7/18/2025 - AI Agents: A New Frontier in Human-Computer Interaction

Description

Chapters

Transcript

0:11 Woohoo!
0:14 [Music]
0:21 Find a night
0:25 every
0:28 [Music]
0:35 like me.
0:38 [Music]
0:40 One of life's little
0:43 mysteries.
0:46 You all right there? You're choking out.
0:48 You're choking out. You gonna be all
0:50 right.
0:52 Somebody didn't do their vocal warm-ups,
0:54 did they?
0:55 [Music]
1:07 So tonight
1:09 I ask
1:12 the stars above.
1:16 How did I ever hear your love?
1:23 What did I do?
1:26 What did I say
1:28 [Music]
1:30 to turn your rain jealous
1:34 my way?
1:36 [Music]
1:43 I'm the guy that never I learned to
1:47 dance,
1:50 never even got one second glass.
1:55 [Music]
1:58 across a crowded room that was close
2:01 enough.
2:04 I could look, but I could
2:08 never touch
2:12 [Music]
2:14 the stars above.
2:17 All right, good people. Happy Friday
2:19 night. Date night. Friday night. Date
2:21 night.
2:24 We are Sans's producer Brandon
2:26 [Music]
2:45 [Music]
2:51 in a westerly.
2:53 direction.
2:57 This car is my train.
3:02 I've been driving. I've been wondering
3:06 what it is I'm running from again.
3:12 Champ, you going to sing? I feel like an
3:15 80-year-old man
3:18 holding on to 29.
3:23 Up ahead on that horizon
3:27 is a California line.
3:30 [Music]
3:38 Up ahead of trucks carrying a wide load.
3:41 Free house. Good night.
3:45 Cute little front door into two windows.
3:47 My love
3:49 ain't sure whether a cry should last.
3:53 You see, I broke a home on myself once.
3:58 This I stumbled to that dog.
4:02 I write a note by the dogs light.
4:06 Said, "Don't you come around here
4:11 anymore."
4:14 Well, I've had enough
4:17 of this freedom of
4:21 Oh, losing Champ. Champ's walking. He's
4:24 had it. He's had enough. He's like, I
4:28 can't take this. Learn a new [ __ ]
4:30 song. learn a new [ __ ] song. It's
4:34 what he said. That's what he said. I can
4:36 talk dog. You know how I can talk dog? I
4:39 got I got access to the new dog talking
4:41 AI. You just put the chip in their head,
4:45 a little uh injection in their neck,
4:47 staple something to their ear, and now
4:49 you just talk to him. And boy is he
4:51 pissed. He does not like me.
4:55 You know, you know what's hard for dogs?
4:57 They they uh
5:01 they want to not do things, but because
5:03 they're dogs and the Pavlovian thing,
5:07 they want to eat
5:09 against their will. They'll come sing
5:11 with us.
5:13 Yeah. Anyway, I don't know. I got access
5:16 to that, so I'm kind of a big deal, you
5:19 know.
5:21 Jeff Fran Flanigan getting a shout out
5:23 from Ocean. Hey, Ocean. What's
5:25 happening? Haven't seen you in a while.
5:26 Hope you're well.
5:30 [Music]
6:11 Oh man.
6:17 Happy Friday night date night everybody.
6:19 Hope you got your nachos going,
6:23 your dirty microwave hot pocket.
6:28 [Music]
7:01 [Applause]
7:03 [Music]
7:15 [Music]
7:19 Oh, yes.
7:22 I've already made a preemptive cheese
7:23 run.
7:27 Trying out mild drinking for Friday
7:29 night. I like it. No hot pockets,
7:31 though. That's a shame. That's a shame.
7:34 Jeff Flanigan realized Ocean said hello.
7:36 Felt like a heel because he had stepped
7:39 away from the computer. Wait a minute.
7:42 You all are not watching this like
7:46 like like a new episode of Seinfeld in
7:48 the '9s. Come on, people.
7:51 [Music]
8:16 I don't know if I'm feeling
8:17 philosophical tonight
8:20 or demoy
8:22 or ornery or hopeful or pessimistic.
8:29 I think tonight will be a reflection of
8:31 all of you.
8:34 Well, hopefully every night is a
8:36 reflection of all of you,
8:38 but sometimes I have an agenda. I just
8:40 I'm feeling agenda free. So, ask away.
8:43 Ask if you're new here. Hello.
8:46 My name's Kyle Shannon. This is my
8:49 trusty guitar. Uh I got a dog over there
8:52 named Champ. Got a family elsewhere. And
8:56 then uh we talk about AI stuff here. Uh
9:00 I am wholly wholly
9:02 100% wholly unqualified
9:06 to teach anything
9:08 but I do it anyway.
9:11 Think of it as as as mansplaining as a
9:14 service. Welcome.
9:19 Like chat GBT. I will confidently answer
9:22 you even if it's wrong.
9:30 The other thing I like to do is
9:31 entertain myself. Uh, not really for
9:34 your entertainment, really just for my
9:36 own personal pleasure. So, it's a lot of
9:38 stupid jokes. It's a lot of repeated
9:40 jokes. A lot of jokes, a lot of things I
9:42 find funny, I'll just share with you
9:46 because I think they're funny. And if
9:49 you think they're funny or not doesn't
9:51 really matter.
9:53 The people that tend to stay here tend
9:54 to be people that like the same stupid
9:56 [ __ ] I do.
9:58 But if you have questions about AI, pop
10:00 them into the comments. Totally
10:02 unqualified. That's it.
10:06 [Music]
10:09 Only have to be 98% right. Yeah, if I'm
10:11 98% right, I beat Chat GPT. Kyle Kyle
10:15 Kyle was actually watching you live on
10:17 YouTube on my big TV screen since some
10:20 months. I'm sorry about that. Yeah, I
10:23 don't know how you explain that to
10:24 friends when they walk in and they're
10:25 like, "What are you watching?"
10:29 But hopefully you have no friends and
10:31 you don't have to deal with that. You
10:33 can just put me up on the jumbotron.
10:39 But I spent last night at a Holiday Inn.
10:44 It's Holiday in Express. Gota You got to
10:47 get your You got to get your marketing
10:48 memes right there, Brian.
10:52 Oh my god, that's funny. Um, all right.
10:59 Jeff Jeff Flanigan, I'm 99 and 4100s%
11:02 wrong.
11:06 [Music]
11:12 [Laughter]
11:14 I like it. I like it when my friends
11:16 watch with me and say, "Is he
11:18 qualified?" And I say, "Just wait.
11:20 You'll see."
11:25 That actually happened last week. That's
11:27 hilarious. Yeah. Like, why are you
11:29 watching this guy? Hey. Hey, Winston.
11:32 Yeah. Listen. No, I No, I listen. I love
11:36 coming over to your place. Um, the guys
11:39 and I were talking. We were just uh we
11:43 were wondering the I I know you gave up
11:46 on basic cable years ago. Uh
11:50 this fell you were watching on a the
11:52 what was it the about the AI stuff.
11:54 Yeah, I know. We know you're very
11:56 technical and you enjoy the technical uh
11:59 aspects. Uh uh
12:02 we're concerned for you for your for
12:05 your health for your mental health.
12:07 We're Yeah. Are are you okay? Because he
12:12 it
12:14 I I mean Yeah. No, you're good. Okay. Or
12:17 no. Hey, we just wanted to check in. We
12:20 just wanted to check in. Make sure
12:21 because we've never seen uh television
12:24 quite like that before. It was It was It
12:26 was uh It was interesting.
12:30 All right. All right. We're going to
12:31 just go then. We'll uh Yeah. No, we
12:33 can't do it next week, but maybe uh
12:35 maybe in August. Maybe in August we'll
12:37 come back and you know, you'll have
12:39 moved on there. No, you won't. Okay. All
12:41 right. No, that's good. That's good to
12:42 know. Good to know. All right.
12:45 H. Hey, Champy. What's up? You want out?
12:48 You want out? You want your cheese?
12:50 Hold, please.
13:10 [Music]
13:23 I did a demo advanced voice at a family
13:26 gathering and one of my grandkids said,
13:28 "That's just creepy.
13:31 You should show your grandkid the the
13:33 waifu from Grock.
13:36 They'll be like, "That's just creepy.
13:38 Can I go talk to her?"
13:46 Hey. Hey, Grandpa. Can I borrow your
13:48 phone again?
13:53 Hey, Danielle. What's happening? Oh,
13:55 what's going on? I was watching TV. Uh,
13:58 no. I think I think you're it was the
14:00 wiser move was the watching TV part. I
14:02 think you were good before. Uh nothing's
14:04 going on yet. We haven't really started.
14:06 I'm trying to figure out what the hell I
14:07 want to talk about tonight.
14:10 Um happy to show stuff. Happy to answer
14:12 questions.
14:14 Happy to wax on philosophic.
14:18 [Music]
14:23 We are definitely entering a new uh
14:28 We're definitely entering a new
14:32 Is era the right word? No, it's not an
14:36 era. This whole thing's an era. We're
14:39 entering a new phase.
14:43 We're entering a new phase.
14:47 [Music]
14:54 the the uh
14:58 the tools like Manis and Genspark and
15:01 the new OpenAI agent thing.
15:06 Um,
15:12 shit's about to get weird. Shit's about
15:15 to get weird.
15:17 because these tools are um
15:30 they're very raw right now and they
15:32 don't really work,
15:36 but if you look at what they're doing
15:37 and what's actually going on under the
15:39 hood,
15:41 they're I think they're going to start
15:43 to improve quite quickly. If you've been
15:45 paying paying attention to Gen Spark in
15:47 particular,
15:48 Gen Spark launches a new feature
15:51 probably about on average every two
15:53 weeks.
15:55 Um, and they're just getting better and
15:57 better and better and better. If you
15:59 haven't played with Gen Spark, it's
16:00 really worth you going and playing with
16:01 it. Paradigm. Uh, I don't know if we're
16:04 in a new Well, I guess we are in a new
16:05 paradigm. The new paradigm is we're
16:08 we're in a we're in a we're entering a
16:12 phase
16:13 where
16:18 how we interact with with
16:22 computers and data and the internet
16:27 is not going to be like we've ever done
16:29 it before. And I don't think any of us
16:32 really know what that means. I don't
16:34 think we we know how to use these tools.
16:37 I don't think we know when to use these
16:39 tools. Um I don't know if we know when
16:42 to trust these tools.
16:44 Um
16:46 I think a lot of people are going to
16:47 make a lot of mistakes and share [ __ ]
16:49 with these tools they shouldn't.
16:52 Um
17:02 yeah, I don't know. It's wild.
17:06 We're entering a paradigm where
17:08 intelligence isn't the bottleneck. Yeah.
17:14 Yeah. And and our relationship with
17:23 Yeah. I don't know. Anyway, if you don't
17:25 know what the [ __ ] I'm talking about,
17:27 I'm happy to go back and explain it. If
17:31 you do know what I'm talking about, Hume
17:33 AI had an update to EV3. I I actually Oh
17:37 [ __ ] You know what? I got a uh I got a
17:40 LinkedIn request from the CEO of Hume to
17:43 go check it out and I didn't do it.
17:44 Maybe that's something we can play with
17:46 tonight. Sorry, you probably addressed
17:48 this. I wasn't there. What's the
17:50 thoughts about YouTube stop paying for
17:52 anything with AI? Is it fair? Um, yeah,
17:55 I think it's I think it's absolutely
17:57 fair and I think it's I think it's so
18:00 there's there's some nuance. So, if you
18:01 don't know, about a week ago, no, two
18:04 weeks ago, YouTube said they're going to
18:07 demonetize any video that is basically
18:12 AI slop.
18:14 So, if it was just sort of generated out
18:16 of a machine,
18:19 um,
18:20 you're not going to be able to monetize
18:22 it. And so what that looks like is
18:24 channels that, you know, spew out
18:28 programmatic iterations of faceless
18:31 videos like, you know, 10,000 videos a
18:34 week or whatever. YouTube knows which
18:36 ones they are, right? What they also
18:39 said though is
18:41 we're not demonetizing all of AI videos.
18:44 We're just demonetizing that crap. And
18:48 if you have handcrafted, if you have a a
18:50 a chain of craft, right, where where
18:53 you've thought about stuff and maybe you
18:55 went to midjourney and you made images
18:56 there and you made this and then you
18:58 made a song over here and you stuck it
18:59 all together and you made a video that
19:01 tells a story, that's still monetizable.
19:04 If you make a music video that's really
19:06 good, that's still monetizable. Now,
19:08 clever
19:11 programmatic
19:13 AI assemblers will be able to replicate
19:16 well-crafted AI content, but I think
19:20 YouTube will still know if there's, you
19:22 know, a regularity to regularity to
19:24 uploads, things like that. I assume you
19:26 can appeal it. But what it starts to do
19:29 is the the YouTube decision starts to
19:34 say there's actually a difference in in
19:39 wellthoughtout
19:42 crafted AI productions and [ __ ]
19:46 And I think that's I think that's great.
19:48 I mean I think that's something that we
19:50 all need to discern. We you can look at
19:53 Tik Tok right now and go, "Okay, that's
19:56 AI news. That's AI influencer. That's
19:59 the, you know, crap crap crap crap." Oh,
20:01 and then you see Kelly Bosch and it's
20:03 all AI with AI music and it's [ __ ]
20:06 art,
20:08 right? You can tell the difference.
20:11 And so I think over time that's just
20:14 going to be part of the part of the
20:16 noise of the internet is there's going
20:19 to be a bunch of AI crap and then
20:21 there's going to be stuff that rises
20:22 above it and we'll gravitate to the
20:24 stuff that rises above it. The stuff
20:26 that moves us, the stuff where we have a
20:28 relationship with the creator. No, no
20:30 different than it is today.
20:32 Um,
20:34 this is new to me. Interesting. Yeah,
20:36 cur curation's important. Um Netflix
20:40 said today they've got one of their
20:43 movies that's live on Netflix right now
20:45 has AI generated scenes in it.
20:51 I that shouldn't come as a shock to
20:53 anyone.
20:56 Hollywood studios are looking for people
20:58 that know how to prompt [ __ ] into
21:01 existence. So, if you're getting good at
21:04 prompting V3 and
21:07 putting together midjourney images and
21:09 runway ML stuff and going to LTX Studios
21:12 and making things, there's there's stuff
21:14 to do there. Which, by the way, if
21:16 you're on LTX Studio or even if you're
21:18 not, they're doing a contest through
21:20 April 6th, I think.
21:23 Um, where first prize is five grand and
21:25 a trip to Germany.
21:28 [Music]
21:40 Is Digital Gods here? I don't know.
21:42 Winston's here. I don't I haven't seen
21:45 Lord Digital Gods, but I haven't been
21:46 looking at TikTok here for a minute.
21:49 April, did I say April? August. August
21:51 6th.
21:57 Uh, Daniel, Daniel, any of my irregulars
22:00 here? Yes, lots of irregulars here. Um,
22:04 the videos that just do AI voiceovers.
22:06 Yeah, those those take no talent. Yeah.
22:11 Now, you could argue, dude, what? Go
22:14 respond. Go respond to who? To what?
22:23 April fools.
22:27 [Music]
22:31 Digital gods usually watches on YouTube.
22:36 Oh, the CEO of Hume. No, I did respond
22:39 to him. I said I said, "Hey, uh, thanks
22:43 for connecting with me. I love your
22:45 [ __ ] I've loved your [ __ ] for years. I
22:47 think what you're doing is great." And
22:48 he's like, "Cool. Go check out our
22:50 shit." and then I didn't.
22:53 So, so I will I will uh I did respond to
22:57 him, but but maybe we'll go play with it
22:59 tonight.
23:04 Oh, the other thing if you work for a
23:06 large company or if you know people that
23:08 work for a large company, um I was
23:10 actually accepted into the Washington
23:12 Speakers Bureau. So, if you go to
23:15 Washington Speakers Bureau, I now have a
23:17 profile page up on on their um if you're
23:22 looking for speakers. So, if you want
23:24 someone to talk about AI,
23:28 I do not bring champ.
23:33 [Music]
24:07 I was watching network chuck.
24:11 setup NB. Oh, N8N. It looks complicated
24:14 but cool. Yeah, I I have not played with
24:18 N8N. It's one it's one of those there
24:21 are there are many things in AI that um
24:27 I keep meaning to go try
24:32 and then I look at it and I'm like I
24:34 don't have the energy right now. Nad is
24:37 one of those where it's like I know it's
24:39 it it's just like make and it's just
24:41 like Zapier and I've made lots of Zapier
24:43 automations and they're jankier than
24:46 [ __ ] and I I think NAN's a little
24:48 slicker, but um it's just another
24:51 [ __ ] thing to learn. It really is
24:54 just another [ __ ] thing to learn. And
24:55 then I look at something like GenSpark
24:58 and I look at something like OpenAI
25:00 agents and I sort of play them out for a
25:02 year or two and
25:06 it um
25:09 we're not going to we're not going to
25:11 need automation tools like that. I don't
25:14 think
25:15 um I think we'll just be able to ask for
25:17 [ __ ] Hang on. My my my light my $20
25:21 million production studio lighting
25:24 package needs to be plugged into USBC.
25:28 Um,
25:32 what are we doing today, Kyle? I don't
25:34 know, Mr. I think we're going to go
25:36 check out Hume. I'm just kind of
25:38 chilling right now. It's Friday night
25:40 date night. We're kind of hanging out.
25:42 We got some folks here. We got some
25:44 folks on YouTube. We're talking about
25:46 how people are inviting their friends
25:48 over to watch me, which I think is just
25:51 a bad call on on on many fronts. Many
25:56 fronts. Many fronts.
25:58 Um,
26:00 I appreciate it. I love being up on the
26:03 jumbotron.
26:05 I'm I'm a CEO who doesn't mind a
26:07 jumbotron shot.
26:12 I think soon Chain of Craft will be all
26:16 AI handled as well. You know, Shane saw
26:19 I'm glad you said that. I
26:22 when I think about the YouTube
26:24 demonetization of automated crap, right?
26:30 You can also automate
26:34 well-crafted.
26:37 You can design a chain of craft and then
26:41 automate it.
26:43 I mean, one of the things that I do a
26:46 lot is I'll go find something that I
26:48 like. I'll go to Chat GPT. I'll have
26:51 Chat GPT analyze the thing I like, break
26:54 it down into a series of prompts, and
26:56 then I say, "Now, give me 20 completely
27:00 different ideas inspired by that."
27:03 Right? different characters, different
27:05 locations, different styles, whatever it
27:07 might be.
27:09 But I'll use it as a jumping off point
27:11 and then I'll do something really cool
27:13 with it.
27:15 That process that I do, that's this sort
27:18 of it's just a way I work and it's like
27:21 the manual version of
27:25 a way to get interesting work out of AI.
27:29 I could absolutely automate that, right?
27:36 Kelly Bosch
27:39 probably has enough
27:42 knowledge of her process that she's
27:45 probably got four or five major chains
27:47 of craft that she repeats. Right?
27:50 There's one where she starts at
27:51 midjourney, then she goes to um to uh
27:56 Luma Labs.
27:58 She does a start and an end frame. She
28:00 goes to Luma Labsa, right? That's one
28:03 chain. She's got another chain where
28:04 she's just going in using midjourney
28:06 video. She's got another chain, right?
28:10 And she knows how to do her settings to
28:12 get the prompts the way she wants them.
28:16 That could be automated.
28:18 So then the question becomes,
28:23 if you take a really interesting chain
28:25 of craft and you automate it,
28:29 does that immediately invalidate the
28:32 automated version of that? I don't think
28:34 it should.
28:36 It still had to be crafted. But
28:41 someone could take that automated chain
28:43 of craft and just automate it themselves
28:44 and then go spew out 10 th 10,000 of
28:47 those a week. I don't know. We're in
28:50 weird territory. We're in weird
28:52 territory, people. Ton Todd is checking
28:55 in. Check. You're in. Congratulations.
28:57 Checking out. All right. Good day, sir.
29:06 Yeah. you didn't miss anything. So, uh,
29:08 have a good night. Have a good Friday
29:10 night if you're indeed checking out.
29:12 That's awesome. But thank you. Hello.
29:18 We need GPT agents.
29:21 Nah, I don't believe it. James Dutder,
29:23 what don't you believe, James?
29:26 What's the topic for tonight? Cam
29:28 Katkin, it's up to you. I I I basically
29:31 have punted on having an agenda tonight.
29:34 I don't have a producer. I'm just kind
29:35 of hanging out.
29:37 What happened with the Tik Tok love
29:40 clicking? Uh,
29:43 I think we got Mr. It in here. It looks
29:45 like there's some tapping going on. I
29:48 don't know if he's using his tappy tap
29:49 machine.
29:53 [Music]
30:32 Um,
30:34 let's see. Can you talk about the recent
30:36 announcements from OpenAI? Sure.
30:40 Um, I made a Champion video in
30:42 irregulars. Okay, Dr. Jay, let's go.
30:45 We'll go watch We'll go watch Dr. Jay's
30:48 uh video in irregulars
30:51 and then Oh, I never have moved my my
30:53 little thing over. Producer Brandon
30:55 would have yelled at me for that.
30:57 Actually, all you people on TikTok
30:58 should have yelled at me for that, but
30:59 that's okay. You just had the backstage
31:02 view for like what's it been an hour?
31:05 Oh, no. I started late. Half an hour.
31:08 Um, I used Manis to design edit
31:11 workflows apparently
31:14 so I could feel so overwhelmed I
31:16 couldn't start to use them.
31:22 Okay, archetypal architect, I think that
31:25 you're on to something. Actually,
31:29 this is what I was saying. I don't think
31:31 we understand.
31:34 I don't think we understand
31:39 what's in our hands right now.
31:43 And I won't speak for you. I don't think
31:45 I understand what's in our hands right
31:47 now.
31:50 When I had a chatbot that you put in a
31:52 prompt and it gave you an answer, I
31:54 could get my head around that. And when
31:56 it got really good and started,
32:00 you know, making images and doing code
32:02 better and doing all that stuff, it was
32:04 good. I got it. I want a program that is
32:07 an emulator for this kind of video game.
32:10 Bang, out it comes. Cool. Makes sense to
32:13 me. Then they come along with these
32:15 reasoning
32:17 models,
32:19 which sometimes you talk to people,
32:20 they're like, "It's not really
32:21 reasoning." Like, "Okay, calm down
32:24 there, Skippy.
32:26 But what it's doing is it's doing
32:27 multi-step
32:29 internal prompting before it answers
32:31 you. So you give it kind of a goal and
32:34 then it goes, "The user asked me for
32:36 this and I went and found this and now I
32:38 think I should do this and then I should
32:39 do that and then I should do that."
32:45 And now we've got these things that do
32:47 that multi-step reasoning combined with
32:49 tool use. And the tool use includes
32:52 things like surfing the web on our
32:55 behalf and taking actions on our behalf.
33:00 And I think we're going to I think we're
33:03 going to end up in this weird situation
33:05 where
33:06 someone like archetypal architect is
33:08 going to go, "Oh, I wonder if I could
33:10 use Manis to create NADN automations.
33:13 Hey, Manis, could you go make me some
33:16 NAND automations?" And and Manis goes,
33:18 "Sure." And it goes out and it reads all
33:20 the documentation and it asks, you know,
33:24 what do you want to do? Uh, I want to do
33:26 this, that, and the other. Okay. And it
33:28 goes off and it makes these automations.
33:31 And then he's sitting there like, uh,
33:35 uh, did I really want that? Did I Now I
33:37 got to go install these things and set
33:39 So I I think I'll just take a nap.
33:42 [Laughter]
33:48 I think most humans don't have really
33:52 complicated
33:56 like hundredstep
33:59 problems to solve that if it were solved
34:02 in 15 minutes they would know what to do
34:05 with the answer because like if it goes
34:07 off and does all that research and make
34:09 makes all these reports what are you
34:11 going to do with them? Are you going to
34:14 publish the report? Are you going to use
34:15 the software? You going to try to sell
34:17 the software?
34:20 If you create a brilliant piece of
34:21 software and you want to sell it, well,
34:24 now you need to make it bulletproof,
34:27 right? That's a whole separate
34:29 activity than making the software in the
34:31 first place.
34:33 And then if you launch it live, then
34:35 you've got to market it. And then you've
34:36 got to deal with tech support. And sure,
34:39 you could use NAND to to create
34:41 automated tech support, but then that's
34:43 another [ __ ] project to do, right?
34:46 So, I just think we're entering this
34:48 world where our capacity to make [ __ ] is
34:52 going to be so significantly beyond
34:55 our capacity to imagine what we might
34:58 want to make and then deal with the fact
35:01 that it just got made for us.
35:06 I noticed that when 40 switches to web
35:08 search or reasoning, it often loses
35:11 context. Yeah, that makes sense. I get
35:14 my jollies from having problems not
35:16 solving them. Exactly. I think a lot of
35:18 people are like that to be quite honest.
35:21 All right. Um Okay. So, let me
35:27 go to Ireulas. Let's go to irregulars,
35:31 people, shall we?
35:33 Hey, irregulars. What you got going on?
35:37 I'm an irregular. I'm an irregular, too.
35:40 You're an irregular? Yeah, I'm an
35:42 irregular. Are you really? How did you
35:45 get in? I just showed up. Me, too.
35:48 That's so funny.
35:52 I thought you had to fill out a form.
35:54 There's no form. You don't have to fill
35:57 out a form. You just show up and you're
35:59 an irregular.
36:01 Who knew?
36:03 Apparently, they frown on being late,
36:05 though. However,
36:07 very strange. Very strange. It It just
36:10 seems to be a Tik Tok channel that
36:14 doesn't always start on time
36:17 and yet everyone's concerned about being
36:19 late.
36:21 It It's very concerning. I I I don't
36:24 think it's normal behavior. It's
36:26 irregular.
36:28 I'm starting to get it. I'm starting to
36:30 get I think I just might have become an
36:32 irregular. Oh my goodness. Oh, this is
36:34 very exciting. Very exciting.
36:41 All right. Um, that little skip was skit
36:46 skip that little skit was brought to you
36:50 by uh old SNL. Um, all right. What am I
36:54 looking for here? Urgent. Spend your
36:56 Friday night with Manis. Credits inside
36:58 and invite codes live tonight. Manis
37:00 founders versus open AI agents
37:04 exclusively for the AI collective. Just
37:06 came across this.
37:08 Of course, your first stop should be the
37:10 AI learning lab with Kyle Shannon. Oh,
37:11 that's very sweet. Um,
37:15 holy [ __ ] The Okay, so Cam Ken, I'm
37:18 coming to you. I'm coming back to I will
37:20 circle back to um the OpenAI
37:24 announcement because Manis and Gen Spark
37:27 are [ __ ] trolling them hard and I'm
37:30 I'm here for it. So, I'm liking this a
37:32 lot. But yeah, there's a there's a Dr. J
37:35 Champy video inspired by Steo's post.
37:40 How do I Where is it?
37:43 Oh, here we go. Canva. Okay. Oh, nice.
37:49 Doggy date night. I like it. Okay. Wait,
37:52 did did this make a new thing? It did.
37:54 Hang on a sec. Copy.
37:57 Go back to here. Go back to here. Go
37:58 back to here. All right. Now I'm sharing
38:01 it, right? Yes. Okay.
38:04 [Music]
38:09 It's doggy date night.
38:12 Who will champ?
38:14 It's doggy date night.
38:22 Okay, this is the kind of Listen, this
38:25 is the kind of automated dril that
38:27 YouTube will not let you monetize.
38:30 You can't just make a doggy date night
38:32 video with cute dogs in it and expect
38:34 that to be considered narrative. This is
38:37 not narrative, but I'm going to watch it
38:38 again because it's cute.
38:44 >> It's doggy date night.
38:46 will choose.
38:49 It's doggy date night.
38:52 >> Wait, you can't see.
38:55 I thought you could see. Can you hear
38:57 this? I think I'm sharing correctly.
39:02 You can see. Yes,
39:07 I love doggy date night. Yeah, but you
39:10 can't monetize this.
39:13 You'd have to put another scene in it in
39:15 a in a title. We can see it and hear it.
39:17 Okay, good. All right, one more time.
39:23 It's doggy date night.
39:26 Who will champion choose?
39:29 It's doggy date night.
39:35 Okay, that is so stupid. It's brilliant.
39:42 Okay, let me go to X X.com,
39:46 formerly known as Twitter.
39:51 It's doggy date night. Uh,
39:56 it's doggy date night.
40:00 Okay, I What am I looking for? I am
40:02 looking for
40:07 Oh, who did that? Winston. Those are
40:09 beautiful. Thank you very much for that
40:11 gift. Open AAI um agent.
40:16 Okay, let's go watch a video. So Cam
40:20 Katkin. So yesterday
40:25 OpenAI
40:27 launched agent
40:30 OpenAI agent. And I'm going to find a
40:32 video here in a second. Yeah, this is
40:35 it.
40:37 Yeah,
40:39 this is a
40:41 >> before the whole
40:42 >> it's a fascinating video on a lot of
40:44 levels.
40:47 All right, so we're going to play the
40:50 marketing video. And mind you, this is a
40:54 market. My AI comedy club vid is less
40:56 entertaining.
41:01 Oh my god. Okay. So, what we're gonna do
41:04 here
41:08 Oh my god,
41:10 I'm sleepy. I'm sleepy, people. It's
41:13 Friday night date night. There's no
41:14 yawning in AI.
41:19 Good lord. All right. Um All right,
41:22 let's watch the video.
41:25 It may be a while for the whole world to
41:28 evolve to a AI agent ccentric worldview.
41:31 And so I think we should do what we can
41:33 to meet the world where it's at.
41:36 >> Okay. So, so you pop into agent mode,
41:40 right? So the little toolbar at the
41:42 bottom, this will be available to plus
41:45 users in the next three days, like
41:48 Monday or Tuesday.
41:50 That's me, too. I've had two naps today
41:52 and I'm going to bed. I can't stay
41:53 awake.
41:56 Who is that? Is that Vicki or or Brandon
41:58 that just posted the AI salon post
42:00 there?
42:07 All right. Okay. So, Monday we should
42:10 have this this new thing called agent
42:12 mode.
42:13 It's me. It's Vicki. Oh, cool. Yeah. Uh
42:16 head head off. Take take a nappy. I I
42:19 had I had um
42:22 I didn't have a nappy at work. I have I
42:25 have napped at work before though. But I
42:27 had a nap. I got home. I had to go take
42:29 photographs of Gabby's art show and then
42:32 she wanted to hang out there and so I
42:35 came home and took a nappy and then I
42:37 went and picked her up and then I came
42:38 here. So that's why I was late. Um Oh,
42:40 that's okay. I don't mind you posting as
42:42 a salon. That's cool. Um where's the
42:46 agent link? So the agent link will just
42:48 be within chat GPT under your tools menu
42:52 in the prompt bar. It'll just be a new
42:54 mode called agent mode. Okay.
42:56 >> Meet the world where it's at.
42:58 [Music]
43:03 My name is John. I work on the deep
43:05 research and agents teams at open. One
43:07 great use case that comes up a lot is
43:10 you have some kind of budget file and
43:12 whenever you do that it's kind of a
43:14 pain. It takes maybe four to eight hours
43:16 and that's kind of your day. I'm going
43:17 to show you an example where the agent
43:20 >> Okay, first of all,
43:22 you you we all have this scenario that
43:25 comes up all the time. We've got like
43:26 this budget and we've got to spend four
43:28 or eight hours working on that budget.
43:32 Really? Does that really come up for a
43:35 lot of people? A lot of times,
43:37 I mean, maybe. And then the example that
43:40 he uses, well, you'll see, is not one
43:43 like it's his budget. It's like the city
43:46 of San Francisco's budget
43:47 >> sources information on the city of San
43:50 Francisco's annual budget, expenses, and
43:52 revenues for the past five years, and
43:54 it's going to compile that all into one
43:56 nicely formatted spreadsheet. It
43:59 >> Okay,
44:01 so one of the things this says to me is
44:04 that the people that are making these
44:06 tools
44:08 don't actually have a handle on how
44:11 they're going to be used.
44:13 They use two examples when they did the
44:16 announcement. They used two examples. At
44:18 work, budgeting is annual. Not all the
44:20 time. Exactly. Exactly. I've been
44:24 posting AI salon vids in YouTube and
44:26 forgot to change accounts. Oh, that's
44:27 cool. You're cool, Vicki. Vicki, I'm I'm
44:29 I'm cool with you posting either way. Um
44:31 I thought it was kind of cool. Um,
44:37 the two examples that they used in the
44:39 demo were book me a vacation.
44:45 Okay.
44:46 I I mean, yeah, that that's a use case.
44:50 Every single time one of these companies
44:53 that does these multi-step things, you
44:56 know, says, "Here's what you can do with
44:58 it." They're like, "You could book a
45:00 vacation." you I know that use case.
45:02 Could someone come up with a [ __ ]
45:04 interesting use case? So he's like, you
45:06 do budgets all the time. And and so now
45:09 I'm going to show you the San Francisco
45:10 budget. Okay, whatever. So So anyway,
45:13 >> goes on by itself. I usually just close
45:15 my laptop, go grab a coffee, maybe I
45:17 have lunch that expenses and revenues
45:20 for the past the city of San Francisco's
45:22 annual where the agent sources eight
45:25 hours. When whenever you you have some
45:27 kind of budget file and whenever you do
45:29 that it's kind of a pain. It takes maybe
45:31 four to eight hours and that's kind of
45:33 your day. I'm going to show you an
45:34 example where the agent sources
45:37 information on the city of San
45:39 Francisco's annual budget expenses and
45:41 revenues for the past 5 years and it's
45:43 going to compile that all into one
45:46 nicely formatted spreadsheet. It goes on
45:48 by itself. I usually just close my
45:50 laptop, go grab a coffee, maybe I have
45:52 lunch. Okay, let's talk about this.
45:58 This is a marketing video.
46:00 The promise here,
46:05 the promise here is, "Oh, once you have
46:10 agent, you just say, "Hey, I need a
46:12 budget whipped up there, Skippy." And I
46:15 and I hit the button and I close my
46:17 laptop and I go off and have a latte
46:20 with the rude barista.
46:25 Okay. Is that how anyone you know works?
46:29 I guess that could be the future. We
46:31 We're just there to sort of initiate a
46:34 thing and then it does the thing and
46:37 then we're going to go relax in between.
46:39 I don't know. I don't quite get it. But
46:41 I get it, but I don't get it. But
46:43 they're basically saying it's going to
46:44 take 15 minutes for this thing to do all
46:46 this [ __ ] So you might as well go get a
46:48 coffee.
46:48 >> So first it needs to find the data. So,
46:50 it probably does a web search to figure
46:53 out where it can find this San Francisco
46:55 city budget information. Once it finds
46:57 the San Francisco
46:58 >> Okay. So, so here's where I want to
47:00 stop. So,
47:01 what you're looking at on screen here is
47:04 this this little window. Notice that
47:06 they've got the same background for this
47:08 window as the advanced voice little
47:11 circle when you talk to advanced voice
47:13 and it's got that little sky thing. It's
47:16 got similar branding here.
47:18 So, this little window is an agent
47:22 within a normal chat context, right? So,
47:24 you're having a chat thing and within
47:26 the chat thing, you're like, I got to
47:27 come up with a budget thing and I'm
47:29 going to do it this and a that and oh,
47:31 why don't I just have this agent go off
47:33 and do all this research on the budget,
47:35 right? And so
47:38 this is if you've used GenSpark or
47:40 Manis, it's essentially an identical
47:43 interface where what this window does is
47:46 shows you the steps that the agent is
47:49 taking. It's off surf surfing the web.
47:51 It's off researching websites. It's it's
47:53 writing code. It's switching to image
47:56 generation. It's switching to generate
47:58 um a spreadsheet. So this window is now
48:02 this playable
48:03 um history of all the steps that this
48:06 agent took along the way but it's in the
48:08 context of a larger chat GPT chat that I
48:11 think is actually quite significant.
48:13 GenSpark and Manis both are these kind
48:16 of standalone operations that you go
48:18 there, you kick off this agentic
48:20 behavior, it does all of its [ __ ] and
48:22 then it gives you a bunch of stuff. Chat
48:25 GPT, this is all going to be within the
48:26 context of a search. So I think I think
48:30 more people are going to experience this
48:32 than experience Manis and GenSpark
48:35 because those are these specialized
48:37 tools that you actually have to think
48:39 about how will I use this, this, and
48:40 that. I think this might be uh something
48:44 a little um a little more intuitive for
48:47 people. Um anyway, let's keep watching
48:50 this. Go city government website. It
48:54 will try to access the PDF files. So, it
48:56 has its own file system and everything.
48:59 Then it needs to extract maybe 200
49:02 numbers from each
49:03 >> Vicki. Right. If I'm working, my manager
49:06 will want me at my desk, not randomly
49:08 grabbing coffee or out for lunch. Yeah.
49:10 They they see out at the cafe. They're
49:12 like, "Hey, Vicki, what's going on?
49:13 Aren't you supposed to be working on
49:14 that that thing?" "Oh, I not to worry.
49:17 My agent's taking care of it." Okay. So,
49:21 Vicki, why do we need you? Let me get
49:24 back to my desk. Exactly. Like, I just
49:27 don't I just don't buy this utopian
49:29 little vision they have here. But
49:30 >> PDF and finally, it will have one
49:33 command that will generate the entire
49:34 spreadsheet all at once. If you go back
49:36 to the chat, you'll see the final
49:38 response. And let me just open it now.
49:41 Yeah, I think it got 98% of the
49:44 information correct. It also formatted
49:47 the Excel workbook as I instructed.
49:49 >> Wait, wait, what did you say?
49:54 What did you say? Goodlook marketing
49:56 dude
49:57 >> command that will
49:58 >> I don't know if he's good looking.
49:59 >> Generate the entire spreadsheet all at
50:01 once.
50:01 >> If you go back to the chat, you'll see
50:03 the final response. And let me just open
50:06 it now. Yeah, I think it got 98% of
50:10 >> Let me just open it now. Yeah, I think
50:14 it got like 98% right.
50:18 Wait, what?
50:21 Which 2% didn't it get right and you
50:24 just opened your laptop and immediately
50:26 noticed it got 2% wrong?
50:31 How How did you know that? Well,
50:37 I would assume that when you get back
50:40 after your double Zippo latte with the
50:42 friend that you don't really like at
50:44 work, your colleague that you pretend to
50:45 like but don't really because they're
50:48 passive aggressive and they don't have a
50:50 point of view that's their own.
50:53 I assume when you get back to to the
50:56 Zippy Latte with that person, you sit
50:59 down at your desk, you open it up and
51:01 you look at all of the [ __ ] that it did
51:04 and you're like, "Oh my god, that's a
51:06 lot. Do I have to look at all this and
51:09 then you look at the spreadsheet that it
51:11 made?"
51:13 How do you know those numbers are right?
51:16 How do you know where the source is?
51:20 I assume
51:23 you as a human need to dig through some
51:25 [ __ ]
51:28 right? So, the the four or eight hours
51:30 it would have taken you to make this on
51:32 your own.
51:34 Now, you've probably got one to two
51:36 hours. If if you actually want to see if
51:38 it's 100% right, you actually need to go
51:41 look at every number in this spreadsheet
51:43 and figure out how it got to that.
51:47 Now, is it possible that it wrote up a
51:49 whole description of how it did it and
51:51 maybe it's got areas of where it's not
51:53 confident about those number? Maybe. I
51:55 don't know. But this is where I'm what
51:58 I'm saying where I don't think that we
52:00 are
52:03 prepared
52:04 for
52:06 what our role is with these things. If
52:10 like if I'm doing something like
52:12 creative ideiation with one of these
52:14 tools, then it doesn't [ __ ] matter.
52:16 Like it's either good ideas or not. But
52:18 if it's something like a spreadsheet for
52:20 a [ __ ] budget,
52:22 oh, it got it 98% correct. Great.
52:26 Is everyone going to check to make sure
52:29 that it got 100% correct? I don't know.
52:33 So, how many of these reports are going
52:35 to be put into annual reports
52:38 that someone's going to actually do the
52:40 math on and go, um, where did you get
52:42 this number? Ah,
52:46 chat GBT
52:53 information correct. It also formatted
52:56 the Excel workbook as I instructed it
52:58 to. In this case, the revisions were
53:00 small, so I just made them within Excel
53:02 cuz it was just a copy paste. But
53:04 absolutely, you can make them in chat
53:06 GPT. I would say just try it out. If it
53:09 can do 90 95% of the actual time
53:13 consuming part of the work, uh, that's
53:15 going to save you a ton of time.
53:17 >> Yeah, that I agree with. Right. If it's
53:19 going to do 90 or 95% of the work, it's
53:21 going to save you a ton of time. Current
53:23 chat GPT maybe gets you to 80%. Maybe
53:26 these new agents get you to 90 95%. But
53:29 you still have to deal with the output,
53:32 right? You still have to use your
53:33 critical thinking.
53:35 Is there going to be a point in the
53:37 future where you don't need to use your
53:38 critical thinking? Probably.
53:41 And that's the point at which they
53:42 probably don't need you for that job.
53:47 So, make no mistake, as cool as these
53:49 things are, and as much as we want them
53:52 to be rock solid,
53:56 the minute you have a tool that can do
53:58 your job as well as you do it,
54:02 why do we have you here?
54:06 And that's something that we're all
54:07 going to have to figure out because I
54:10 think there's an answer to it. The
54:12 answer is I'm going to use these tools
54:14 to take, you know, what I do for my job
54:18 to the next level.
54:20 I think who ends up staying in jobs are
54:23 the people that go, "Oh, okay. If all
54:25 that shit's automated, rather than
54:27 heading to the cafe with with Betsy from
54:30 accounting,
54:33 I'm going to reinvent how we do
54:35 something and deliver more value to the
54:37 client or save the company money and I'm
54:39 going to go tell my boss about it. I I
54:42 think I think that's how you use these
54:44 tools to to, you know, amplify
54:49 yourself, your work, your livelihood.
54:52 Um, now really interesting thing.
54:57 So, this thing's out.
55:00 It uh let me let me talk a little bit
55:02 more about it. Um, so for Cam Catkin and
55:04 for anyone else who's interested in the
55:06 way these things worked work um
55:13 chatbt will never be able to retire as
55:16 well as I do. Exactly.
55:22 Um,
55:24 so what OpenAI did with with Agent is
55:28 they realized that they had two
55:31 different tools that were both very
55:33 powerful but limited. They had deep
55:36 research, right? If you've used deep
55:39 research, you turn on the deep research
55:41 button and you say, "Hey, go find all
55:43 the marketing claims that my competitors
55:46 make." and it goes off and it researches
55:48 websites and does stuff and it writes
55:50 you up a report. And then you have
55:52 operator and what operator does is it
55:54 can surf the web visually and it can go
55:56 look at web pages and click on interface
55:59 elements and fill out forms and analyze
56:02 pictures and whatever it does.
56:05 Deep research can't see the pictures,
56:08 can't navigate
56:10 um complicated interfaces.
56:13 operator is slow and can't do quick
56:17 analysis of lots and lots of websites,
56:20 right? So, they both have strengths and
56:22 weaknesses and they realize they're kind
56:23 of complimentary. So, what agent is is
56:27 it's basically
56:29 a virtual machine wrapper
56:32 that its primary job
56:35 is to navigate which tools to use. So,
56:38 it might go, oo, I'm going to go use
56:39 deep research to to find a bunch of
56:41 [ __ ] Then I'm going to go to this site
56:44 at the Gap and find the latest in men's
56:46 fashion and describe them. Then I'm
56:49 going to write a Python tool to bring it
56:52 all together into a little interface.
56:54 Then I'm going to get this spreadsheet
56:56 that I got from my deep research and and
56:59 turn that into an Excel spreadsheet.
57:00 Then I'm going to put that all together
57:01 into a presentation with charts and
57:03 graphs. So it's got access to lots and
57:06 lots of different tools. And its job,
57:08 what it's been trained on is how to use
57:11 which tool when.
57:14 Right now,
57:17 it's which is cool. Like, it's okay.
57:19 [ __ ] awesome.
57:21 They're now not doing the reinforcement
57:24 learning just on the core model. They're
57:27 now doing the reinforcement learning on
57:30 the navigator,
57:33 right? the the thing that is going to
57:35 figure out which of all of the tools I
57:37 have access to am I going to use
57:40 um agent also has access to APIs. So
57:43 you're going to be able to do things
57:44 like connect it to your Google Drive,
57:46 connect it to your Gmail, connect it to
57:48 anywhere where you've got a connector,
57:50 which is a [ __ ] privacy and security
57:53 nightmare that we won't talk about right
57:55 now. But it's but it's it's this very
57:57 powerful coordination layer figuring out
58:01 all of the different tools it needs to
58:02 use to accomplish your goal. That's what
58:05 it is. Cool.
58:08 If you've used GenSpark or if you've
58:10 used Manis, you've used Agent. They're
58:14 the same thing. And I would argue that
58:17 Manis and Gen Spark are way ahead.
58:22 And
58:24 what we've seen in the past two days
58:25 since this announcement is that they
58:28 know that Manis
58:33 Manis AI
58:37 let's see
58:39 um
58:42 here they're bragging about millions of
58:45 real world sessions.
58:48 Here are the lessons we learned about
58:50 context engineering for AI agents.
58:53 Um,
59:01 let's see. Maybe it's in replies.
59:11 I don't see it here. Um
59:17 so so when when OpenAI launched Agent
59:21 um Manis did a tweet that said welcome
59:24 to the club and then GenSpark also did
59:27 one and then GenSpark today announced a
59:30 million um a million dollar contest.
59:34 Let's see. Genspark.
59:43 [Laughter]
59:46 Milliondoll
59:48 Gen Spark side by side comparison show
59:51 AI showdown.
59:53 $100,000 on the line. Same prompt, one
59:58 shot, side byside results. We think
1:00:01 we're pretty good, but we're hunting for
1:00:03 more use cases where chat GPT might have
1:00:06 an edge. Help us make Gen Spark better.
1:00:11 So, they are [ __ ] absolutely going
1:00:13 after um
1:00:17 going after OpenAI's agents. Now, here's
1:00:20 the deal. They should be scared. They
1:00:23 should be scared. GenSpark and Manis.
1:00:29 Most people haven't heard of them. If
1:00:31 you're in AI, if you watch this channel,
1:00:33 you've absolutely heard of them. You've
1:00:34 probably used them. Um,
1:00:38 hang on a sec. Get black bar going there
1:00:41 for the people on the Tik Tok. Uh,
1:00:46 Kyle, soon you're reaching 15,000 likes
1:00:48 in Tik Tok again. Very nice. 15,000
1:00:51 likes. Beautiful.
1:00:53 Already done. 25,000 likes.
1:00:56 Um, okay.
1:01:06 It's hard to overstate the advantage
1:01:09 that
1:01:11 chat GPT has.
1:01:15 um
1:01:18 big corporations, big companies,
1:01:21 um even ones that are Microsoft shops
1:01:24 and have access to Copilot
1:01:28 often also have ChatGpt subscriptions,
1:01:32 enterprisewide
1:01:34 OpenAI chat GPT subscriptions
1:01:37 because nobody likes C-pilot.
1:01:41 People don't like it. Like Microsoft is
1:01:43 having trouble selling what they sell
1:01:45 because the companies that they're going
1:01:47 into are like, "Oh yeah, our people use
1:01:49 chat GPT. We get we got them licenses.
1:01:51 That's what they use. They don't they
1:01:52 don't they can't figure out how to use
1:01:54 co-pilot."
1:01:56 So the fact that this agent experience,
1:02:00 the manis and genspark-l like experience
1:02:02 is going to live inside a chat GPT chat,
1:02:06 there's a competitive advantage there
1:02:08 that's going to be really really hard
1:02:11 for the Gen Sparks and the Manises of
1:02:13 the world to overcome. Not to mention
1:02:18 the security and privacy issues. Manis
1:02:21 is a Chinese company. Gen Spark I don't
1:02:23 think is, but I'm not 100% sure on that.
1:02:25 But Manis is a Chinese company.
1:02:29 Both of these things, as well as the
1:02:31 Perplexity Comet browser that's, you
1:02:34 know, can can access all of your tabs
1:02:37 and all of the things you've logged
1:02:39 into.
1:02:41 All of them are security nightmares,
1:02:44 right? because you're like, I want you
1:02:46 to go look at, you know, all of my
1:02:48 tweets on my Twitter account and and
1:02:52 post responses to them. It'll go do
1:02:54 that, but you've got to give it access
1:02:56 to your Twitter account with full write
1:02:59 access. And then if you want to like
1:03:00 book the travel thing like they they
1:03:02 demo, you've got to give it access to
1:03:04 your credit cards, to your travel
1:03:06 accounts, to your right, all that stuff.
1:03:09 And it's going to go buy that [ __ ]
1:03:10 that's going to ask you permission.
1:03:16 So,
1:03:17 so I think it's I think it is smart for
1:03:20 them both to um sort of kick OpenAI in
1:03:24 the teeth while they're just getting
1:03:26 started. Um I don't think that advantage
1:03:28 is going to last all that long. Chat GPT
1:03:31 is incredibly wellunded. They're
1:03:33 incredibly smart and they're going to
1:03:34 figure this shout [ __ ] out incredibly
1:03:36 fast. But right now,
1:03:40 if you want to use OpenAI's agent, you
1:03:44 don't need to. Just go to GenSpark. If
1:03:46 you've not played with GenSpark, go to
1:03:49 GenSpark right now and just play with
1:03:50 it. So, um, so that's that's my take on
1:03:53 this stuff. I think agent is a is a a
1:03:58 strong initial
1:04:02 um attempt
1:04:04 at doing what GenSpark is already doing
1:04:08 really really well and Manis is doing
1:04:09 really well. I think the chat GBT
1:04:13 version is going to really suck
1:04:15 initially. It's going to be slow. It's
1:04:18 going to be janky. I think it's
1:04:20 essentially what Gen Spark says. In
1:04:23 fact, they did. Let me see.
1:04:27 Um, Jen Spark
1:04:30 prompt
1:04:32 chat GBT agent.
1:04:36 Let me see if I can find this.
1:04:55 One of the
1:04:58 one of the Gen Spark engineers.
1:05:03 One of the Gen Spark engineers. I can't
1:05:05 find it right now. Is this it?
1:05:12 Does this guy work at GenSpark?
1:05:17 No, I think he I think he just copied
1:05:18 what the Gen Spark guy did. But
1:05:20 basically, um, they took the exact
1:05:24 prompt that that the open Oh, wait. I'm
1:05:26 not sharing my screen. Damn it. That's
1:05:29 why I need a producer.
1:05:33 [Laughter]
1:05:36 I can't believe I've been I've been
1:05:38 talking about all this [ __ ] without
1:05:39 share sharing my screen. That's okay.
1:05:41 Um,
1:05:43 so GenSpark um
1:05:46 took the exact prompts that they did in
1:05:48 the in the OpenAI demo and they're like,
1:05:51 "Oh, look, our thing did it way faster
1:05:54 and way better, right?" So, so I think
1:05:58 that um I think that the the OpenAI
1:06:01 one's going to suck for a while and and
1:06:03 that's okay. But again,
1:06:05 this is not about where we are in the
1:06:08 present. I look at these tools and I and
1:06:11 I'm like this is what
1:06:15 some version of what these do is where
1:06:19 everything is headed.
1:06:23 So, so our whole idea of you need to
1:06:25 learn prompt engineering and you need to
1:06:27 learn this and you need to do that and
1:06:29 you need to understand all these
1:06:30 different tools.
1:06:32 We're not going to need to understand
1:06:33 all these different tools. If you want
1:06:34 to build
1:06:36 applications that leverage all this
1:06:38 stuff, then yeah, you're going to need
1:06:40 to know some of that [ __ ] But even
1:06:41 then, you're going to get assists.
1:06:45 But just as a general enduser,
1:06:48 we're just going to ask for [ __ ] and
1:06:50 these tools are going to go off and make
1:06:52 it. So, if if one of my strong
1:06:55 recommendations and one of my if if if
1:06:58 I've got weekend homework for you,
1:07:01 even if you've tried GenSpark and Manis
1:07:03 in the past, I would go back and really
1:07:06 start to explore what are all the
1:07:09 different ways I personally might use
1:07:12 something like Manis or Genpark
1:07:15 to do [ __ ] for me. Like, think about
1:07:18 things you might use it for.
1:07:20 start to understand how you would use
1:07:22 that differently than how you use chat
1:07:24 GPT.
1:07:26 All right. I think it's a big big big
1:07:29 big deal. I you know like this is the
1:07:32 the new phase that I think we're
1:07:34 entering is is going to be centered
1:07:36 around these things. Genpark uses GPT
1:07:40 models via API. So OpenAI gets paid when
1:07:42 GenSpark does it. That's funny.
1:07:46 We're at 20,000 likes already. Good.
1:07:49 Fantastic. Only a matter of time before
1:07:51 model makers eat any company.
1:07:54 Uh but needs the model to run. Yeah.
1:07:56 Listen, I think I think Genpark and
1:07:59 Manis are going to be my instinct is
1:08:02 they're going to be successful
1:08:04 companies. Um I don't see them,
1:08:09 you know, I see them being becoming more
1:08:11 niche
1:08:13 successes than really broad successes,
1:08:16 but we'll see. Or they get acquired,
1:08:18 right?
1:08:19 like Gen Spark's doing what they do so
1:08:21 good. It wouldn't surprise me if OpenAI
1:08:23 goes, "Yeah, here's a billion dollars.
1:08:25 Bring all that IP over here. We'll take
1:08:26 it."
1:08:29 Um
1:08:32 Sam told us there'd be days like this,
1:08:37 we have you here to check the missing
1:08:39 2%. Yeah, exactly.
1:08:51 Um, all right. Questions that make that
1:08:55 all make good sense.
1:08:57 I know it was a lot. I rambled for a
1:08:59 long time and I thought I was sharing my
1:09:01 screen when I wasn't.
1:09:04 I didn't have producer Brandon yelling
1:09:06 at me. Is it possible to make sub
1:09:08 projects in each projects folder in chat
1:09:11 GPT?
1:09:14 Yeah. Well, you you mean a separate
1:09:17 a separate agent? So So the way it's
1:09:20 going to work is this. For 20 bucks a
1:09:22 month, you get 20 agents per month.
1:09:29 So
1:09:30 So basically like an agent a day, right?
1:09:33 Think of it like that. Or you know on
1:09:35 the weekend you do, you know, seven or
1:09:36 eight of them.
1:09:38 How I can imagine it happening is
1:09:42 you're working in chat GBT on something
1:09:44 and then you say, "Oh, I've got this
1:09:46 more complicated thing. I've got I I
1:09:48 need an output
1:09:50 as part of this project I'm working on."
1:09:53 And that would be a perfect thing for me
1:09:54 to to generate an agent to go do that
1:09:57 task, right? and you're like, "Okay, go
1:10:00 find, I don't know, the latest public
1:10:03 art installations in Denver, how much
1:10:06 the commissions were, who who issued the
1:10:09 commission, and how long it took from um
1:10:12 application until granting of the thing,
1:10:15 and then turn that into a report for me.
1:10:17 So, I'm putting together a thing for
1:10:19 some art installation I want to do in
1:10:21 the city of Denver. And one component I
1:10:24 want is a little report to know what's
1:10:27 the lay of the land. So I could see
1:10:29 working within chat GPT having that
1:10:32 realization and go let me spin up an
1:10:35 agent. So you spin up the agent that
1:10:37 little window is now in the context of
1:10:39 your chat
1:10:41 and you keep chatting. That agent is off
1:10:44 doing its thing. It may take 15 minutes,
1:10:46 right?
1:10:48 And and you're still chatting doing
1:10:50 other things. So I think within a single
1:10:52 chat or within a group of chats like in
1:10:54 a project, you could have multiple
1:10:56 agents off doing things. So yes is the
1:10:59 answer. Again, I haven't used it, but
1:11:01 that's just based on what I'm seeing and
1:11:04 knowing how they've done other things. I
1:11:06 think that's how it's going to work.
1:11:12 Workflow hack from Josh Groves on on X.
1:11:17 Drop any deep research report into Gem
1:11:21 Spark AI slides. Yeah. Beautiful,
1:11:24 perfect, coherent slides in less than 10
1:11:26 minutes. Yeah. Um,
1:11:29 GenSpark slides are really good. They're
1:11:33 really good. They're welld designed.
1:11:34 They're well put together. I think that
1:11:36 hack is a good one where you have deep
1:11:39 research, go do the deep research to
1:11:41 make sure you got the research part of
1:11:42 it right. Because what I found is if you
1:11:45 just ask GenSpark to go make you a slide
1:11:48 deck on whatever topic, it'll just make
1:11:51 [ __ ] up or it'll like it's I think it's
1:11:54 better to have some sort of coherent
1:11:57 vision for what you want the deck to be
1:11:59 and then yeah, let it go do that. That's
1:12:01 a that's a really good hack. Is that
1:12:04 agent function what's coming on Monday?
1:12:06 Yes.
1:12:08 Yeah. And in fact, let me go just check
1:12:10 right now and just see on the off chance
1:12:12 I have it. I doubt I do.
1:12:19 New chat.
1:12:21 No, I don't have it yet.
1:12:27 I wonder if I have it over on Teams.
1:12:31 Uh, search connected apps tools, deep
1:12:36 research connectors. Nope,
1:12:40 I don't have it. And I was gonna I was
1:12:42 gonna like reup for a pro GPT account
1:12:45 for a month, but I'm like, ah, [ __ ] it.
1:12:48 I don't care. Like, it's
1:12:52 I've got access to Manis. I've got
1:12:54 access to GenSpark, and I don't really
1:12:57 use them. Again, I'm I'm still in this
1:13:00 place where
1:13:03 I don't really know what to use these
1:13:05 things for.
1:13:11 I'll figure it out. We'll figure it out.
1:13:16 I think it's important to understand
1:13:18 what they can do. I think when we figure
1:13:22 out what these things can do and get
1:13:25 good at them,
1:13:28 I think we are going to occur to other
1:13:31 people who are not using them
1:13:33 like wizards,
1:13:35 which I think is cool. I think we should
1:13:37 all be like wizards.
1:13:40 I don't have it yet either, Kyle. I need
1:13:42 a cartoon avatar persona. All right. So,
1:13:46 Mr. The way you do that is go to um any
1:13:51 of the image generating tools, chatbt
1:13:55 midjourney ideoggram,
1:13:58 generate a cartoon character,
1:14:01 and then um
1:14:05 put your script together in chat GPT.
1:14:09 Use 11 Labs to pick a voice to to to
1:14:12 make an audio track of that thing
1:14:14 talking. And then go to Hedra
1:14:18 Hed
1:14:20 and upload the audio track of your
1:14:23 script and your character.
1:14:26 And
1:14:28 you have it.
1:14:30 I've done it in here many times. Video
1:14:33 avatar to promote my product. Yep. Make
1:14:36 an image,
1:14:38 write a script, pick a voice in 11 Labs,
1:14:41 have it read that script,
1:14:43 upload the the audio file and the image
1:14:48 file into Hedra and say, "Make me a
1:14:52 video." Done.
1:14:54 Done.
1:15:02 And just like that, Kyle caused everyone
1:15:04 to yawn. What did I do? Oh, because I
1:15:06 yawned.
1:15:14 But seriously, Cam Catkin, go play with
1:15:16 Gen Spark. If you want to if you want to
1:15:18 see what's coming with agent, go play
1:15:21 with Gen Spark. I think with Gen Spark,
1:15:24 you get like one agent a day for free.
1:15:33 Let's go do how we doing on time. We'll
1:15:36 do we'll do one one quick little thing
1:15:38 here.
1:15:40 Let's do the the Josh Grove um Groves
1:15:44 little hack.
1:15:46 So, I'm going to go to chatgpt. Oops, I
1:15:49 got that wrong. Let me change this.
1:15:57 All right. So, I'm going to go grab deep
1:15:58 research.
1:16:02 I'm going to say um
1:16:07 find me
1:16:11 the
1:16:18 five
1:16:20 most unexpected
1:16:24 unexpectedly
1:16:27 awesome
1:16:35 sites
1:16:36 to see in Denver
1:16:39 and five
1:16:43 remarkable
1:16:47 um
1:16:50 things to do.
1:16:54 All right, that's kind of shitty, but it
1:16:57 should happen pretty quick. And so now,
1:17:00 because we're doing deep research, this
1:17:02 thing's going to kick off
1:17:04 to tailor this. Do you? Yeah. Quirky and
1:17:07 offbeat. Quirky and offbeat. Yes.
1:17:12 And no. Meow Wolf.
1:17:17 Been there. Okay. Are you interested in
1:17:20 art? Um, art history. Um, you pick
1:17:26 topic.
1:17:32 Will you be traveling solo? We'll say
1:17:34 this is with
1:17:38 family.
1:17:44 All right, there we go. Okay, so now
1:17:47 we've given a little context.
1:17:54 How's the agent with Excel? Can it
1:17:56 perform complex tasks? Yes. Yes. So,
1:18:01 so agent. All right. So, this is off
1:18:03 starting research. And now, if you
1:18:05 haven't seen deep research, see where it
1:18:07 says starting research here? I should be
1:18:10 able to see what it's doing once once it
1:18:13 gets started here. I should be able to
1:18:15 open a window and we can watch what it's
1:18:16 doing. So, what we're going to do here,
1:18:19 let me give you a little transparency
1:18:21 into where my crazy straw of a brain is.
1:18:23 We're going to have this go research
1:18:25 cool things to do in Denver. I'm going
1:18:26 to then take that output. We're going to
1:18:29 go over to GenSpark and say, turn this
1:18:31 into a slide presentation. All right.
1:18:34 Um, okay. Looking at sources. How do I
1:18:39 You used to be able to go look at this.
1:18:41 Why can I not look at it?
1:18:52 Oh, there we go. You had to click on
1:18:54 looking for sources. So, all right. I'm
1:18:56 piecing together a list of things.
1:18:59 I looked at at atlasobscura.com.
1:19:04 Inspecting unique places, 80 nostalgia
1:19:07 store, mortuary t um turned restaurant.
1:19:12 That's That's old. I've been there.
1:19:14 Okay.
1:19:16 Um, do you come on live every day at
1:19:19 this time? Every weekday? Yeah, weekdays
1:19:21 at 8:00 p.m. Mountain time. And
1:19:23 sometimes I'm a little late because
1:19:25 life, you know, life happens. Um, but
1:19:28 yeah, I'll be back here Monday. Monday
1:19:30 at 8 Mountain.
1:19:33 Be careful. 100,000 likes are coming.
1:19:36 Um, but Sam, what you should also do is
1:19:39 if you have not, you should go to
1:19:44 um that URL, the salon.ai,
1:19:47 and click on join our community. So,
1:19:49 this is a community of AI
1:19:51 creators and entrepreneurs and
1:19:53 tinkerers, people that are trying to
1:19:55 figure this [ __ ] out. A lot of people
1:19:57 that are at at the AI salon come here as
1:19:59 well. Um we call them the irregulars
1:20:01 that kind of hang out here on a regular
1:20:03 basis. So, so go check out the salon
1:20:07 um and uh and come hang out here with
1:20:10 us.
1:20:12 Okay. So, our deep research is doing its
1:20:15 thing,
1:20:23 which is cool. All right. Questions,
1:20:27 thoughts, thoughts, thoughts, thoughts,
1:20:29 thoughts, thoughts. What are you making?
1:20:31 So, what I'm making right now
1:20:34 is
1:20:36 a bit of the theme tonight. So, one of
1:20:38 the nicknames for the channel is Chad
1:20:39 Add. So, the chances of me ever
1:20:42 completing a sentence in an evening are
1:20:45 kind of low. Um, the chances of me
1:20:47 staying on topic are highly unlikely.
1:20:51 Um, but the chance that I'll stumble
1:20:53 into something interesting, pretty good.
1:20:56 Um, the
1:20:59 what we've been talking about tonight is
1:21:01 OpenAI's agent. And if you want to get a
1:21:04 essentially a fancier precursor
1:21:08 or or a fancier example of what agent is
1:21:10 and is going to be, go use GenSpark or
1:21:14 Manis. Manis.im or Genspark I think.ai.
1:21:19 Um so what we're doing right now is um
1:21:22 Josh Groves on X said, "Oh, here's an
1:21:25 idea. take it take a deep research
1:21:28 output and put it into Manis that
1:21:30 generates slides and have it generate
1:21:32 you a slide deck. So what I'm doing is
1:21:34 I'm doing a quick deep research project
1:21:36 on Chad GPT. I'm going to take the
1:21:38 output of that and then upload that into
1:21:41 GenSpark and turn it into a deck and
1:21:43 we'll see it. I've done it but literally
1:21:46 don't even prompt GenSpark. Just
1:21:48 dropping the research and it'll do its
1:21:50 thing. Oh, that's cool. Awesome. Okay,
1:21:52 I'll try that. We'll see how it does.
1:21:55 Sea slug of doom is late. Unbelievable.
1:21:58 It's Friday night date night. You You
1:22:00 can't be late. What? Tardiness is not an
1:22:05 option. This is AI we're talking about.
1:22:08 Things move faster than that.
1:22:11 Good lord.
1:22:13 What next? You tell me you went out and
1:22:15 had a lovely dinner.
1:22:18 What about an overcooked hot pocket in a
1:22:20 dirty microwave? That's the tradition
1:22:22 here.
1:22:23 Oh my god. I feel like no one
1:22:26 understands the rules. No one
1:22:28 understands the rules. They just they
1:22:29 just show up. They show up whenever they
1:22:31 want. Willy-nilly Mr. K all day. Deep
1:22:34 research chat GBT3 wrote me 60-page
1:22:37 proposal document to make myself That's
1:22:40 cool. To make yourself what? Um
1:22:45 a job I want. Oh, that's very cool. That
1:22:48 is so cool. essentially to be a
1:22:50 traveling tech coach for districts that
1:22:53 don't have one. You would be so good at
1:22:55 that, dude.
1:22:57 Hey, Mr. K.
1:23:00 I I have I since you're here, I I can I
1:23:03 beg you for something? What are you
1:23:05 doing next Wednesday? We were supposed
1:23:06 to record You and Ann Murphy and I were
1:23:09 supposed to record something last Sunday
1:23:11 or the Sunday before and we missed it.
1:23:13 It didn't happen. What are you doing
1:23:15 Wednesday at 4? You want to do something
1:23:17 live with me anyway? Ping me
1:23:22 if you can do it. 4 PM my time. Um, next
1:23:25 Wednesday. If you can do it, it would be
1:23:28 great. You'll save my butt. It's It's
1:23:31 her birthday anniversary. It's her
1:23:33 anniversary birth. I would love to.
1:23:34 Okay, cool. So, I will I will uh I'll
1:23:36 ping you on LinkedIn or on the on the AI
1:23:39 salon, Mr. K. That'd be great. Cool.
1:23:41 Perfect.
1:23:43 See slug of doo. By the way, Kyle, when
1:23:45 can we expect to see you on Jumbotron
1:23:48 with your HR manager and the salon logo
1:23:50 clearly visible? Listen, I said this at
1:23:53 the beginning of the show. I am I I am a
1:23:55 CEO is proud to be on a jumbotron. Put
1:23:57 me on a jumbotron anytime.
1:24:02 Oh my god. That one that one is just
1:24:05 like what what was what was the best
1:24:07 thing about that was you know you got
1:24:11 the contrite letter. I've made a
1:24:13 mistake. There was a lapse in judgment.
1:24:15 D, but he was such a [ __ ] CEO. He had
1:24:18 such the arrogance of a CEO. He goes,
1:24:21 but I am very disappointed that, you
1:24:24 know, this was done without my
1:24:26 permission. Like, he had to be a douche.
1:24:27 He just had to be a douche. Like, on top
1:24:31 of being a douche. He had to be a
1:24:33 douche.
1:24:38 Okay. Uh, quirky and offbeat family
1:24:41 stuff. Here we go. Um, can I download
1:24:44 this?
1:24:48 Let's see. Sources, share,
1:24:53 download
1:24:54 PDF. Okay, beautiful. Look, OpenAI
1:24:58 actually has like decent decent
1:25:01 exporting functions now. Who knew?
1:25:06 Okay.
1:25:11 The International Church of Cannabis.
1:25:14 Immersive Art Venue. Little Man Ice
1:25:16 Cream. Been there. 5280s Retro Toy and
1:25:20 Pop Culture Shop. Buckhorn Exchange.
1:25:23 Remarkable family-friendly activities.
1:25:26 Candy Factory Tour. This is cool.
1:25:30 Here's a bunch of [ __ ] All right. So,
1:25:33 we got a PDF. So now what we're going to
1:25:37 do is we're going to go to Gen Spark,
1:25:42 Gen Spark.
1:25:47 And so
1:25:50 when you go to Gen Spark, you can all
1:25:52 see this, right?
1:25:54 Hang on. Let me get this thing. That's
1:25:56 not part of the interface. Let me make
1:25:58 this bigger. Plus+. Wait, I'm not there.
1:26:05 Um,
1:26:10 sign in.
1:26:19 [Music]
1:26:27 Now look at all these things that Gen
1:26:28 Spark has. AI slides, AI sheets, AI
1:26:32 docs, AI pods as in podcasts, AI chat,
1:26:36 AI image, AI video, deep research, call
1:26:40 for me.
1:26:42 You can have you can have Gen Spark
1:26:46 make phone calls to all your favorite
1:26:49 restaurants and ask them if they've got
1:26:51 Prince Albert in a can or whatever
1:26:55 and it will do it. Uh it's got a
1:26:57 download thing. It's it's got all
1:26:59 agents. So what we're going to do is
1:27:01 we're going to do AI slides. We're going
1:27:02 to click on AI slides.
1:27:07 And then
1:27:10 let's see. I click on the little
1:27:14 browse local files.
1:27:18 There's our quirky and offbeat Denver
1:27:20 things. Open.
1:27:22 And I'm say like uh
1:27:25 giving a presentation
1:27:30 in five minutes.
1:27:33 Make it pretty and make it snappy
1:27:39 and
1:27:42 It should look like
1:27:46 Denver.
1:27:48 No. No. Let's just say make it pretty
1:27:50 and make it snappy. Let's see. Let's see
1:27:52 if it can figure out what the [ __ ] it
1:27:54 should look like. To Josh Groves's
1:27:56 point.
1:28:00 Chef Kelly. Chef Kelly loving the hot
1:28:02 pockets.
1:28:05 Chef Kelly. Now, actually, this is
1:28:08 really interesting. As a chef, you could
1:28:11 probably take a hot pocket, preheat the
1:28:14 oven, put it in an air fryer or the
1:28:16 oven, but you just got to admit that
1:28:18 it's just not the same experience as
1:28:21 microwaving it in a dirty microwave in
1:28:23 the little sleeve, right? The crisping
1:28:25 sleeve. You know what I'm saying?
1:28:29 I think even as a chef, you've got to
1:28:30 acknowledge that there's there's ways to
1:28:33 do this.
1:28:34 [Laughter]
1:28:37 Okay. So, um,
1:28:43 so GenSpark. Now, what did it do here?
1:28:46 Time to build this. So, they created a
1:28:48 cover slide
1:28:50 using a Denver skyline. Cool.
1:28:54 Nothing like the toxic sleeve. Exactly.
1:28:57 It just gives it a different flavor.
1:28:59 Right. If you've got it in the chemical
1:29:01 sleeve in the dirty microwave,
1:29:04 right, that's got the leeching
1:29:07 radio fields.
1:29:09 There's just something about that. You
1:29:11 can't replicate that in the oven.
1:29:14 If you were to cut it up in chunks and,
1:29:16 you know, sauté it, it's just not going
1:29:18 to be the same. It's not a hot pocket.
1:29:22 It's some sort of pastry delight. Sure,
1:29:26 it's not a hot pocket.
1:29:30 All right.
1:29:33 Some sort of pastry delight.
1:29:36 Chef Kelly, I'm gonna need you to whip
1:29:38 up a pastry delight for me.
1:29:41 Which, by the way, Chef Kelly, since
1:29:43 you're here, I should probably show the
1:29:45 good people. She put together a um an
1:29:49 avatar of herself, like a 3D character
1:29:52 avatar talking about food additives,
1:29:56 dyes in foods, and it's really good.
1:29:59 Kudos on putting that together. I
1:30:00 thought it was really well done. Jen
1:30:02 Spark's decks are mediocre, but Gamma is
1:30:05 good. Yeah, well, Gamma is more specific
1:30:07 to design. Um,
1:30:12 yeah, Gamma Gamma is quite a good
1:30:14 product as well.
1:30:17 Um,
1:30:20 do we have previews yet? Oh, we do.
1:30:24 Look, there's Denver in the background.
1:30:26 There's our first slide. Quirky and
1:30:27 offbeat family adventures in Denver.
1:30:29 Beyond the typical spots. Why go off the
1:30:32 beaten path?
1:30:34 Very cool.
1:30:38 Nice.
1:30:43 There you go.
1:30:45 Love it. This is looking great.
1:30:49 You can watch it write the code.
1:30:59 We got placeholders for all that [ __ ]
1:31:01 So anyway,
1:31:19 look over here. It's absolutely
1:31:21 brilliant. This this the section divider
1:31:24 is fire. The Colorado family adventure
1:31:27 backdrop. So, it's going and searching
1:31:29 for actual photos of these locations,
1:31:36 selecting the photos to put in there.
1:31:39 That's pretty [ __ ] slick. Like, this
1:31:41 is [ __ ] you would do, right?
1:31:45 How specific was your prompt, Kyle?
1:31:46 Sorry I'm late to the party. It was not.
1:31:49 It was
1:31:51 Well, okay. So, what I did, Corey, was I
1:31:53 went to Deep Research and I said, "I
1:31:55 want five quirky sites and five offbeat
1:32:00 locations or some whatever I don't know,
1:32:02 activities in Denver." And so, Deep
1:32:04 Research, I did that in Chat GPT. Deep
1:32:07 Research went off and just wrote me a
1:32:09 report
1:32:11 of here's some interesting places.
1:32:13 I took that PDF,
1:32:17 uploaded it to GenSpark, and I said,
1:32:19 "Giving a presentation in five minutes.
1:32:22 Make it pretty and make it snappy.
1:32:24 That's it."
1:32:28 And then uploaded the the the PDF.
1:32:32 And so it's now going and finding all
1:32:35 this stuff. It's writing the code
1:32:39 to generate these things. Are we done?
1:32:42 View and export. We're done.
1:32:48 Play slides.
1:32:51 So,
1:32:53 this is essentially Jim Ross did this.
1:32:55 Jim Jim Ross was was asked to jump in um
1:33:00 to a keynote tweet
1:33:06 was supposed to happen,
1:33:08 the guy dropped out that was supposed to
1:33:10 do it and they reached out to Jim Ross
1:33:12 and said, "Hey, can you do this keynote
1:33:13 for us?" And he said, "Sure." And he did
1:33:17 what I just did here. He did in front of
1:33:20 the room. He said, 'I want to do, you
1:33:23 know, a presentation to this audience.
1:33:27 You know, I I assume he had a PDF there
1:33:29 based on this PDF I've got. And it
1:33:32 generated a a keynote that he then gave.
1:33:37 So, in front of the audience, he had one
1:33:39 of these tools generate the slides for
1:33:42 the keynote presentation that he then
1:33:44 just delivered. I thought that was
1:33:46 [ __ ] genius. And then of course at
1:33:49 the end of the presentation, what did
1:33:50 people want to talk about? Nothing about
1:33:53 what what his topic was. They were like,
1:33:55 "How did you do that?" Anyway, beyond
1:33:58 the typical tourist spots, discover
1:33:59 hidden hidden gems and memorable
1:34:01 experiences the whole family will love.
1:34:04 Um, why go off the beaten path? Unless
1:34:07 unleash your family's curiosity.
1:34:09 Denver's packed with surprises outside
1:34:11 the usual hotspots.
1:34:13 International
1:34:15 Church of Cannabis offbeat attraction.
1:34:17 The Money Museum. Oh, I've been to that.
1:34:19 That's cool. Um, Dinosaur Ridge.
1:34:23 Um, the Museum of Transportation.
1:34:26 The iconic Big Boy locomotive at the
1:34:29 Forny Museum. Look at this. Right. It
1:34:32 went and found the picture. It gave it a
1:34:34 caption. Home to over 600 transportation
1:34:37 artifacts.
1:34:42 Unconventional family activities.
1:34:47 Rhino Art District. That's where I work
1:34:48 is in Rhino. I know that that mural
1:34:51 actually.
1:34:53 Plan your own offbeat adventure. There
1:34:55 you go.
1:35:02 Boom. Done.
1:35:06 Crazy
1:35:09 crazy crazy.
1:35:13 So,
1:35:17 so this is the kind of stuff this the
1:35:20 kind of stuff where um
1:35:26 what do you do with capability like
1:35:28 that? I don't know
1:35:34 where are the B you know like a lot of
1:35:37 what this channel for a lot of the way I
1:35:40 think about this channel is that my job
1:35:45 or the job that I've assigned myself I
1:35:47 don't know that this is my job the job
1:35:50 I've assigned myself in this channel is
1:35:53 find stuff I think is interesting go
1:35:56 play with it
1:35:59 and figure out where the boundaries are,
1:36:03 right? Figure out where the edges are of
1:36:06 what works, what doesn't, what's janky,
1:36:08 what's not janky,
1:36:11 so that as the tools evolve
1:36:16 and something new comes along, we can
1:36:18 look at that and immediately know, oh,
1:36:21 that's better. That's good.
1:36:25 Right? because we know what it was
1:36:26 before.
1:36:28 All of these tools like GenSpark and
1:36:30 Manis and and OpenAI agent
1:36:34 are really janky right now. There's lots
1:36:36 to there are lots of edges to discover
1:36:38 where they're not good at [ __ ]
1:36:41 But I have a feeling they're going to
1:36:42 evolve very very quickly. And I think
1:36:47 exploring different ways to use these
1:36:50 automated multi-step tools
1:36:53 that are not the same [ __ ] everyone else
1:36:55 has been doing, I it feels to me is
1:36:58 really really important.
1:37:04 So
1:37:05 there's your um homework for the
1:37:08 weekend. go to Manis, go to GenSpark,
1:37:11 play with them over the weekend, and
1:37:13 that way when OpenAI agent shows up next
1:37:17 week, it should show up Monday, but it
1:37:19 might be, you know, next week sometime,
1:37:23 have a good sense of what it can do and
1:37:26 and compare it to what Gen Spark did,
1:37:29 what Manis did,
1:37:32 and start to get your head around these
1:37:34 things are different. What can I do with
1:37:35 them? What can I do with them? What can
1:37:36 I do with them? And my request would be
1:37:38 if you have good ideas and breakthroughs
1:37:40 like Josh did tonight, um, come share
1:37:43 your ideas. Um, super sweet. You can
1:37:46 also fact check it, check it after it
1:37:49 finishes and edit the slides. Cool.
1:37:50 Yeah. Like one thing I would do with
1:37:52 that is I would tell it I want it to
1:37:54 make a slide per destination. It kind of
1:37:57 consolidated some of them in the here's
1:37:59 some [ __ ] to do with your family. Like
1:38:01 break those all out as separate things.
1:38:03 But yeah, that's that's a really good
1:38:04 point. Yeah, none of this stuff that it
1:38:06 makes is final form. You can say, "I
1:38:09 don't like the design. Go make it look
1:38:11 more hip and colorful," and it'll go do
1:38:13 that, right? So,
1:38:15 absolutely crazy. All right. Um,
1:38:20 so let me Oh, you know what? Let me see
1:38:24 if I can go find Chef Kelly's thing and
1:38:27 then I'll share that.
1:38:32 Chef Kelly
1:38:36 since she's here.
1:38:41 Kelly Anderson, right?
1:38:50 Videos
1:38:53 introducing my alter ego. Okay, hang on
1:38:56 a sec. Let me change my sharing.
1:39:01 Share screen.
1:39:04 Okay. Um,
1:39:09 so if you haven't seen this,
1:39:13 um,
1:39:17 Chef Kelly's an irregular. She comes to
1:39:19 these things regularly.
1:39:22 Get it? and um
1:39:26 has basically
1:39:29 taken
1:39:32 her brand, her ideas, learned enough of
1:39:35 this AI [ __ ] and this is her first kind
1:39:38 of alter ego video, and I got to say,
1:39:41 for for a first video out of the gate,
1:39:43 like this is pretty [ __ ] remarkable.
1:39:45 It's really [ __ ] good. So, um so I
1:39:48 won't play the whole thing because it's
1:39:49 four minutes, but go check it out. She's
1:39:51 uh Kelly Anderson on LinkedIn and the
1:39:54 the name of this post is introducing my
1:39:59 alter ego. All right.
1:40:03 >> Can you believe that red dye number
1:40:04 three has been linked to cancer since
1:40:06 the 1980s?
1:40:08 Somehow it's remained in your food, your
1:40:10 drinks, your medications for over 40
1:40:13 years. Why? Because the FDA, the
1:40:16 regulatory government system that's
1:40:18 supposed to protect us, was never really
1:40:21 designed to do so in the first place.
1:40:23 This is the story of how easy it is for
1:40:26 dangerous additives to get FDA approval,
1:40:29 stay legal, and end up in your kitchen.
1:40:32 >> Like good editing. I assume, Kelly, that
1:40:35 you've done video editing before because
1:40:36 the the editing's good, the B-roll is
1:40:39 good, the use of the music is good. It's
1:40:41 just really well done.
1:40:44 >> Yes. Source Camp has shown this to
1:40:46 several people. The like
1:40:50 one of the things that I like most about
1:40:52 this is
1:40:58 this kind of output for me starts to
1:41:00 become an incredibly strong
1:41:02 counterargument
1:41:05 to all of the [ __ ] Like I got a lot
1:41:08 of hate on LinkedIn when I when I put
1:41:10 out um a LinkedIn post about chain of
1:41:13 craft. Meaning if you
1:41:18 if you
1:41:21 have a creative process where you know
1:41:24 you know you know you have a story you
1:41:26 want to tell and you put together you
1:41:28 know a series of activities that involve
1:41:30 generating images and music and this and
1:41:33 that and then you stitch it all
1:41:34 together. you can end up with something
1:41:36 good. And almost all of the the feedback
1:41:40 that came back in to that was, you know,
1:41:42 you're a loser for daring to say that
1:41:44 that AI could be in any way creative.
1:41:49 And what those people are are are
1:41:51 missing the boat on it was the whole
1:41:54 point of the article. They're missing
1:41:56 the boat that
1:41:58 the ability to push a button and
1:41:59 generate an image is not 100% of the
1:42:04 process with AI, right? It's never that.
1:42:07 It's like Kelly had this story she
1:42:10 wanted to tell. She she had to create
1:42:12 the character. She had to think about
1:42:13 the script. She had to create the
1:42:15 script. She had to create the B-roll.
1:42:17 She had to create, you know, the voice.
1:42:20 She had to edit it all together.
1:42:22 >> Dies began. And if if someone were to
1:42:25 watch this,
1:42:29 if what Chef Kelly never said was, "I
1:42:32 used AI to make this." There's nothing
1:42:35 in this that I can see that immediately
1:42:38 goes, "This is AI." And quite frankly, I
1:42:42 don't think it's anyone's [ __ ]
1:42:43 business what tool she used to make
1:42:46 this, right? Like, this is just a good
1:42:49 video. cold tar derivatives literally
1:42:51 the byproduct of burning coal in the
1:42:55 early 20th century hundreds were used in
1:42:57 every
1:42:58 >> right like there's a good example like
1:43:00 >> literally the byproduct of burning coal
1:43:04 >> like is that is that stock footage
1:43:07 should did she just have one of the
1:43:09 video tools generate that I don't know
1:43:13 does it matter no right the the the chef
1:43:18 is talking about burning coal. We see we
1:43:21 see that it brings that idea to life.
1:43:24 >> Hundreds.
1:43:24 >> And now we move on to the next part of
1:43:26 the story. That's just good
1:43:28 storytelling.
1:43:31 It's irrelevant what tool she used or if
1:43:33 she got that from a clip library or if
1:43:35 she generated it.
1:43:38 It's irrelevant whether she generated
1:43:41 this character with AI and did the voice
1:43:43 with AI or hired a 3D animation studio
1:43:48 out of Hollywood and spent $50,000
1:43:51 creating a rigged model.
1:43:53 Do we care? No.
1:43:56 Like it's compelling. Like
1:43:59 just good storytelling. I got a good
1:44:02 character here. There's clear branding
1:44:03 in the kitchen. Like I know what the I
1:44:05 know what everything is about here.
1:44:08 So
1:44:11 this is a really good example if if you
1:44:14 want to share with people um the future
1:44:18 of AI. The future of AI is we stop
1:44:21 talking about AI,
1:44:24 right? We stop talking about AI. We just
1:44:26 talk about, hey, I saw this cool video
1:44:28 on on on dyes. Did you know that that
1:44:31 they've known these things cause cancer
1:44:33 since the 80s and we're still sticking
1:44:34 them in our bodies?
1:44:37 That's what to talk about.
1:44:41 How did you use a Hey, Kelly, what tool
1:44:43 did you use for that? Oh. Oh, you use
1:44:45 Mid Journey. Yeah. Yeah. No, I've used
1:44:47 Mid Journey. Yeah. No, it's I you know,
1:44:49 I think I I tried it once. I didn't like
1:44:51 it.
1:44:54 That shouldn't be the [ __ ]
1:44:55 conversation. Period. End of story.
1:45:02 It's not meltdown Mondays, but I'm
1:45:03 taking this opportunity
1:45:07 I'll put on my pink [ __ ] bow. I'll do
1:45:09 it.
1:45:17 All right. Well,
1:45:24 literally over half the posts on
1:45:25 LinkedIn today were created by AI.
1:45:28 People just said don't say they used it.
1:45:30 Yeah, exactly. They [ __ ] hypocrites.
1:45:34 By the way, I think this is amazing. I
1:45:35 absolutely agree. If a company like
1:45:38 Pixar did the exact same thing, people
1:45:40 would think it was amazing. Well, I but
1:45:42 the my whole point about Kelly's thing
1:45:44 is I think that thing's just really
1:45:46 good. I think it's amazing as is
1:45:48 independent of how she made it.
1:45:52 And that's that's kind of that's kind of
1:45:55 my point of where we need to be headed
1:45:56 with all this stuff.
1:45:59 Like even with the Manis and and
1:46:01 GenSpark and and AI agent thing, like
1:46:05 what are we using them to do? If what we
1:46:08 can do is we can use those in a way that
1:46:10 it outputs something that you can put in
1:46:13 the world and people go, "Oh, wow.
1:46:14 That's really good.
1:46:16 Who cares how you got there? You're the
1:46:19 one that manifested it. You're the one
1:46:20 that learned the tools well enough to
1:46:22 generate a something that they can then
1:46:25 look at and just enjoy.
1:46:30 So anyway, yeah, there you go. If this
1:46:33 came out two years ago, people would
1:46:34 have thought I was an animator. Exactly.
1:46:38 They still may. Do you know, Kelly, that
1:46:41 I would I would argue that probably more
1:46:44 than half of the people that see that
1:46:46 video have no clue, zero clue,
1:46:52 that how you made that is even possible.
1:46:57 that they look at that and you are
1:46:59 effectively you've hired someone that
1:47:02 used to work at Pixar.
1:47:07 We're we're in a very
1:47:10 weird place in history. It's also I
1:47:13 personally feel this is a very magical
1:47:14 place in history. It's why I come go
1:47:17 live five nights a week
1:47:20 because we're never going to be in this
1:47:22 moment in history again. Do you know
1:47:24 that three years from now, Kelly, or
1:47:27 maybe it's probably three three years
1:47:30 from now,
1:47:33 AI will have saturated the world like
1:47:36 the the internet took, you know,
1:47:38 whatever it was, probably
1:47:42 15 years, 20 years maybe, to saturate
1:47:45 the world, right? And now everyone knows
1:47:47 what the internet is. They know what the
1:47:49 worldwide web is. They know how to
1:47:50 communicate digitally. They know what
1:47:52 social media is. They know how to
1:47:54 connect data sources and share and right
1:47:57 that's all it's all just fabric of
1:47:59 communication now
1:48:02 three or four years from now people will
1:48:05 know that if you want something in the
1:48:07 world you literally just ask for it and
1:48:10 out it comes
1:48:12 but we're in this weird gap time where
1:48:16 some small percentage I still think it's
1:48:18 less than 10% some small percentage of
1:48:21 people actually know what's going on
1:48:23 with AI and 90% of the people don't have
1:48:26 a [ __ ] clue
1:48:30 and they look at what we're doing and
1:48:32 they're like, "Uh, how did they do
1:48:33 that?" Wait, Chef Kelly, I thought she
1:48:36 was a Is she like super loaded? Is she
1:48:40 Did she hire an animator? She made that
1:48:42 herself. How's she find time to cook?
1:48:46 You know, right? Like, we're in this
1:48:48 weird time. So, anyway, congrats on
1:48:51 that.
1:48:53 All right,
1:48:54 the creator is here. Well done. I know.
1:48:56 Exactly, right? Chef Kelly is here. So,
1:48:59 yeah, she just made that thing. Really
1:49:00 beautiful. All right, I'm gonna get out
1:49:02 of here. Um
1:49:04 trying to think what's going on next
1:49:06 week. Not a ton. Um
1:49:10 the week of July 28th. So, so July going
1:49:15 into August, I'm doing a fiveday
1:49:19 um AI crash course.
1:49:23 It's going to be five, it's going to be
1:49:24 like AI overview, AI in business, AI and
1:49:27 creativity
1:49:29 something something and then like a Q&A
1:49:30 day, a fiveday crash course. Not next
1:49:33 week, but the following week. I have
1:49:36 some work to do. Um but I want you to to
1:49:39 spread the word about it and I'm I'm
1:49:40 going to start promoting it next week.
1:49:43 Um, but anyway, I'll be back here
1:49:45 Monday. What I want you to work on over
1:49:47 the weekend is if you can do something
1:49:49 like Chef Kelly did, whip that up over
1:49:50 the weekend. That would be great. Um,
1:49:52 but I want you to use GenSpark and Manis
1:49:55 if you haven't used them before,
1:49:56 especially, but even if you have, go
1:49:59 back and and start thinking about
1:50:01 creative ways that you can use a tool
1:50:04 like that that that just does something
1:50:07 in a completely unexpected way. That
1:50:09 that to me feels important. and and uh
1:50:12 it'll it'll set you up well for when
1:50:15 agent shows up in chat GPT. All right.
1:50:19 All right. That's it, people. I'm
1:50:21 leaving. Have a fantastic weekend. Hope
1:50:23 you had fun tonight and didn't burn burn
1:50:26 your tongue on your hot pocket. All
1:50:27 right. Peace out.