somewhere in a boardroom
1:48:51 >> getting dressed for school down the
1:48:53 hall. I turned the volume up and let the
1:48:56 falldiled
1:48:59 and turned away. I poured my coffee and
1:49:02 thought, "What do I say?"
1:49:06 And the world keeps spinning faster than
1:49:10 it did before.
1:49:13 And I'm standing at the window wondering
1:49:16 what we're spinning toward. Have you
1:49:19 read the AI news today?
1:49:23 Agn's
1:49:28 what it seems.
1:49:31 Have you read the
1:49:34 news today?
1:49:38 That's really cute. That's really good.
1:49:39 Um,
1:49:41 so wait, let's see. So, the comments
1:49:44 here are pretty funny. Todd Waller, my
1:49:46 open claw is Chuck. I don't have a cool
1:49:47 story about why. Oh, yeah. By the way,
1:49:49 go into um where did I put it? Was it in
1:49:53 community feed?
1:49:59 Yeah. Go to into community feed. What
1:50:01 did you Why did you name What's the name
1:50:04 of your AI buddy? whether it's chat GBT
1:50:06 or OpenClaw, just go put that in there.
1:50:09 Um, let's go back to look what I made.
1:50:13 So, that was that was uh Silver Fox's
1:50:15 song. Here's Dr. Jay's How Fast.
1:50:38 I laced them up like it was sacred
1:50:40 ground. Brand new souls, not a scar for
1:50:43 sound. Kitchen floor turned Olympic
1:50:45 track. Heart said go. But the world held
1:50:48 back. I hit the hallway at a lightning
1:50:50 pace. Felt like time. Had a different
1:50:53 face. But no one cheered. No one laughed
1:50:55 out loud. Just quiet walls in a distant
1:50:58 crowd.
1:51:00 Did wonder get left behind? Who decided
1:51:06 joy had an age or line? Not a single
1:51:10 soul asked how fast I could run. Didn't
1:51:14 say let's see. Didn't say that looks
1:51:17 funwhere
1:51:20 between
1:51:26 the practical stuff.
1:51:29 I was built for the race, for the wild,
1:51:31 for the blast, for the question that
1:51:33 matters.
1:51:41 >> Nice. Love it.
1:51:43 Beauty, beauty, beauty. And then what do
1:51:45 we have here? We've got LDG. Back in the
1:51:49 70s, your uncle LDG had a band called
1:51:52 Infinity
1:51:53 because I was lazy at names.
1:52:00 All right. So, what's this? So, this is
1:52:02 I rummaged through some stuff and found
1:52:04 a copy of this old song on cassette tape
1:52:06 and digitized it, uploaded it to Suno
1:52:08 and re-engineered it. Now, I'm
1:52:10 desperately looking around my uh house
1:52:13 to see if I saved any other songs. So,
1:52:15 cool. So, this is this is a song that
1:52:17 existed that just got re reunized.
1:52:21 That's pretty cool.
1:52:52 It wasn't long ago. know that I was
1:52:54 happy being blind.
1:52:58 No news was very good news. Out of sight
1:53:01 was out of mind.
1:53:04 But I grew tired of me. Life had spilled
1:53:07 a seed that grew and grew and grew into
1:53:12 a monster of desire. I was thrown into a
1:53:16 rude awakening.
1:53:21 Do you know how much we love you? Do you
1:53:28 know how much we care?
1:53:33 Do you know how much we love you? Even
1:53:39 though you're
1:53:45 >> nice. That's awesome.
1:53:49 Friaking amazing. Um, sea slug of doom.
1:53:53 Crazy forecast. We will all be sitting
1:53:55 in the confessional booths like
1:53:59 like thx 1138. AI will be soothing us
1:54:02 and showing us no indication of actually
1:54:04 listening. By the way, I talked about
1:54:07 this on uh on office hours today. If you
1:54:11 have not played with boardy.ai
1:54:14 be it's not even really played with.
1:54:18 go to boardy.ai
1:54:21 and start interacting with it and do the
1:54:25 onboarding phone call.
1:54:29 And so what what Bordy.ai is is
1:54:34 Anthem Rock. Yeah, that was definitely
1:54:36 Anthem Rock. Um what Borty.ai is is a
1:54:39 super connector that will introduce you
1:54:41 to people that you're looking to meet.
1:54:43 So do the onboarding call because it is
1:54:47 impressively.
1:54:51 It doesn't feel like AI.
1:54:55 The Australian accent helps, but it but
1:54:57 it's it's it's really well done. He made
1:55:01 a connection with me on LinkedIn. I had
1:55:03 no idea it was an AI connection. Yeah.
1:55:06 So, what happens is you tell Bordy all
1:55:09 about your business and who you are and
1:55:11 what your values are and what you're
1:55:12 looking for and who you want to meet.
1:55:15 And then you can just keep talking to
1:55:17 him. He starts emailing you and you
1:55:18 email back and forth with him and you
1:55:20 educate him and he'll say, "Hey, I found
1:55:22 someone that you might be interested in
1:55:24 meeting." And it'll make the intro. And
1:55:26 so, I've now had four human connections,
1:55:31 human conversations about StoryVine. And
1:55:34 then I just I just accepted one
1:55:38 two days ago and he made the intro and
1:55:41 then the guy Ash that he introduced me
1:55:44 to said sure set up a time. Um so so
1:55:47 anyone that he's connecting you with has
1:55:50 also gone through the onboarding
1:55:52 process. So, it's a really interesting
1:55:54 kind of professional dating app that's
1:55:57 got a lot of friction in it and a lot of
1:56:00 high context where you're giving this
1:56:03 thing, you know, what to what to play
1:56:05 with. Kyle, did you do it for your
1:56:07 musical, too? Well, I'm just starting.
1:56:09 So, what I I started with Story Vine and
1:56:11 then I just added in my speaking stuff
1:56:13 and now I'm going to do it for my
1:56:14 musical. So, that's a it's a really good
1:56:16 call. Um, so, so yeah, I'm I'm starting
1:56:19 to take this thing more seriously
1:56:21 because it's starting to make actual
1:56:22 connections. It's wild.
1:56:25 All right, so last thing we want to do
1:56:26 tonight, I want to bring up um, producer
1:56:29 Brandon is down in Florida with Rick
1:56:33 McCaulay. So, they're in the same
1:56:35 household. Hey guys.
1:56:36 >> Hello.
1:56:37 >> Yeah. What's happen in Florida?
1:56:40 >> We're gonna do a hackathon tomorrow for
1:56:42 12 hours. We're going to change the
1:56:44 world, man.
1:56:46 >> Again. again and always you guys are are
1:56:48 uh killing it. So it's a uh what's the
1:56:51 is there a theme of the hackathon?
1:56:53 >> So we so funny story we thought it was
1:56:55 going to be healthcare and we had this
1:56:57 whole health care plan about curing
1:56:59 cancer and it was be amazing
1:57:01 on my I was on the plane on the way
1:57:03 down. I got an email. Hey, we switched
1:57:05 things up. It's enterprise now. So I'm
1:57:07 like oh crap. So now we're going to
1:57:10 >> what's enterprise even mean? like
1:57:13 enterprise AI, like how to use
1:57:14 enterprise
1:57:17 workforce. So, I don't know if they
1:57:18 changed sponsors or what happened, but
1:57:21 the healthcare thing went out the
1:57:23 window. So, we're uh we're actually
1:57:25 borrowing a page from your book um with
1:57:28 the great repurpose and Rick's upskill
1:57:30 USA. We've got a real solid plan. I
1:57:33 can't wait to talk about it next week
1:57:34 after we win.
1:57:36 >> Yes.
1:57:36 >> Yeah. You know, you know what would be
1:57:38 what would be a good thing to put in
1:57:40 there if you don't already have it in
1:57:41 there is um how to identify the people
1:57:45 to keep now that are going to be
1:57:47 valuable in the future. That whole thing
1:57:49 of who's valuable today might not be the
1:57:51 person valuable in the future.
1:57:53 >> So finding, you know, how do you find
1:57:56 your secret cyborgs? How do you find the
1:57:58 people with curiosity and adaptability
1:58:00 and like all those attributes that are
1:58:02 not the hard skills? um something in
1:58:05 there because that to me feels like a
1:58:07 bit of voodoo voodoo magic where
1:58:11 >> if you're not thinking long and hard
1:58:13 about what are the skills that make
1:58:15 someone good at AI you're just going to
1:58:17 be thinking like oh Fred's the most
1:58:19 valuable because he's really good at
1:58:20 whatever skill X is. So we're going to
1:58:23 keep Fred but Fred might be really rigid
1:58:26 and not curious and not adaptable. So as
1:58:29 things change Fred's going to be in a
1:58:30 nightmare, you know.
1:58:33 Yeah. And so without spilling too much
1:58:35 candy in the lobby, what we're aligning
1:58:38 on is judging the criteria that we're
1:58:41 aligning on is what parts of your job do
1:58:45 you want to get rid of with AI?
1:58:47 >> What parts do you want to automate with
1:58:50 AI and augment with AI? And what parts
1:58:52 do you want to own? This is the part
1:58:54 this is my secret sauce. this is the
1:58:56 part that I'm the AI is never going to
1:58:58 touch because we've talked to a lot of
1:59:00 creatives where they like I'll let AI do
1:59:03 that but I won't never let it do that
1:59:06 and we just scale that up to the
1:59:07 enterprise level.
1:59:08 >> So it's taking all the junky stuff and
1:59:10 take the noise out of the signal and
1:59:12 then amplify your humanity. So you get
1:59:14 rid of the answering emails, you
1:59:15 automate the stuff that's taskbased, you
1:59:18 know. So it's basically skills is like
1:59:20 the super word now. I think agents is
1:59:22 last weekend, right? But this is now
1:59:24 skills. So you get skills like social
1:59:27 media management. You just that's a
1:59:28 skill set. Build a website. Boom. So you
1:59:31 just say put that in the way and then I
1:59:33 design what I want to get done and then
1:59:35 it goes and does it. So agenting skills
1:59:39 are are programmable black boxes that
1:59:41 you put in a workflow.
1:59:43 >> Yeah, that's really cool.
1:59:44 >> Where the where the AIable parts of your
1:59:47 workflow are and then let lets you
1:59:49 amplify you which is spending more
1:59:51 customer time.
1:59:52 >> That's really cool. That's really cool.
1:59:54 And if you could do some sort of thing
1:59:56 where it's a timeline of like,
1:59:58 >> you know, have someone put in what their
2:00:00 job looks like today.
2:00:01 >> Here's what your job's going to look
2:00:02 like three years from now, but then what
2:00:04 are the milestones to get there?
2:00:07 So, so you go through and do your
2:00:08 signature work on you set up the agent
2:00:11 places where you're going to interact
2:00:13 and where you're not the audit and the
2:00:15 redress.
2:00:16 >> Then it tells you what efficiency you
2:00:18 gain. So, you just gain 12 hours per
2:00:20 week.
2:00:21 >> You get Fridays off.
2:00:23 >> That's the goal.
2:00:24 >> And you double the productivity of your
2:00:26 work product
2:00:27 >> and you take away the junk you never
2:00:29 wanted to do in the first place. Right.
2:00:31 We all need to do that in life, right?
2:00:33 >> So, here's one thing I'll I'll I'll add
2:00:35 in. Um,
2:00:37 >> don't just make it about efficiency,
2:00:39 >> okay?
2:00:39 >> Make it make it about here's new things
2:00:42 you're going to be able to do in in your
2:00:45 with your skills. So, so one piece is
2:00:47 you make what you currently do more
2:00:49 efficient, but then you also discover
2:00:51 here's new things I can do with those
2:00:53 hours that I gained.
2:00:55 >> Yeah, that's it. For me as a teacher,
2:00:57 it's can I get the student to tutor
2:00:59 themselves and then do that all night
2:01:01 long and come in. Then I find out what's
2:01:04 the eeky guy for that student
2:01:06 >> that makes them superpowered and then
2:01:08 send them on their froto journey and I'm
2:01:10 just Gandalf on the side and watching it
2:01:11 all happen, baby.
2:01:12 >> Yeah, that's really cool. 4 day work
2:01:14 week has people excited. Cam Katkin's
2:01:16 excited about that.
2:01:18 >> I think I think Brandon just quoted a
2:01:20 Meatloaf song. That's hilarious.
2:01:24 So, speaking of trying new things and
2:01:27 thinking about where your skills are
2:01:28 taking you, we're gonna get off because
2:01:31 Kyle, I know you have some announcements
2:01:32 about next week in the AI learning lab.
2:01:35 >> I do. So, yeah. Okay, cool. Thanks,
2:01:37 guys. Great seeing you. Good luck
2:01:38 tomorrow.
2:01:39 Um, so next week I am in DC on So, so
2:01:44 we'll be here Monday night next week,
2:01:47 but then Tuesday, Wednesday, I
2:01:50 definitely can't be here.
2:01:54 Friday, I definitely can't be here.
2:01:56 Thursday's a tossup. So, so next week is
2:02:00 probably going to be Monday and maybe
2:02:03 Thursday. Um, and then the following
2:02:06 week I'm in Anaheim for social media
2:02:08 marketing week and that is also going to
2:02:11 be play it by ear, but probably maybe
2:02:14 one night or two nights that week as
2:02:17 well. So, the next two weeks are going
2:02:19 to be really thin. So, you're going to
2:02:20 be on your own.
2:02:22 Tomorrow, Nikki Weiss is doing an LOL on
2:02:27 building her app and what she learned
2:02:29 along the way, like things not to do or
2:02:31 things that broke. So, that should be
2:02:32 really interesting and powerful. She's
2:02:34 she's a powerful, passionate woman who
2:02:37 is going to teach what she's been
2:02:38 learning. That is literally the
2:02:41 definition of of generous leadership is
2:02:43 is share people what's working, what's
2:02:45 not working, things like that. Um,
2:02:50 weekend homework, I would go play with
2:02:53 Claude Designer. It's it's a new
2:02:55 paradigm. It's a new something. It's not
2:02:59 um it's not an LLM as we as we know it.
2:03:03 It's an LLM in conjunction with a design
2:03:06 system. So, go play with that. Start
2:03:09 understanding what that is and what it
2:03:10 looks like. If been if you've been using
2:03:12 AI tools in Figma, it's probably
2:03:14 similar, but it it doesn't have the
2:03:18 whole world of of templates that you go
2:03:20 pick what you want. I want to do a card.
2:03:22 I want to do this. I want to do that.
2:03:24 Again, we're just in this world where
2:03:25 you can just ask for [ __ ] now and it
2:03:27 will go invent it. So, I would go
2:03:30 explore with that all weekend. And then
2:03:32 finally, especially if there's there's
2:03:34 some people who here who are back that
2:03:36 used to be here,
2:03:39 please go into the AI salon mighty
2:03:41 network. So So Brandon, if you could pop
2:03:43 up the URL, community.thesalon.ai
2:03:47 and go look for the ambassador
2:03:49 application. So we're looking we're
2:03:51 starting an ambassador program. So
2:03:54 people who um can represent the AI salon
2:03:57 at local events and online and within
2:04:01 the AI salon community like welcoming
2:04:03 people, things like that. But we also if
2:04:06 you're an AI ambassador, we're doing we
2:04:09 want to do things that amplify you that
2:04:11 make you more visible, tell your
2:04:13 stories, capture your stories and
2:04:14 promote you and and do things like that.
2:04:17 So the ambassador program, we're just
2:04:19 designing it right now. We just want to
2:04:21 understand who's interested in being an
2:04:23 ambassador. So if you love the AI salon
2:04:26 and you want to be an ambassador and you
2:04:27 want us to help market you and who you
2:04:30 are, join join the ambassador pro or
2:04:33 just you know fill out the form that
2:04:34 you're interested in it in it and as we
2:04:38 design it um we'll bring you into the
2:04:40 conversation. All right. Beautiful. It's
2:04:43 in salon announcements by the way in
2:04:45 that channel in salon announcements.
2:04:48 Okay, everyone. Fantastic. Hope you had
2:04:51 fun tonight. Um Oh, wait. We have one
2:04:53 more thing. Let's go look and see if
2:04:56 Yeah, the uh the the next
2:05:00 um
2:05:02 notebook LM of the AI learning lab
2:05:07 has has rendered. So, let's see if it
2:05:10 sucks.
2:05:13 That's such a pessimistic Let's see if
2:05:16 it sucks. That's that's a very Gen X
2:05:18 kind of way of I'll assume it sucks
2:05:21 until it doesn't. Prove to me you're not
2:05:23 going to suck.
2:05:26 Early generative AI flooded the internet
2:05:29 with a very specific kind of highly
2:05:31 produced absurdity. You could command a
2:05:34 computer to write a country song about
2:05:36 purple pickles or render a
2:05:38 hyperrealistic clip of a dog cruning
2:05:40 into a microphone. Kyle Shannon launched
2:05:43 the AI learning lab as a nightly live
2:05:46 streamed project to track these bizarre
2:05:48 tech updates. But over three years,
2:05:50 these reviews transformed into something
2:05:52 else.
2:05:52 >> Wait, sea slug of doom. Like the Rolling
2:05:54 Stones used the Hell Angels, Hell's
2:05:56 Angels as security. That kind of
2:05:58 ambassador. That's exactly what we're
2:06:00 looking for. Yes.
2:06:01 >> A support group for a quiet creeping
2:06:04 human crisis. To understand how this
2:06:07 technology forces us to redefine our
2:06:09 purpose, we have to map out the five
2:06:11 distinct eras of the last 3 years. In
2:06:14 mid 2023, the arrival of chat GPT meant
2:06:18 anyone with a keyboard possessed the
2:06:19 baseline ability to write code, draft
2:06:23 professional essays, and analyze complex
2:06:25 data sets. This graph represents the
2:06:28 traditional path to technical mastery.
2:06:30 But look at how the 10,000hour rule is
2:06:33 suddenly intersected by a sharp drop to
2:06:35 near zero time. Skills that used to take
2:06:38 years to acquire were democratized in an
2:06:40 afternoon. While this felt like a period
2:06:43 of newfound superpowers, it also began
2:06:45 the process of turning specialized
2:06:47 human.
2:06:48 >> See sl of doom. Dang vocal fry. I know
2:06:51 it's pretty bad, isn't it?
2:06:52 >> Expertise into a cheap, abundant
2:06:54 commodity. By 2024, text evolved into
2:06:58 highfidelity audio and video. Multimodal
2:07:01 platforms like Sora, Sunno, and
2:07:03 Mitjourney allowed users to spin up
2:07:05 entire environments from a single
2:07:07 prompt. These tools weren't perfect.
2:07:10 They were often surreal and glitchy.
2:07:12 Kyle and his community learned to
2:07:14 embrace the jank, the philosophy of
2:07:16 accepting these hallucinations as a
2:07:18 playground for experimentation rather
2:07:20 than a technical failure. Solo creators
2:07:23 used this playful mentality to stitch
2:07:25 these imperfect tools together. A single
2:07:27 person could now build massive
2:07:29 >> All right, that one's that one's in the
2:07:31 neighborhood of usable. So, um, cool,
2:07:35 Groovy. I know it's late. Um, have a
2:07:38 fantastic weekend, everybody. Um,
2:07:40 appreciate you being here and hanging
2:07:42 out. Uh, have a good night. Have a good
2:07:44 weekend. And I will see you Monday. And
2:07:46 then next week's going to be light. So,
2:07:48 plan on stuff other than this.
2:07:52 You deserve it. Bye.
2:07:56 Let's see.
2:07:58 Ending there. Oops.
2:08:01 It's fast.