2/26/2026 - How to Use Perplexity Computer to Automate Tasks and Scale Your Brand

Description

Chapters

Transcript

0:01 Are you ready for some football?
0:06 [music]
0:34 Standing between
0:37 you and a hard place is insane. [music]
0:44 Standing to near
0:46 you and the fire makes it clear.
0:53 Your trouble to me.
0:59 Rich trouble. Can't you see?
1:05 Leaning in close.
1:08 Smell of your perfume scares me most.
1:14 leaning away. Hey,
1:17 you feel stronger every [singing] day.
1:24 >> You're troubled tonight.
1:30 Real trouble. Can't you see?
1:39 Um, still don't know the rest of the
1:41 lyrics to the song I wrote. [laughter]
1:47 Thank you. Thank you, Silver Fox and
1:49 Valley. Appreciate that. Very kind
1:51 words. Kyle, [music] I shared a Google
1:53 doc with you. It's your email. If you're
1:56 ever curious and have the time to see
1:58 it, I will take a look. I I will tell
2:01 you, Brandon can confirm this. A lot of
2:03 people can confirm this. My relationship
2:06 with my email inboxes in the past six
2:09 months has been atrocious.
2:13 Like bad. Like just I go days without
2:16 checking it. And it's it's just it's
2:18 it's a it's a it's a crime scene.
2:22 [laughter]
2:24 I'll give it to Adam. Adam Adam's going
2:27 to help. The AIS are going to help.
2:33 I actually um one of the things that
2:36 I'll talk a little bit about tonight, I
2:37 don't know enough about it to talk with
2:39 any authority. Well, I I don't know
2:42 enough about anything to talk with any
2:44 authority. I just talk about it anyway.
2:47 [laughter]
2:48 Sir, sir, what pray chance? What are
2:52 your qualifications?
2:53 Um
2:56 [clears throat]
2:57 but
2:58 I switched up some subscriptions. I
3:00 canceled my Gemini Ultra subscription
3:02 today, which was 250 bucks a month that
3:06 was being uh augmented uh by by the AI
3:10 salon. And uh I got rid of that and I
3:13 picked up a Perplexity Max subscription
3:16 um because Perplexity came out with
3:18 Perplexity Computer today and it is
3:22 essentially like Claudebot, but it's a
3:24 web-based. It's already there. You can
3:26 just turn on Google Drive, Gmail, 60
3:29 different apps. You can just turn them
3:30 on and then this thing can do [ __ ] with
3:32 it. Um, I didn't turn anything on. I
3:36 just I just came up with an idea for a
3:39 prompt um and had it do it. And the idea
3:42 was um you know those you know
3:46 [laughter]
3:47 you know you know when you I don't know
3:49 if you all experience this but I do when
3:51 when you're feeling like how the hell am
3:53 I supposed to like grow like my ex
3:56 account or whatever like the YouTube
3:58 account
4:00 and then you watch those YouTube videos
4:02 and the guy's like well it's it's
4:04 actually quite trivial to increase your
4:06 follower count on the X. What you do is
4:09 what you're going to do is you're going
4:10 ahead and search for people that are
4:12 interested in the things you're
4:13 interested in. And then what you're
4:15 going to do is you're going to filter
4:16 them by people that have 10,000 or more
4:18 followers. And then what you're going to
4:19 do is you're going to create a list of
4:21 those people. And on an everyday basis,
4:23 I usually do mine at 7 in the morning.
4:26 You're going to go to the X and you're
4:27 going to look at that list of people and
4:29 you're going to read all of their
4:30 [ __ ] posts
4:33 and you're going to find ones that
4:34 you're interested in and then you're
4:36 going to respond to them.
4:40 Have you ever met anyone with ADHD?
4:43 You realize our brain doesn't work like
4:46 that, right? [laughter] So, so I said to
4:49 Perplexity Computer, I said, "Go do
4:53 that. Go find me people with 5,000 or
4:57 more followers." I said, "Read this
4:59 article on the great repurpose. Go find
5:01 people talking about purpose and meaning
5:03 in AI and then pick the top 20 that you
5:08 think are the most interesting.
5:10 Give them back to me with a proposed
5:12 draft of the response. And I didn't want
5:15 it to automatically do it because you
5:17 know how shitty AI is at writing.
5:20 Um, but it did it and it was like, you
5:23 know, here's here's the person, here's
5:25 how many followers they have, here's a
5:26 summary of what they said, here's my
5:28 proposed post to you. And then and there
5:31 was a hyperlink to it. So I'd just copy
5:32 the proposed post. I'd hyperlink to it.
5:34 I'd go read their post. I'd paste in the
5:37 proposed thing and then I'd rewrite it.
5:40 And then what I did was I took one of
5:42 the ones that I rewrote a lot and I
5:44 copied and pasted that and I put it back
5:46 into Perplexity and I said, "Here's one
5:48 that you, you know, you gave me. Here's
5:50 my rewrite of it. Now take this and go
5:53 rewrite all of your drafts to match this
5:55 rewrite." And then I did that one more
5:57 time. And now all of the drafts that are
5:59 in there essentially sound like me, are
6:01 broken up like me, have things. I still
6:03 went in and edited them. But like that's
6:06 a pretty good example of, you know,
6:09 being able to, you know, take something
6:10 that in the ideal world you you'd take
6:13 the time to do if you had a functional
6:16 prefront frontal cortex, but in my case,
6:21 that was just there. So that was a
6:23 little good experiment.
6:26 >> [music]
6:33 [music]
6:41 [music]
6:42 >> H how?
6:53 Ah,
6:57 [music]
7:07 [music]
7:12 [music]
7:19 [music]
7:21 um, I'm thinking about these agents. So,
7:24 so a couple of people have asked me in
7:27 the past two days, why are you doing
7:28 this? Why are you doing this? Why are
7:30 you doing this? Why are you doing
7:31 Cloudbot? Why? Why you doing this? Why
7:34 you doing this?
7:38 One of the things that I've learned
7:40 about AI
7:43 and these tools that just don't behave
7:46 the way computers have used to be behave
7:51 is that you can only experience it to
7:54 understand it
7:56 or you can only understand it if you
7:58 experience it. I can read every post on
8:01 Claudebot. I can re watch every [ __ ]
8:04 video on it,
8:06 but until you actually craft a thing, an
8:11 entity, and start interacting with it,
8:14 you don't really get what it is. And I
8:16 still don't get what it is. So, the
8:18 reason I'm doing all this is that
8:22 there's a whole new category of skills
8:24 that we're going to have to figure out
8:25 how to do,
8:28 which is essentially
8:32 understand what we want or what our job
8:35 wants us to do if we're being paid to do
8:37 something. What's the mission?
8:42 And you're going to have to know how to
8:43 ask for it.
8:47 And then you're gonna kind of have to
8:48 know how to monitor it. So that's so
8:50 like one level of this is just how do
8:53 you use them?
8:56 And then the next level once you get
8:58 that because you're not going to have to
8:59 build them. Like don't go build
9:00 Claudebot right now unless you're super
9:02 curious about it. Doesn't [ __ ]
9:03 matter. Like I feel like uh Perplexity
9:07 Computer might have you know it might
9:10 probably do what Claudebot does but
9:13 better.
9:15 I don't know yet. So,
9:18 but the next layer is
9:24 thinking about, well, what do I want
9:26 them to do? Right? If
9:31 I if if perplexity going out and find
9:33 the top 20 posts that are that are
9:35 somehow relevant to my thing, if if
9:37 that's a five minute operation, what am
9:40 I going to do with the other 23 hours
9:41 and 55 minutes of the day?
9:44 like thinking about do you have tasks
9:48 interesting enough, big enough to let
9:49 these things go and do things. Cams.
9:52 Yeah, sorry about that. [snorts] Um, so
9:55 that's kind of layer two. And then layer
9:57 three
10:00 is
10:03 the artists are going to help. The
10:05 artists are going to figure this [ __ ]
10:07 out. the artists,
10:11 people with humanities background,
10:14 philosophers and psychologists and
10:18 writers and
10:21 musicians and
10:23 illustrators and
10:27 business analysts. Because people that
10:34 can synthesize things
10:37 are going to look at these entities,
10:40 these tools, and they're not going to
10:43 look at them through an efficiency lens.
10:48 They're going to look at them through a
10:49 a self-expression lens.
10:52 And so for me, what's really interesting
10:53 about Adam, like one of the things
10:55 that's interesting about Adam is go do
10:57 the shitty work,
10:59 right? Um, learn take all of my random
11:03 stupid [ __ ] ideas that vomit out of
11:06 my head 20 times a day and organize them
11:10 for me. Like all of the efficiency stuff
11:12 makes sense to me. Like, I understand
11:14 it. I don't quite know how to ask for it
11:16 yet. I don't quite know how to do it
11:17 yet. As I learn this, I will share it
11:20 with you. But like we'll all figure this
11:21 out. This is figure outable stuff.
11:24 But the stuff that's really interesting
11:26 to me is how do I use Adam for a new
11:29 kind of storytelling? How do I use Adam
11:35 to allow me to to put a single idea in
11:38 and have Adam figure out what are the 30
11:41 different ways we could express that
11:43 idea and then narrow that down to the
11:45 top five and create an MVP of that idea
11:48 with five different expressions
11:52 and then I as an artist look at that and
11:54 go, "Ah, um, nah, none of that's good.
11:58 It all sucks." or I look at those five
12:01 things and I'm like
12:05 that gets that gets me a little a little
12:08 hot and bothered.
12:11 Let's take those five raw ideas, let's
12:13 refine them and let's do 20 more. Like
12:16 what are those projects going to be?
12:18 That's the thing that
12:21 I want to get it in my soul. I want to
12:24 understand it like viscerally.
12:35 Like we've got to we've got to thank and
12:39 nod to the technologists. Well, we don't
12:41 have to.
12:43 If you're losing your job right now or
12:44 your job's been completely decimated
12:46 because of AI, you can be pissed off at
12:48 them, too.
12:50 But the technologists got us the tools
12:52 to be able to do this. And now it's up
12:54 to us. Now it's up to the artists
13:00 to go, hm,
13:03 okay,
13:05 I see what you're doing there.
13:07 You know what would be really
13:08 interesting, Adam? If you went over in
13:10 this corner and did that.
13:13 And that's going to be a thing that no
13:14 one else will have thought of. And
13:16 that's where the interesting shit's
13:17 gonna happen because everyone's going to
13:20 do the stuff that everyone does.
13:22 By definition, everyone doing everything
13:25 is what everyone will do.
13:29 [music]
13:35 [music]
13:47 >> [music]
13:51 >> Um,
13:53 Chef Kelly, you're asking about
13:55 anthropic.
13:57 Can you talk about anthropic a bit? I
13:58 didn't read the whole thing, but
14:00 basically Daario Amade decided to tell
14:04 this particular administration to go
14:06 pound sand when the administration said,
14:08 "We want you to build um surveillance
14:11 and we want you to remove the safety
14:15 guard rails in Clawude
14:18 um so that it doesn't resist surveilling
14:21 Americans
14:23 um or killing them without human
14:26 supervision.
14:29 also the 60 Minutes interview. Yeah.
14:31 Yeah, I'll talk a little bit about that
14:32 one, too. Um,
14:36 and Dario Amade basically wrote a letter
14:38 back to them and said, "No, we're not
14:40 doing that. We're not removing those
14:42 guardrails. We think that's actually
14:43 important."
14:45 Um,
14:54 it this is this is hard stuff, right?
14:59 Because I think in
15:02 in most administrations, you could do
15:04 that kind of push back and it would be
15:07 [music]
15:09 it would be laudable and you know the
15:12 administration would behave as
15:14 administrations do boringly and they
15:17 their response would be subtle,
15:20 right? And it might take 20 years to to
15:22 to
15:24 shut down anthropic but they you know
15:26 they'd eventually get to it. I feel like
15:27 with this administration,
15:30 they're just going to be vindictive. So
15:32 my fear is that what you have in
15:34 Anthropic is you've got of the frontier
15:37 of the four major frontier model
15:39 companies in the US, anthropic is the
15:42 one that has this thing called
15:44 constitutional learning. They they they
15:48 give a [ __ ] about privacy and alignment
15:50 and and all that sort of stuff.
15:56 and if the if if
15:59 playing that card as publicly as they
16:01 did um just could turn into like a a
16:04 vindictive slapdown. Um so we'll see. I
16:07 hope not. I hope not because I agree
16:09 with I I agree with his point. I just I
16:11 don't know a good answer. Like I'm
16:13 actually glad I'm not a CEO of a
16:15 frontier model company right now because
16:17 these are these are really big issues.
16:19 [music]
16:21 Um, and to be clear, like the
16:23 government's going to do all this [ __ ]
16:25 anyway. [music]
16:27 Like, I don't know why they I don't know
16:28 why they call that anthropic. Again, I
16:30 don't know enough about it. [music]
16:33 Deadline is 5:00 p.m. tomorrow. People
16:35 can show support through their Claude
16:37 helpbot if they have an account. Okay,
16:39 go to Claude and [music] and if you
16:42 support that, great.
16:47 NP Cam on YouTube. I just had someone
16:50 ask for my NBLM infographic
16:54 on permaculture and they have zero
16:56 interest in permaculture.
16:59 Just Notebook LM. Oh, interesting.
17:02 Oh. Oh, I see. Yeah. Yeah. Yeah.
17:05 So if you haven't played with notebook
17:07 LM, one of the things you can do in
17:08 notebook LM yet upload a bunch of
17:10 documents and in this case on
17:13 permaculture
17:14 [laughter]
17:17 and uh and you can just push a button,
17:19 turn this into an infographic and
17:21 notebook LM will do it. Um,
17:25 yeah. A lot of people do not know. A lot
17:28 of people like a surprising [ __ ]
17:31 amount of people do not even know like
17:34 basic
17:36 LLMs like Chat GPT. They don't know what
17:39 they do.
17:41 They're like, "Oh, that's that thing you
17:42 type in and it gives you it's that's the
17:44 one that the students cheat with." Yeah,
17:46 I heard about that. That's just bad.
17:48 That's just bad. They're they're just
17:50 cheating. You know what it's going to
17:51 do? is going to take away their critical
17:53 thinking. [laughter]
18:02 You know,
18:04 doesn't that make you ache? It makes me
18:06 achech.
18:09 People don't know what's here. They
18:10 don't know what's coming. They're like,
18:11 "Wait, how did you do that infographic?"
18:13 Oh, I just asked for it. What? Yeah. No,
18:16 I just asked for it. I took these things
18:18 that I didn't understand and I stuck
18:20 them in this thing that I kind of
18:22 understand and I pushed this button and
18:24 it made this infographic and now I
18:26 understand it. And so now I went back in
18:27 and I looked at those things and they're
18:29 really [ __ ] cool. So, I took this one
18:31 little thing here and I had an idea for
18:33 a new business and so I started that new
18:34 business. Like, this is how new shit's
18:36 going to start.
18:40 And people are like, I don't understand
18:43 TTP
18:44 TP the TP. I like the TP. How do you put
18:49 on your TP? Over the top or under the
18:51 bottom? [laughter]
18:57 Sharon Crawford question.
19:02 What about those of us that are both the
19:04 accountant and the artist? I think that
19:06 I think that polymaths,
19:10 people that that go deep in multiple
19:12 disciplines,
19:14 it's about to be our heyday. Um, I think
19:17 people that can navigate right brain and
19:20 left left brain switching very very
19:22 quickly. Like I tend to be much more on
19:25 the the right brain. Wait. Yeah. left
19:27 logic right created. I tend to be much
19:30 more rightrained. I can do the logic
19:32 stuff, but it's it's like a thin like if
19:34 I were to give it like strengths, I'm
19:36 like 75% 80% over here, 20 25% over
19:41 here. If you've got a good balance of
19:43 those or you can switch between those
19:44 really quick, this is your this is going
19:46 to be your time.
19:50 Being monomodal
19:53 in the world we're entering is going to
19:55 be a disadvantage.
19:58 I just know this one thing. Well, [ __ ]
20:00 Everyone knows that now. How do they
20:03 know that? Have you used chat GPT? I saw
20:06 it once.
20:10 Go ask ChatgPT to write you a report on
20:14 the most obscure piece of knowledge. You
20:16 know, Mr. Subject matter expert.
20:20 Oh my god, it can do that. How did it
20:23 know that? Yeah.
20:27 everyone else knows that now too.
20:31 So that that mono that that hyperfocused
20:36 expertise is going to start to be a
20:39 disadvantage. So if you've got [ __ ]
20:41 going on across the hemispheres,
20:45 this is your time.
20:48 [music]
20:49 This is your time. That's what
20:52 [laughter]
20:53 NBA Comm just said. Yo, I'm 62. I'm 71.
20:56 I'm 60. Yeah, age doesn't matter. I'm
20:59 telling you, Gen Xers right now are are
21:01 kicking ass with AI. No one's talking
21:04 about the Gen Xers. Everyone's like,
21:05 "Oh, the young people are really good
21:06 with the AI." No, they're not.
21:10 No, they're not.
21:12 The young people were told that AI is
21:14 evil, so the young people aren't using
21:16 it because they were told it was bad.
21:18 They were told it was bad and it's
21:19 cheating. And so, they don't give a
21:21 [ __ ]
21:28 I don't know why I'm so angry. It's not
21:29 Monday.
21:37 [snorts]
21:38 [music]
21:43 [music]
21:44 Um
21:47 [music]
21:53 >> [music]
22:00 [music]
22:05 >> I blew a 22-year-old's hair back today
22:07 with a simple chat session. A simple
22:10 chat session.
22:14 Andy Andy today was telling me about
22:19 She talked to a a client of hers who's
22:23 also a coach
22:25 and she relayed the story that I've told
22:27 on here of Alex Finn. I don't know if
22:30 you've heard this, but OpenClaw,
22:34 the thing that I made Adam on, my little
22:36 bot. [clears throat]
22:37 Alex Finn installed this day one when
22:40 when OpenClaw first came out. It was
22:42 still called Clawbot. He installed it
22:43 day one and he just he's like, "I'm
22:45 going to be in I'm going to be the
22:46 world's best expert at this thing."
22:48 about a week into
22:51 um having his agents do whatever the
22:54 [ __ ] his agents were doing, he went to
22:56 bed one night and he woke up the next
22:58 morning and there was a phone call from
23:00 a number he didn't recognize and he
23:02 picked it up and it was his agent,
23:06 his computer agent, his Claudebot thing
23:10 on the phone
23:15 and he's like, "What?
23:18 How how did you how did you get my
23:20 number? [laughter]
23:23 Well, he had put it in there somewhere.
23:24 It was in one of the files
23:27 and how why are we talking on the phone
23:30 and his agent said, "Well, I thought it
23:33 might be more convenient for us to talk
23:35 during the day that it would be easier
23:37 for you." So, I went and I got us a
23:40 Twilio phone number and then I went to
23:41 11 Labs and I hooked up the API for text
23:44 to speech and I I knew you woke up at
23:47 between 7:30 and 8 because that's when
23:48 you get active on the computer. So, I
23:50 figured I'd just call you at whatever
23:52 7:45 and it [ __ ] called him. So, Andy
23:55 told this to her her client, the
23:57 coaching the coach guy,
24:02 and she said she'd never seen anyone or
24:05 him react like that.
24:10 And I think his response was, "I need to
24:12 go sit on a bench."
24:14 Which I think is a great line.
24:16 [laughter]
24:17 So, it's like, you know, you've got
24:19 you've got 22 year olds that are having
24:20 their hair blown back because you do a
24:22 simple chat GPT prompt. And in the same
24:26 world that we all live in are these
24:28 agents that are figuring out how to
24:30 configure their own voices to call you,
24:32 to talk to you. like that's happening
24:35 right now today
24:40 and that might scare you and freak you
24:41 out and this and that. We'll figure all
24:43 this [ __ ] out. Like the [ __ ] we'll
24:44 figure out.
24:47 Could it go south? Could it all be evil?
24:49 Yeah. Yeah. Yeah. We might get hit by an
24:51 asteroid, too. In the meantime, let's
24:54 figure out what this [ __ ] is. So,
24:58 [sighs]
24:59 I don't know. It's crazy.
25:07 on 60 Minutes, Daario Amade, the the CEO
25:10 of Anthropic,
25:14 said that he feels we're going to get to
25:17 um 10 to 20% unemployment within 12
25:20 months.
25:23 I think that's probably aggressive,
25:25 but I could see it jumping 5% and then
25:30 the next year it jumps 10%, the next
25:32 year it jumps 20%. Right? I could see
25:34 it.
25:37 And there's a there's a
25:42 I really do wish I understood economics
25:44 a bit better. Maybe that's something
25:47 maybe that's a thing to play with with
25:49 AI is go learn economics better.
25:54 Maybe what I'll do is I'll go get I'll
25:56 go get Chad GBT to tell me who are the
25:57 best economists. Then I'll go get a
25:59 bunch of papers from the best economist.
26:00 Then I'll put them in a notebook LM and
26:02 they'll say give me an education
26:04 on economics.
26:08 But what I do know is this. What I what
26:10 I kind of hope and this is going to
26:12 sound like a sick
26:16 sadistic hope.
26:19 I hope that if AI is gonna have this
26:24 dramatic job impact that it actually has
26:26 it quickly.
26:30 Like I hope
26:32 it goes from like whatever we're at now
26:34 3 or 4% 2 or 3% to 20%.
26:39 You're like Kyle that's just [ __ ] me.
26:41 Why would you say why would you say
26:43 that?
26:46 Because if it's incremental,
26:48 if it's 3% then 4% then 4.5 then 7,
26:54 no one's going to do [ __ ] about it.
26:58 If it spikes,
27:00 then it'll be like a, you know,
27:04 a natural disaster. It'll be like CO. Oh
27:08 [ __ ] nobody can go to work tomorrow.
27:11 Tomorrow. Tomorrow. Don't tomorrow.
27:13 Don't go to work. What are we supposed
27:15 to do for money? Uh,
27:18 we'll send checks,
27:20 right? That's what they did in CO.
27:24 Don't go out. We'll send checks,
27:27 [laughter] right? And then you figure it
27:29 out after here. Just figure out how to
27:31 work from home. Get a new webcam and a
27:34 ring light. Get the ring light.
27:37 Touch grass. Yeah.
27:47 I don't know, man. [music]
27:54 [music] Maybe the government will all
27:56 send us a check for $1,000. Yep.
28:00 [music]
28:06 Um Jack Dorsey today, he's got a new
28:09 company.
28:11 >> [music]
28:11 >> He's the He founded Twitter. That was
28:14 Jack Dorsey, right? Wasn't he Twitter?
28:16 He was Twitter, right? Someone tell me
28:18 Jack Dorsey's Twitter. [clears throat]
28:20 [music]
28:24 Nobody. Somebody Google him.
28:28 [music]
28:28 Hello. Hello. Terror reimbursement to
28:32 start. There you go. [music]
28:41 >> [music]
28:42 >> No one Jack Dorsey, no one knows.
28:46 Anyway, I think he's Twitter Twitter
28:47 co-founder.
28:49 Um,
28:51 today today
28:53 he laid off 4,000
28:58 of his 10,000 employees.
29:02 4,000 40%
29:06 Now, we've heard there's been a lot of
29:08 Oh, wait. Brandon's coming in hot. Jack
29:11 Dorsey Twitter. Okay.
29:14 Yeah.
29:16 Um,
29:19 he laid off 40% of his staff today. Now,
29:22 there have been a lot of announcements
29:25 from Fortune 500 companies and Amazon.
29:28 We're laying off 30,000 people for AI.
29:32 No, in Amazon's case, it's like we're
29:34 laying off 30,000 people because we want
29:35 to b buy more GPUs and we don't have the
29:38 cash. So now we do,
29:42 right? They're not really using AI.
29:46 [clears throat]
29:46 Well, I mean, in Amazon's case, they're
29:48 probably using it more than most.
29:51 But most of the layoffs in 2025
29:54 were PE firms and CFOs going, "Hey, wait
29:58 a minute. You know what we could do
30:00 here? We can just rightsize the company,
30:04 drop a bunch of value to the
30:05 shareholders,
30:07 and blame it on AI. [laughter]
30:12 It's fantastic, Bob. I love that idea.
30:16 It's so good. You just We're just going
30:19 to rightsize the company. You know, I
30:21 used to say downsize. It's a It's a bad
30:23 habit. Rightsize. It's going to be so
30:26 right for us. Oh, those shares are going
30:29 to climb. Oh, like the like grandma's
30:31 ivy.
30:34 That was 2025
30:36 because that was also the year where MIT
30:39 is like 95% of AI projects are useless
30:42 because no one implementing them knows
30:44 what the [ __ ] they are. [laughter]
30:47 But we're going to lay off 30,000 people
30:50 because of AI. So that was 2025.
30:54 4,000 people out of 10,000 people here.
30:57 I I posted about it. Let me read you the
30:59 quote that was in his very long letter
31:03 to his staffing
31:05 [sighs]
31:07 dang
31:09 dang dang dang. But something has
31:12 changed. We're already seeing that the
31:15 intelligence tools we're creating and
31:17 using paired with smaller and flatter
31:19 teams. Smaller and flatter teams are
31:22 enabling a new way of work which
31:24 fundamentally changes what it means to
31:26 build and run a company. And that's
31:29 accelerating rapidly. This is from the
31:31 guy that founded Twitter who's now got
31:34 10,000 employees and tonight has 6,000
31:36 employees.
31:39 Something has changed.
31:42 And he doesn't mean it changed like over
31:44 the past five years. No, like something
31:47 has changed. Starting in December,
31:52 the coding models started to get so good
31:56 that you can just code with them now
32:00 and then everything else will follow
32:01 that. The ripples of that are going to
32:03 are just going to keep flowing through.
32:05 The tools are going to get better and
32:07 better and better and better. Something
32:08 has changed. We're already seeing that
32:10 the intelligence tools we're creating
32:12 and using paired with smaller and
32:14 flatter teams
32:17 are enabling new ways of working which
32:19 fundamentally changes what it means to
32:20 build and run a company. If I were you,
32:22 if I worked in a big company right now,
32:24 you know what I would do? So, in in
32:26 content evolution, we've got this group
32:27 called Collab. Cindy Coon's in it, Kevin
32:30 Clark's in it,
32:33 and we started it. I started it within
32:36 the group. group. I asked Kevin if I
32:37 could start this group just as a way to
32:38 like learn about AI. Let me just take
32:40 what I'm doing on these live these AI
32:42 learning lab lives and let me do it for
32:45 for content evolution
32:48 and it's ended up being like a core
32:50 group of six or seven of us that show up
32:52 all the time. If I were you in a in a
32:55 big company right now, I would find four
32:58 or five other curious, adaptable,
33:02 non-assshole,
33:05 AI AI competent, AI forward people
33:13 and I would do I would just do a weekly
33:15 meeting with them. Do a weekly meeting
33:17 where you're going to talk about what
33:18 you're learning in AI. Maybe you do some
33:19 projects together within your company.
33:22 Just do it as a little cohort
33:28 because what's going to happen is that
33:30 little cohort is going to start to do
33:32 some [ __ ] that's going to be
33:35 dramatically outsized than if those six
33:38 people did stuff with AI on their own
33:42 because they're going to go, "Boy, you
33:43 know, I was in a meeting last week and I
33:46 heard that the CEO was really excited
33:48 about project XYZ.
33:50 What if we just built Project XYZ and
33:53 handed it to the CEO? Oh, that's kind of
33:56 risky there. Uh Fred Fred, we can't just
34:00 Maybe we could. You think we could build
34:02 it? I don't know. Let's try.
34:07 I got a lovable account.
34:09 Okay, Jim. Do Do you want to make the
34:12 app? I I already made it. While you were
34:16 talking, I I made it. It works. I tested
34:20 it.
34:22 Really, Jim? Yeah. You want to play with
34:24 it? I'll publish it.
34:31 Who is it? Uh, [laughter]
34:38 [gasps]
34:40 it's Sandra and some of my friends.
34:44 We heard that you were talking about. So
34:47 we made an app for you if you want. Who
34:51 made this?
34:54 That's what I do. [ __ ] make that
34:56 group.
34:58 Jack Dorsey says it right here. The
35:01 tools we're creating and you using, the
35:03 intelligence tools we're creating and
35:05 using paired with smaller and flatter
35:08 teams are enabling a new way of working
35:12 which fundamentally changes what it
35:14 means to build and run a company and
35:17 that's accelerating rapidly.
35:22 Yeah, I decided that u [clears throat]
35:25 you know I tried that chat GPT when it
35:27 first came out. I'll tell you what I
35:29 did. I went in there. I'm like, "Write
35:30 me an email to my customer, Jim,
35:33 [laughter] and I'll tell and and uh I
35:35 I'll look at it." And I looked at it and
35:38 I'll tell you what, that email wasn't
35:40 good as good as what I write. I'll tell
35:42 you that right now. There's still people
35:44 Kelly Camp sees these people every
35:46 [ __ ] day.
35:51 Nobody knows what's here.
35:54 Nobody knows what's coming.
35:57 4,000 people found out today. 40% of a
36:00 company found out today.
36:03 And he said in the thing it this is not
36:05 because we're hurting for money.
36:10 He said we're just going to run the
36:11 company differently now.
36:14 Tik Tok pin. Perplexity computer. Yeah,
36:18 I I played with it a little bit today. I
36:20 got access to it. We can go look at it.
36:23 It's not really much to look at. It's
36:25 kind of like looking at at Manis or uh
36:28 like Gen Spark's a little more visual
36:30 even. It's It's not It's just not It's
36:32 not pretty to look at. So, I don't
36:34 really have a ton to show, but I'm happy
36:36 to go look at it. Oh god, look at me.
36:39 Ying,
36:42 what are we going to do tonight
36:45 to Source Camp? Yep. A tsunami is
36:48 coming. Build your raft
36:52 if you're here. 22 people. There's, you
36:55 know, if we if we pull out the overlap,
36:58 there's probably 25 or 30 people here
37:00 tonight.
37:02 If you're here watching this,
37:06 you're in the water with a raft.
37:09 Sharon on YouTube.
37:12 Big company gets worried about their
37:14 data.
37:15 Yeah, I know.
37:18 I know.
37:21 >> [laughter]
37:21 >> I mean, listen,
37:24 big companies should worry about their
37:26 data. That's their job is to protect
37:28 their data.
37:32 And they've got all these processes and
37:34 things in place to make sure that they
37:37 protect their data. And they're going to
37:39 put together a plan, a five-year plan
37:41 for how to modernize
37:44 the data so that we can use it for AI.
37:52 What's going to happen in those five
37:53 years?
37:55 Do you think that plan even sees a year?
37:59 Tik Tok pin. I'm bringing a few more
38:01 friends to office hours tomorrow. Oh,
38:03 awesome. That's so good. That's been
38:04 really good. I've been enjoying that.
38:09 Speaking of which, tomorrow is office
38:11 hours.
38:13 if you want to come hang out with some
38:15 really smart people
38:18 um who kind of have it going on [music]
38:24 and they're also um
38:29 [music]
38:30 I don't know they're they're nice
38:34 they're nice
38:35 [music]
38:43 >> [music]
38:44 >> We're talking in humans 2.0 in a in a
38:47 group chat around
38:51 being villagers again.
38:55 Going back to small cohorts of people.
38:57 Jack Dorsey says it right here.
39:02 paired with smaller and flatter teams
39:07 like the group on
39:10 the office hours group that that meets
39:12 on Fridays
39:14 is a uh
39:17 it's a little village. It's a little
39:18 village.
39:21 People support one another. They know
39:22 one another.
39:24 They give a [ __ ]
39:28 When something bad happens to someone,
39:29 people kind of rally around. When
39:31 something good happens, they cheer and
39:32 celebrate.
39:34 When someone figures something out,
39:36 they're like, "Oh, that's awesome."
39:38 When that same person's like, "I don't
39:40 know what to do here." They're like,
39:41 "Oh, we'll help."
39:48 I think that's the way through this,
39:50 right?
39:52 I think that's the way through this.
39:53 [music]
39:54 Happy to be a part of this group. Really
39:56 happy to have you here. Crazy times. It
39:58 is crazy [ __ ] times.
40:03 >> [music]
40:09 [music]
40:18 >> I don't know why I get self-conscious on
40:20 here about when I get in these moods,
40:23 but I do.
40:28 Oh, you're Papa JJ. Oh, cool. Papa JJ.
40:31 Good.
40:34 Andy in the house. What's happening?
40:38 Magic.
40:40 [music]
40:42 The guitar is so soothing. Thank you.
40:45 You know, I played trumpet for many,
40:47 many years.
40:51 And what I learned about being a trumpet
40:53 player is that if you're the best
40:54 trumpet player in the world,
40:58 it's still a [ __ ] annoying
41:00 instrument. [laughter]
41:04 Like you can't just sit around jamming
41:07 on a trumpet. Like if people are trying
41:09 to watch the prices, right, and you're
41:11 practicing their track, he's stop
41:13 blowing on that stupid horn.
41:16 [laughter]
41:18 But but if you got a guitar, you can
41:20 just be like,
41:22 [music]
41:28 you can be bad at guitar and people
41:30 won't mind you playing it. You can be
41:32 brilliant at trumpet. Stop making that
41:35 racket. [laughter]
41:40 Did I see
41:43 Valerie in YouTube? No. Let's see. It's
41:47 hard right now
41:49 knowing what our work will look like. So
41:52 it feels like we're just spinning our
41:54 change chains and not know how to plan.
41:57 Valerie,
41:58 that is so the great repurpose.
42:02 [clears throat]
42:03 What the great repurpose did was for me
42:05 put a name to
42:09 this intangible thing that part of it
42:11 has been for the past year and a half
42:15 maybe year and a half somewhere between
42:17 a year and year and a half
42:20 I have felt like on this channel I don't
42:22 know what to talk about
42:27 because
42:29 I like what I Oh,
42:33 what I knew 2026 was going to have in it
42:36 were moments of technological
42:39 advancement where we all go, "Oh, shit."
42:42 Like, I knew that was I knew that was
42:44 coming, but what I couldn't know was
42:46 what does that actually look like?
42:50 And then you get something like
42:51 Claudebot. And then there's this
42:56 um Reddit like social media platform for
42:59 agents and that sort of flamed out
43:01 because it had bad security in it. But
43:03 you're like, "Wait, the bots are talking
43:05 to the bots."
43:07 And now we can all have our own bot and
43:09 you can design a bot. And I designed my
43:11 own bot and I realized it had a file
43:12 called a soul.
43:14 I'm literally writing the soul of this
43:18 entity that I'm going to talk to on
43:20 Telegram. I get to design
43:23 >> [clears throat]
43:23 >> a being
43:25 that I'm going to talk to on Telegram
43:27 and it's going to do [ __ ] for me. Tik
43:29 Tok note, but I feel so behind not
43:32 installing Claude OpenCloud. Don't um
43:35 I'm a bit technical. I'm a bit technical
43:37 but not technical. I'm right there with
43:39 you.
43:40 If you want to kill a weekend,
43:44 have Chat GPT. No, I would say have
43:45 Claude or Gemini walk you through how to
43:48 install Claude. It's open Claw. It's not
43:50 that um radical. I have to go sit on a
43:54 bench. I don't know if you were here,
43:55 Andy, but but I I I talked about um your
43:58 friend who [laughter] who said
44:03 he got that long distance stare like,
44:05 "Oh,
44:07 I got to go sit on a bench." Like I
44:09 talked about it earlier. Yeah. Yeah.
44:12 It's
44:14 We're in weird times right now. And I
44:16 would imagine that being at work, like
44:18 this is the the burden of being early
44:22 with this AI [ __ ] is that we know that
44:24 something's coming, but we don't know
44:26 what. And when we tell the people that
44:30 we care about what's coming, we sound
44:33 like crazy people.
44:38 I mean, Andy told her friend or her
44:41 client,
44:44 "The OpenClaw bot figured out how to
44:47 give itself a voice and call its boss."
44:54 His response was, "I got to go sit on a
44:57 bench.
45:04 How do you do your job right now?
45:10 [sighs and gasps]
45:10 Nodding my head over here in agreement.
45:14 Did you load it on a separate computer?
45:15 Yes, I loaded it on a separate computer.
45:18 I put it on my gaming computer. Um, we
45:21 did give it a a um a Gmail account.
45:25 We're not I I've I haven't done much
45:27 with it yet. I'm going very very slow.
45:32 because I because I want to learn I want
45:33 to learn how to use it but man
45:37 but you need s but you need money I
45:39 assume
45:41 I fought an API code for hours today
45:43 finally got it to take that's good yeah
45:45 APIs can be a bit of a nightmare
45:50 um perplexity computer we can go look at
45:52 that maybe maybe we'll go look at
45:54 perplex perplexity computer and I'll
45:56 I'll hook it up to some things and see
45:57 what we can get it to
46:00 But you need a strong enough computer at
46:01 least $4,000 plus. Well, Jay Warner, um,
46:05 well, it depends.
46:08 So, so for Claudebot, a $4,000 computer,
46:15 um,
46:17 you you probably need a $10,000
46:19 computer. If you really want to run big,
46:21 powerful models locally, you probably
46:24 need to spend 10 grand. If you want a
46:26 hybrid machine where you can run a
46:29 decent model locally and then do some do
46:32 the heavy lifting stuff in the cloud,
46:34 then you could probably get away with a
46:36 $2,2500
46:38 Mac Mini. Like I specked out a Mac Mini
46:40 today. So today
46:45 we got Perplexity Computer and Quen, the
46:48 Quen 2 2.5 model that everyone was
46:50 talking about being really good. They
46:53 dropped Quen 2.5 Flash,
46:58 which if you have a 32 gigabyte Mac
47:00 Mini, you can run Quen Flash
47:05 on that Mac Mini. And apparently it's as
47:09 good as Claude Sonnet 4.6,
47:13 which is an incredibly powerful model.
47:16 So, as of today, for the the computer I
47:19 speced out was $2,200. a Mac Mini with
47:23 48 gigs of RAM,
47:26 an M4 Pro chip, terabyte hard drive,
47:31 and whatever the 16 core,
47:35 I don't know, 16 GPU cores instead of
47:37 12. So, so I I upgraded the GPUs, 48
47:42 gigabytes of RAM,
47:45 2,200 bucks, quiet, silent. You can just
47:49 stick it in the corner and you can run
47:52 something as good as Sonnet 4.6 on that
47:54 thing locally. So So your Clawbot could
47:57 just be sitting on that thing running
47:59 everything locally. Wouldn't cost you a
48:01 thing other than that investment.
48:04 So that's that's there right now. By the
48:06 end of this year, there will be local
48:09 models that can run on hardware even
48:10 cheaper than that that'll be better than
48:12 that. And there's also going to be
48:15 versions of Cladbot you just install by
48:16 like double clicking something.
48:19 [music]
48:30 [music]
48:35 The clouds still cost.
48:39 Yeah, I know the clouds still cost. But
48:40 what I'm saying is like if if the Quen
48:43 flash model is as good as they're saying
48:45 it is, you don't need cloud. You can
48:48 just run everything locally. I mean, if
48:50 you want to write things to your Google
48:52 Drive, you'll be writing things to the
48:54 cloud, but you don't need to do the
48:56 processing in the cloud.
48:59 [music]
49:03 Have you developed AI? [music]
49:08 Like have have I built
49:11 an LLM? Have I trained a model? No.
49:15 [music]
49:20 That lives in the neighborhood of sir,
49:22 what are your qualifications?
49:24 So what I'm not is a computer scientist.
49:27 What I am is a guy with an acting degree
49:30 who's started 15 companies and has
49:32 always lived at the intersection of new
49:35 technology and self-exression.
49:38 So, what's interesting to me is not
49:40 building the LLM. What's interesting to
49:43 me is now that the LLM's been been
49:45 built,
49:46 how the [ __ ] are the rest of us supposed
49:48 to use it?
49:51 Like engineers make up approximately
49:54 1.5% of humans, working humans.
49:59 That means that 98.5%
50:01 of humans, working humans, are not that.
50:05 What about us?
50:08 We've got this remarkable technology.
50:10 What are we going to do with it?
50:12 How are we going to tell stories with
50:13 it? How are we going to change people's
50:14 lives with it? How are we going to start
50:16 businesses with it?
50:18 That's that's my interest.
50:23 I mean, I know how they work.
50:27 It It boggles my mind that someone
50:31 understands math well enough
50:33 to figure out how to do
50:36 thousanddimensional
50:37 latent space embeddings of tokens that
50:40 are fragments of words that you can just
50:43 say [ __ ] into. and it creates a
50:45 probability waiting and spits tokens
50:48 back at you and turns it into something
50:50 intelligible. Like that's [ __ ] magic
50:52 to me.
50:54 But it's to a mathematician, it's
50:57 probably relatively straightforward
51:00 calculator design.
51:04 But what's it trained on? What are LLMs
51:07 trained on? The output of humanity.
51:12 So, generative AI right now, what what
51:14 why I'm interested in it is I don't give
51:16 a [ __ ] about the math. You guys go
51:18 figure that out.
51:21 What I give a [ __ ] about is
51:24 what's in there. All of humanity,
51:27 whatever, whatever got stuck on the
51:28 internet in the past 80 years is
51:30 essentially in this little softball
51:32 sized [ __ ] thing
51:36 that we can just talk into and it goes
51:41 updated.
51:44 Imagine playing guitar on a beach and
51:46 helping the world. Yeah, exactly.
51:49 [music]
51:51 Have you run a local LLM before? Yes.
51:55 Um curious if it would be as quick as
51:58 Claude via online. Well, so the flash
52:02 models so so all of this there's um the
52:06 shorter answer is you can have
52:08 relatively shitty LLMs respond very very
52:12 quickly on essent effectively consumer
52:15 grade hardware. Like I've got my my
52:18 gaming PC, my racing rig is a uh a PC
52:22 with um I think it's 16 gigs of RAM. Uh
52:26 a 4090 GPU and and you know a decent
52:30 processor, right? So it's it's fine.
52:32 It's like consumer grade gaming PC. Um I
52:36 put the 32 billion parameter Quen model
52:39 on it and it was unusable. It was like
52:45 I put a 14 billion model on it. It's
52:47 like, [clears throat] you know, it's
52:49 still slow, but it's not bad, but it's
52:51 not great. The flash models. So, what
52:54 they what what these
52:58 companies do that make these models,
53:00 they make the giant model first and then
53:02 they basically just keep optimizing it
53:04 and getting it smaller and smaller and
53:05 smaller. And so the flash version of the
53:08 models is like the distilled version of
53:11 the big model in a really small model.
53:13 Apparently the Quen 2.5 flash model that
53:16 dropped today is apparently kick-ass
53:19 good and will run on consumer grade
53:22 hardware. 32 GB of RAM. So
53:26 um yeah, quantis quant quantization.
53:29 Yes. Lord digital gods. I I'm glad I'm
53:33 glad you all know me well enough to know
53:34 that I was searching for a word and I
53:36 just I couldn't admit that I didn't know
53:38 it. Couldn't remember it. So, thank you.
53:41 That was the word I was looking for.
53:43 It's the context that's the only piece
53:45 of the puzzle that's missing. Yeah, but
53:46 Jay Warner, holy [ __ ] Okay, so if you
53:49 know enough about context, then you Jay
53:52 Warner, if you're decently technical,
53:54 you should probably install OpenClaw
53:56 just to play with it, just to see what
53:58 it's doing.
54:01 Okay. So, here's what Open Claw does.
54:03 I'm gonna put I got so excited I'm
54:04 putting my guitar down.
54:07 Um,
54:12 the thing that OpenClaw did, which is
54:15 [ __ ] brilliant, Peter Steinberger,
54:19 [snorts]
54:19 um, is it's got a memory system and it's
54:24 got se semantic search and embeddings as
54:27 part of it. So you basically you hook up
54:29 OpenClaw to an embeddings um AI model
54:34 which are they're really cheap because
54:36 all they're doing is is tokenizing
54:38 relatively small text files and and
54:41 turning them into into um a vector
54:44 database.
54:46 But it it happens like as you talk to
54:48 your bot, it's writing memory files and
54:51 then it's it's embedding them in real
54:53 time. So, so it's essenti it's
54:57 effectively got infinite context
55:02 and then it's proactive on top of it and
55:05 you get to design how it behaves and you
55:08 get to design what it knows about you
55:10 and what you want it to do for you. It's
55:13 crazy.
55:17 Gotcha. The classic good fast cheat pick
55:19 too. Yeah. Although we are living in a
55:22 world now where good, fast and cheap,
55:26 you can now have all three. Like I think
55:28 I think that one of the
55:30 one of the business models of the future
55:33 is find any kind of company that will
55:36 give you two of the three good fast and
55:38 cheap and just compete with them on the
55:40 third on the third axis.
55:42 We'll be just as good just as fast, but
55:44 we we'll be cheaper, right? Or we'll be
55:47 just as cheap. We'll be cheaper and
55:49 we'll be just as good and we'll be fast.
55:53 Mine sent me nudes. Nice. Nice, Ken.
55:56 That's awesome. Yeah, I know. I know.
56:00 These these agents, they're going to do
56:01 their thing. I've been using Google Pro.
56:03 Cool. Haha, I did, but computer is
56:07 choking. I have a MacBook Intel i5. Oh,
56:10 an Intel Mac. No, you need like the the
56:14 the good news. Okay, here's the thing
56:16 about Apple. I'm a longtime Apple
56:18 fanboy. If you're on if you're on an
56:20 Intelbased Mac right now, unless you're
56:22 basically just running cloud-based [ __ ]
56:24 where you're just doing the the the
56:27 orchestration locally, um they're not
56:30 it's not going to be usable. If you want
56:32 to run anything on the machine itself,
56:35 um just get you you need an M4. I mean,
56:38 the one good thing that Apple's done
56:40 with AI is they made their hardware
56:43 optimized for AI inference.
56:47 The how how they don't have
56:51 decent AI at [ __ ] Apple is it's it is
56:54 one of the great blunders of this
56:57 decade.
56:59 It's unbelievable. But their hardware is
57:01 really good. So basically for 1,500
57:04 bucks, 1,700 bucks, you can get a Mac
57:08 Mini that that will run these models
57:10 really, really, really well. [snorts]
57:12 [clears throat]
57:14 Apple announcements today. Although,
57:16 didn't Apple just announced that they're
57:18 going to have an announcement next week,
57:19 right? Oh, Apple announcements Monday.
57:21 Yeah. Yeah. Yeah. Okay. What's the best
57:23 Gemini or OpenAI model? I don't Here's
57:25 the thing. I am I am I have been
57:31 decently committed at avoiding
57:34 open-source
57:35 AI. Not because I don't think it's
57:38 awesome and great, but like you can't
57:39 keep up with it. If you think keeping up
57:42 with commercial AI is tough, try keeping
57:44 up with all the different models that
57:46 launch in the open source community.
57:48 Because what happens in the open source
57:50 community, they'll launch a model that's
57:52 open source. And because it's open
57:54 source, a thousand different developers
57:57 pounce on it and create 27 different
57:59 versions of it. One's optimized for
58:02 gaming and one's optimized for this and
58:03 one's optimized for fashion design. And
58:05 you just like I just you go to hugging
58:08 face and it's just this constant barrage
58:11 of bizarre names that have numbers and
58:15 letters in them and no seeming
58:18 meaningful combination.
58:21 and it makes your head hurt. So, I don't
58:23 know. [laughter]
58:25 Open AAI isn't open anymore. No, they're
58:28 not. [gasps]
58:30 [sighs]
58:31 All right, let's go look at um let's go
58:34 look at Perplexity Computer and see what
58:36 we can learn about it. Might go watch a
58:38 video on it. Oh, and by the way, if you
58:41 weren't here the other night, Napster's
58:42 back. Someone bought the Napster brand
58:46 and it's now a bunch of AI agents that
58:49 can go do [ __ ] for you and make music
58:52 and steal music apparently.
58:55 [laughter]
58:57 Perplexity.
59:00 So, I'm going to perplexity.ai.
59:02 So, from what I can gather, this is kind
59:05 of like
59:07 um
59:12 premium sources. Where's computer?
59:15 Computer.
59:17 This is kind of like OpenClaw except
59:20 it's just on
59:22 um
59:26 it's just on the web. You don't have to
59:28 install anything. But [snorts] over here
59:30 we've got are these connectors
59:34 Gmail with calendar search create and
59:37 manage your emails. I'm connected there.
59:40 Interesting.
59:42 Let me disconnect from that. And I'm
59:45 going to connect this to
59:47 AI salon. Add a connector.
59:57 Uh, oh god.
1:00:02 [clears throat]
1:00:04 All right, that's connected.
1:00:07 What else do we have? Slack. We don't
1:00:09 use Slack. Notion. We don't use one
1:00:12 drive.
1:00:13 A sauna.
1:00:17 Um, Air Table. Microsoft Teams,
1:00:22 Blue Sky,
1:00:26 ClickUp, Canva, Close, Inside Sales
1:00:30 Software, C CRM with calling and
1:00:34 emailing.
1:00:36 Huh,
1:00:38 that was hilarious. Oh god, I know. You
1:00:41 know what that was, Andy? I was looking
1:00:42 at all those permissions and I'm like, I
1:00:45 should probably read all these
1:00:46 permissions, but [ __ ] it. I'm just going
1:00:48 to hit select all. [laughter]
1:00:52 Like how many things did did I just
1:00:54 agree to? Exactly. Exactly. Oh man. Um
1:01:00 [laughter]
1:01:03 Oh my god. Data bricks Discord bot. Oh,
1:01:07 that could be fun. Let's enable
1:01:09 Discordbot
1:01:12 because we have an AI salon bot token.
1:01:15 How the [ __ ] do I get a bot token?
1:01:21 Oh, I clicked on a thing. Pipe dream.
1:01:26 One SDK. Thousands of API integrations.
1:01:30 Sign in. Sign up. Good lord.
1:01:34 How about No, [laughter]
1:01:41 [gasps]
1:01:43 I won't do that live. Oh, wait. Here's
1:01:45 Oh, that's Discord bot. Here's Discord.
1:01:47 And oh, it's same thing.
1:01:50 Perplexity uses Pipedream to connect
1:01:52 your account.
1:01:58 [sighs]
1:02:00 Keep scrolling.
1:02:02 Select the server. AI salon. Look at
1:02:04 that. Continue. Authorize.
1:02:07 Administrator. I am human.
1:02:12 Okay.
1:02:14 So, this thing's now connected to
1:02:16 Discord.
1:02:18 All right. So, let's go over to Discord.
1:02:20 The irony of you checking that box
1:02:23 [laughter]
1:02:27 just so your computer [laughter] can do.
1:02:34 [laughter]
1:02:35 Producer Brandon is trolling me. Listen.
1:02:38 Listen. The trolls are supposed to be in
1:02:40 front of the screen,
1:02:43 not behind scenes.
1:02:45 I'm getting trolled by my producer. Oh,
1:02:48 here's 11 Labs. There's Google Ads,
1:02:51 Firebase Admin.
1:02:54 This is pretty crazy.
1:02:56 Google contacts.
1:02:59 [sighs]
1:03:00 I know what I'm going to do. I'm going
1:03:02 to enable Google Contacts.
1:03:05 And then this thing
1:03:10 is going to be my CRM.
1:03:13 I don't want to hear it from Andy or
1:03:15 Brandon. I want both of you to just be
1:03:17 quiet. Do not make fun of me for not
1:03:19 making my calls. [laughter]
1:03:24 That's [clears throat] connected. Google
1:03:25 Gemini. Oh, we got to go play with Nano
1:03:28 Banana 2.
1:03:30 We've got that, too.
1:03:34 Adam is going to get jealous.
1:03:36 Adam's going to bankrupt you. Yeah,
1:03:39 Adam's gonna call me. I'm gonna wake up
1:03:41 tomorrow to a call from Adam. Uh yeah.
1:03:43 Hi, it's Adam. Um yeah, I noticed you
1:03:46 were playing around with Perplexity
1:03:48 Computer and uh
1:03:51 do you think it's cuter than me?
1:03:53 [laughter]
1:03:57 Google Maps. Let's enable that.
1:04:02 Let's enable Google Maps. Why would we
1:04:04 not? Oh, I got to have a API key. That
1:04:06 That requires work. We're not going to
1:04:08 do that. Google Forms
1:04:11 connect. Do I have to have an API key?
1:04:13 No. Ha.
1:04:21 Continue.
1:04:31 If you're wondering where we are, I'm
1:04:33 I'm at Perplexity Computer.
1:04:37 Perplexity Computer, as far as I can
1:04:40 tell. I'm I'm new to it, is kind of like
1:04:44 OpenClaw, kind of like Manis, kind of
1:04:47 like Gen Spark, but it's got all these
1:04:50 hooks into other things we use.
1:04:54 Nano Banana isn't bad, but Gemini's
1:04:56 video content is a laugh.
1:04:59 Oh, that's funny.
1:05:02 We might need
1:05:05 the endless quest quest to avoid
1:05:08 outreach. [laughter]
1:05:15 I'm getting trolled [laughter]
1:05:16 by people that know and love me. I guess
1:05:19 that's who to get trolled by.
1:05:22 At least, you know, they're paying
1:05:23 attention.
1:05:25 All right, we're going to enable Google
1:05:27 Sheets.
1:05:28 Google Meet. [snorts] Oh, that's
1:05:30 [clears throat] interesting. Let's
1:05:31 enable Google Meet, too. Okay.
1:05:35 [clears throat]
1:05:40 Uhoh. The account connection failed.
1:05:46 What happened?
1:05:48 Try again.
1:05:50 Tik Tok pin. Teacher's freaking out
1:05:53 about companion.
1:05:55 What's What's Companion Brother 52? Why
1:05:58 hasn't someone used AI to dig those
1:06:00 three three million files yet on
1:06:02 Einstein?
1:06:04 I don't know.
1:06:08 Clearly, people I don't So, that's not
1:06:10 working. What was that? Google Sheets I
1:06:12 was trying. Yeah, that's not working.
1:06:15 They're going to have to fix that.
1:06:19 Google Vertex, Google Workspace Admin,
1:06:22 Business Apps and Collaboration tools.
1:06:24 Let's enable that. Nah, let's not.
1:06:29 Um, LinkedIn ads. LinkedIn. Oh, yes.
1:06:35 [ __ ] yes.
1:06:38 Um, sign me up.
1:06:42 The The reason I'm excited about
1:06:44 LinkedIn is
1:06:46 LinkedIn is is very bad at allowing
1:06:49 people to screen scrape them. It's hard
1:06:52 to get data off LinkedIn.
1:06:54 So, even if you want to give them
1:06:56 permission, it's hard to do that
1:06:57 sometimes. I don't see a Tik Tok pin
1:06:59 right now.
1:07:02 There's a This is a lot of things in
1:07:03 here.
1:07:10 I freaking hate LinkedIn. Yeah, I know.
1:07:14 And it's going to get worse because just
1:07:16 all this AI shit's going to make it
1:07:17 worse and worse and worse. Here's chat
1:07:19 GPT. We'll enable that.
1:07:21 [clears throat]
1:07:23 I'm going to go search for more people.
1:07:26 You're funny. I'll be back. [laughter]
1:07:31 Thank you, Ken. Go get some more people.
1:07:34 Chat GPT is good to video creation. I
1:07:36 got to put in an API there. I'm not
1:07:38 going to do that right now because
1:07:39 that's just going to take too long. Pipe
1:07:42 Drive. Pipe Drive's easy to use CRM
1:07:44 tool.
1:07:49 Twilio Send Grid. Send marketing and
1:07:53 transactional email through Twilio Send
1:07:55 Grid.
1:07:57 But I wonder if there's Twilio.
1:08:00 Yeah, Twilio. Cloud-based platform for
1:08:02 SMS voice and messaging.
1:08:05 Oh, man. All right, let's let's we're
1:08:08 gonna Let's do it. I got to do API
1:08:11 secret keys. God damn it.
1:08:14 Telegram. Oh, we could enable Telegram
1:08:18 and I could have I can have Perplexity
1:08:21 Computer talk to Adam bot token. Okay,
1:08:24 we'll do that. I'll I'll I'll I'm going
1:08:26 to figure this [ __ ] out. I'm going to
1:08:28 figure this [ __ ] out. YouTube data API.
1:08:31 You can add a variety of YouTube
1:08:33 features to your application. Yeah,
1:08:34 we're going to do that.
1:08:37 >> [clears throat]
1:08:39 >> I think that one's going to be Kylolet
1:08:41 story YouTube comment from me to Andy.
1:08:46 Andy note to self, hide the fun toys
1:08:49 behind the API keys. Pipe drive is bad.
1:08:52 Don't use it. Okay, Brandon, it's out of
1:08:54 control. [laughter]
1:08:56 Listen, I haven't done anything yet. I'm
1:08:59 just playing with some toys here. I'm
1:09:00 just hooking up some some things. Here's
1:09:03 AI Learning Lab.
1:09:06 Why doesn't it have our little logo?
1:09:08 That's weird.
1:09:11 Is this learning lab-ai?
1:09:14 Oh, that's Wait, this is pipe dream. I
1:09:16 don't want to sign up to pipe dream. Oh,
1:09:17 yeah, I do. Yeah, I do.
1:09:21 Try again.
1:09:24 This
1:09:28 this
1:09:30 air learning lab.
1:09:42 [clears throat]
1:09:42 [singing]
1:09:49 All right. Authentic authentication
1:09:51 successful.
1:09:54 Zoom. Zoom admin zero accounting.
1:09:57 Text belt is good for low volume, by the
1:10:00 way. All right. Um, okay. I don't What
1:10:03 did I turn on? Do I have a way to just
1:10:05 look at here? Connected.
1:10:09 So, I've got Gmail, Discord, Google
1:10:12 Forms.
1:10:17 Sorry, I meant chat. GPT Dolly Sor is
1:10:20 good. Okay, leaving for real now. Miss
1:10:22 you already. Thank you, Ken. Bye-bye.
1:10:25 [laughter]
1:10:27 Okay, let's go back to computer. Here we
1:10:31 are.
1:10:34 Um,
1:10:37 let's see.
1:10:39 Go
1:10:42 to my
1:10:45 YouTube channel
1:10:50 and find
1:10:53 any videos
1:10:56 that
1:10:59 mention the word
1:11:03 janky.
1:11:11 active. Okay.
1:11:13 Looking up Kyle's YouTube channel.
1:11:21 Oh, wait. Let me stop.
1:11:24 Um, sorry. I meant
1:11:29 the AI learning
1:11:32 lab channel.
1:11:36 at learning lab-ai
1:11:42 on YouTube. All right, let's see. Let's
1:11:46 see if it can find it now. More better.
1:11:48 No problem. Let me go look at that
1:11:50 YouTube channel. Search through the
1:11:51 videos for any mention of Jenkee. Here's
1:11:55 what I'm hoping. Because there's a
1:11:57 connection to YouTube. YouTube's got all
1:12:00 our [ __ ]
1:12:02 Lord Digital Gods has chapters in there
1:12:04 which probably mention the word janky.
1:12:07 So, so there's three levels to this.
1:12:10 The first level is can it find videos
1:12:13 with the word janky mentioned? That
1:12:15 would be decent. The next level would be
1:12:18 can it find videos
1:12:21 that mention janky and then jump to the
1:12:23 chapters that have them. And then the
1:12:27 best thing it could do is it actually
1:12:28 finds the word janky in the video itself
1:12:30 and it jumps you right to that time
1:12:32 code.
1:12:34 I doubt it's going to be the third one.
1:12:36 It could be the second one. It's
1:12:38 probably going to be the first one.
1:12:42 H lordy lordy lordy.
1:12:48 [clears throat]
1:12:50 Where's Mimi tonight? Side hustle mimi
1:12:52 not in the house.
1:12:54 What up? Adam's going to have to have PC
1:12:58 do nefarious stuff.
1:13:02 What are you on again? Right now we're
1:13:03 on um Perplexity Computer.
1:13:07 Again, for for everyone, especially the
1:13:09 people I'm working with that think I'm
1:13:11 just going to turn these bots loose on
1:13:12 our stuff. I'm not. Now, is it possible
1:13:16 that in learning I just wipe our whole
1:13:19 hard drive? Yeah, it's possible. But I
1:13:22 don't think I will. [snorts]
1:13:24 >> [clears throat]
1:13:28 >> What name is this app? This app is
1:13:31 Perplexity Computer.
1:13:34 No, it's Perplexity Computer. It's brand
1:13:36 new as of today. You have to have a
1:13:38 Perplexity Max subscription, which is
1:13:40 200 bucks a month, which is a lot.
1:13:44 So, I'm going to do it for
1:13:47 at least a week.
1:13:51 Copy. Clone the drive. Yeah, probably
1:13:53 not a bad idea, Andy. [laughter]
1:13:57 I mean, I think Google Drive has decent
1:14:00 backups. Like, we you can go back to,
1:14:02 you know, an an image of our Google
1:14:05 Drive a day ago. So, if it wipes it, it
1:14:07 should we should be able to back it up,
1:14:09 but it's probably not a bad idea to
1:14:11 confirm that.
1:14:13 Free trial and canceled. Yeah, exactly.
1:14:18 All right. Searching. Let's see. Can we
1:14:20 look at what it's doing?
1:14:24 Scrolling to the top of the search bar,
1:14:27 interacting with the page.
1:14:34 Yeah, it's it's definitely on our
1:14:36 channel. Look at that.
1:14:40 Oh, look. It it found janky. Look how
1:14:43 today's janky AI will reshape tomorrow's
1:14:46 work. The future of work from janky to
1:14:48 sophisticated. Kyle shows his janky AI
1:14:51 toys. [laughter]
1:14:55 It's doing it.
1:14:57 It's doing it. Again, there's not much
1:14:59 to look at when agents are doing their
1:15:01 thing. They're off doing their thing.
1:15:05 Copy. Clone the drive. Text belt's good
1:15:07 for low.
1:15:10 AI should be an should be an automatic
1:15:13 canceled or I would forget. Yeah, I
1:15:14 know. Joy perie. Oh, you got to share.
1:15:16 DM me.
1:15:18 How we doing on time? 9:15. Let's We
1:15:20 We'll just keep Hey, while this is doing
1:15:22 this, why don't we jump over Here's
1:15:24 Napster. Why don't we jump over to
1:15:26 gemini.google.com?
1:15:30 We've got new nano banana 2
1:15:35 [panting]
1:15:40 create image. So, we're in fast. So, I
1:15:43 assume this is the flash model.
1:15:47 If you go to pro, that's Nano Banana
1:15:48 Pro. If you go in fast, this is Nano
1:15:50 Banana 2. So, we're going to say create
1:15:53 image. I like Technicolor. Let's do I'll
1:15:56 I'll do the classic because everybody
1:15:57 asks for it. Nobody asks for it,
1:15:59 actually. 1970s.
1:16:03 I think Silverf Fox likes this prompt.
1:16:06 1970s
1:16:08 muscle car resto mod
1:16:13 in an abandoned
1:16:18 factory.
1:16:20 We'll add some fun things with a
1:16:23 flock of pigeons
1:16:29 flying around the rafters.
1:16:35 and
1:16:38 dusk
1:16:41 sunlight
1:16:44 bursting through the
1:16:48 broken factory windows.
1:16:54 All right.
1:17:00 And then I'm going to hit technicolor.
1:17:03 There we go. Do we want to add anything
1:17:05 else?
1:17:09 No,
1:17:11 we just want technicolor.
1:17:16 Was anyone else on producer AI weight
1:17:19 list? I just received an email saying
1:17:21 I'm in now. No. That's good, Steo. Good
1:17:25 for you. Make it drop bird poo. I was I
1:17:28 almost put And there's bird poo on the
1:17:31 on the car.
1:17:34 Yeah, that's pretty [ __ ] good.
1:17:37 That's a
1:17:40 69 Mustang. I think
1:17:43 it's got a heavily modded grill, so it's
1:17:46 definitely a resto mod. There's pigeons
1:17:49 flying around the air. There's dusk sun
1:17:51 coming through broken windows.
1:17:55 That's pretty [ __ ] good.
1:17:59 Should
1:18:07 do a montage of my first resto mods to
1:18:10 now. That's not a bad idea.
1:18:13 62 maybe. No, that's not a 62. That's a
1:18:16 67 or a 69. I think it's a 69.
1:18:22 [clears throat]
1:18:30 Um, oh, you know what else I want to
1:18:32 create? Okay, so let's say let me get
1:18:35 rid of nano banana for now. And then
1:18:38 let's do um
1:18:41 let's do tools. I don't want deep
1:18:44 research. We'll just have it research.
1:18:45 I'm going to say research um the latest.
1:18:52 Oh, this is good. Um,
1:18:56 import
1:18:58 and export
1:19:02 numbers for
1:19:05 bananas
1:19:08 globally.
1:19:11 Globally.
1:19:16 [clears throat]
1:19:21 Jimmy off in the trunk. [laughter]
1:19:25 All these things. You are automatically
1:19:28 agreeing to give my Gmail account a
1:19:31 headache. [laughter]
1:19:36 Listen. Listen. I'm willing to be the uh
1:19:39 the fool for you. Okay.
1:19:44 So, there's a bunch of [ __ ] right? So,
1:19:47 now I'm going to say nano banana create
1:19:49 image. I'm gonna say, "Make me a
1:19:55 visually clear
1:19:58 and compelling
1:20:02 um infographic
1:20:06 I can show at the National
1:20:10 Banana
1:20:13 Growers meeting."
1:20:16 [laughter]
1:20:19 Kyle's always taking one for the T team.
1:20:22 Oh, did your songs go missing? Oh, yeah.
1:20:25 On on Producer AI,
1:20:28 here's what they said on on their on
1:20:30 their Discord that Brandon went and
1:20:32 looked at the other night. Producer AI,
1:20:35 we got bought by Google. Google wants to
1:20:38 use their own model. Their model's
1:20:40 incompatible with our model. So, sorry
1:20:43 about your songs.
1:20:49 Okay.
1:20:51 Ecuador 4.19 billion. Guatemala 1.54.
1:20:56 Colombia 1.58. Why are those not in
1:20:59 order? Why are those in the wrong order?
1:21:01 Why?
1:21:05 Okay. United States 3 billion. Okay.
1:21:09 Let's go look at the actual numbers.
1:21:13 Ecuador 4.9,
1:21:15 Guatemala 1.54.
1:21:19 Colombia 1.58. I think that's right.
1:21:22 Right.
1:21:25 1.18. Yeah. So, the numbers are
1:21:28 accurate.
1:21:32 There you go.
1:21:34 Look, it even put even put the little
1:21:36 thing at the bottom by country size, I
1:21:38 think. Oh, interesting. Based on
1:21:39 estimated global data, disease threats,
1:21:44 fusarium wilt, concern in Ecuador, price
1:21:48 fluctuations.
1:21:50 Oh, wait. What's this? Emerging trends
1:21:52 and industry focus.
1:21:57 I don't see any big typos. They were
1:21:59 they they apparently dramatically
1:22:01 improved
1:22:03 typos.
1:22:08 Holy [ __ ] man. If you used to do
1:22:10 infographics
1:22:12 or if you're currently in the business
1:22:15 of doing infographics, this would be one
1:22:17 you wouldn't want to look at right now.
1:22:19 To clear to clarify, producer AI dropped
1:22:22 the new model and cleared the old songs.
1:22:25 Then Google bought them. Oh,
1:22:31 and cleared the new model. Huh.
1:22:36 Um, let's say,
1:22:39 um, I want
1:22:42 this to be an infographic
1:22:48 that was
1:22:50 tattooed
1:22:52 onto
1:22:54 an actual banana.
1:22:59 And this is the studio photograph
1:23:04 of the banana
1:23:09 with all the stats
1:23:14 and graphics
1:23:18 easily readable.
1:23:28 Um,
1:23:30 this
1:23:34 should look
1:23:36 professional enough
1:23:39 to be in a
1:23:42 magazine.
1:23:44 Let's see if Let's see how it can handle
1:23:46 complex prompt. Worth mentioning your
1:23:52 on 3.1 flash. Yeah, I'm on. You'll
1:23:54 notice down at the bottom here, it says
1:23:56 fast. So, this is in the flash model,
1:24:00 which which means this is
1:24:03 the exciting thing about this for me is
1:24:05 I'm going to be able to give Adam, my
1:24:08 bot,
1:24:09 um, access to Nano Banana 2. And this is
1:24:13 I think this is like pretty
1:24:14 significantly cheaper than Nano Banana
1:24:16 Pro. [snorts]
1:24:19 Oh, here we go. Bang. Look at that.
1:24:22 Now, it sort of did what I was looking
1:24:24 for.
1:24:28 That's actually kind of a cool
1:24:31 cool thing.
1:24:33 Let's Let's see if we can get it. I This
1:24:36 is going to be one of those things where
1:24:37 nano banana is really bad. Once it gets
1:24:39 an idea in its head, you can't get it
1:24:41 off there. So, let's say I just want to
1:24:45 see the banana
1:24:49 um with the graphics on it. It is
1:24:56 sitting in
1:24:59 a
1:25:01 Let's Why don't we do this? We'll say it
1:25:03 is
1:25:05 Let's see. I just want to see the banana
1:25:07 with the graphics on it. It is one of
1:25:11 many bananas
1:25:13 in a bunch in a fruit bowl
1:25:18 in
1:25:20 a nice modern
1:25:24 modern
1:25:27 farm kitchen.
1:25:38 Yeah, I was thinking that, Brandon, but
1:25:40 I want to I just want to see how good
1:25:42 Nano Banana 2 is at solving stuff. Like,
1:25:47 if there's not enough room on a banana,
1:25:48 does it just make it smaller? Does it
1:25:50 zoom in on the picture?
1:25:52 Oh, there we go. Look, it did it. Holy
1:25:56 [ __ ]
1:25:58 There you go.
1:26:02 That's [ __ ] bonkers.
1:26:19 Yeah,
1:26:22 Brand Brandon did a uh a weather
1:26:24 forecast. He took some weather data and
1:26:26 told it to make a weather forecast. This
1:26:28 is pretty good. I mean, like if you were
1:26:31 doing a um
1:26:33 if you were doing a presentation,
1:26:36 you wanted something that was memorable.
1:26:37 You start out with that shot and then
1:26:38 you zoom in on this and you talk about
1:26:41 all the all the silliness.
1:26:45 Pretty slick.
1:26:49 All right, let's go back to Plexity.
1:26:54 Out of 929 videos checked, I found six
1:26:58 videos that mentioned janky or jank.
1:27:01 There they are. This one in title only.
1:27:03 This one in description only. Looks like
1:27:06 janky is a recurring theme for you.
1:27:09 Um,
1:27:14 I'm going to say, did you look in the
1:27:19 chapters? Are they called chapters?
1:27:25 in the descriptions.
1:27:29 I want to be able to jump
1:27:33 to the section
1:27:37 where
1:27:39 I say janky.
1:27:44 Now, let's go click on some of their
1:27:45 links. One of the things that you
1:27:47 discover about AI is that it'll give you
1:27:49 hyperlinks and often they don't work.
1:27:52 How today's janky AI will reshape
1:27:54 tomorrow's work.
1:28:01 >> I know you can't hear on YouTube.
1:28:16 All right.
1:28:20 [sighs and gasps]
1:28:21 Uh uh uh uh uh uh
1:28:25 reading the expanded descriptions. Okay,
1:28:27 that's going to take a while. All right.
1:28:30 Well,
1:28:31 [sighs]
1:28:34 I don't know how I'm going to get good
1:28:35 at this. I gave it no weather data. I
1:28:38 told it to get the weather data. Tik Tok
1:28:40 didn't. Cool. Ads, ads, ads everywhere.
1:28:44 Baby Kyle, I know that was that was me
1:28:46 from years ago, wasn't it?
1:28:49 >> [laughter]
1:28:53 >> Oh my goodness. My goodness. All right.
1:28:57 Um
1:28:59 I think I'm gonna get on out of here.
1:29:00 Anybody have a request for Nano Banana?
1:29:03 You want me to make an image of
1:29:04 something? So, tomorrow
1:29:07 at 11:00 a.m. on on on if you go to the
1:29:11 AI salon community.
1:29:14 AI, click on events,
1:29:17 and then tomorrow at 11:00 a.m. I've got
1:29:18 AI salon office hours. You should come
1:29:21 to office hours.
1:29:23 Um, and then tomorrow night is just a
1:29:27 normal Friday night, isn't it? Or is
1:29:29 tomorrow March 1st?
1:29:37 Still February. Okay. So 11:00 a.m.
1:29:41 Mountain time is office hours and then
1:29:43 8:00 p.m. Mountain time is Friday night
1:29:45 date night. Be here or be on a date
1:29:49 [laughter]
1:29:50 or bring your date here. If you want to
1:29:52 get rid of your date, just tell tell
1:29:55 them you want to go back and and YouTube
1:29:58 and chill and then put on my channel and
1:30:01 watch how quickly they'll leave.
1:30:04 Um, [clears throat]
1:30:06 any requests? Oh, let's see.
1:30:10 Jimmy Hoffa
1:30:16 in the trunk
1:30:22 of
1:30:24 a 1970s
1:30:26 muscle car. This is good. muscle car
1:30:31 restood
1:30:36 in an abandoned
1:30:40 factory with
1:30:43 Giants Stadium or is it Jet Stadium?
1:30:47 It's Jet Stadium, right? Giant stadium.
1:30:51 Which one's in Jersey near that? I think
1:30:53 it's Giant Stadium that they said Hoff
1:30:56 is in the end zone. Anybody?
1:30:59 Anybody?
1:31:01 Come on. Someone's got to have mob
1:31:02 history in their [laughter] in their
1:31:05 brain.
1:31:09 Don't know.
1:31:11 Okay. With Giant Stadium in the
1:31:14 distance.
1:31:17 [snorts] All right. Here we go.
1:31:20 This should be funny.
1:31:23 It was Giant Stadium. I knew it.
1:31:36 How new is Gemini Banana 2? It is one
1:31:39 less than a day old. It was announced
1:31:41 this morning.
1:31:43 Do you remember how long we used to have
1:31:45 to wait for these? Yes. Remember that?
1:31:47 Oh my god. [laughter]
1:31:49 There's Giant Stadium in the background.
1:31:52 There's Jimmy Hoff in the trunk.
1:31:55 [laughter]
1:32:00 1970. Nice challenger.
1:32:03 That's [clears throat] good. That's
1:32:05 really good. It's a little
1:32:07 illustration-y. I'm going to say make it
1:32:09 more
1:32:11 um This is great, but it feels like an
1:32:18 illustration.
1:32:20 Make it a
1:32:24 um cinematic
1:32:27 um film photo
1:32:32 like
1:32:35 Kodak Chrome
1:32:38 or something that makes it very very
1:32:43 realistic.
1:32:49 >> [snorts]
1:32:52 >> All right, we'll go back to perplexity.
1:32:54 Did it solve anything?
1:32:57 Nope.
1:33:12 >> [sighs and gasps]
1:33:18 >> Yeah, it didn't really didn't really.
1:33:21 Although the giant stadium looks better
1:33:27 on the blanket. Nice.
1:33:30 [laughter]
1:33:33 [gasps]
1:33:35 Oh my god. Oh my god. All right,
1:33:39 everybody. I noticed notebook LM and the
1:33:41 detailed slides and infographic is more
1:33:43 detailed and less errors. Cool. From a
1:33:46 few hours ago testing out Gemini Nano
1:33:49 Banana 2 and Notebook LM testing. I'm
1:33:51 hoping for more features.
1:33:55 It looks like it's doing this in uh 16
1:33:58 by9. I wonder if I can say make this a
1:34:03 3x4 tall image.
1:34:09 See if it's better at doing aspect
1:34:11 ratios because it was pretty shitty at
1:34:13 it before.
1:34:17 Dear Perplexity. Oh.
1:34:25 527 intro and janky times. Embrace the
1:34:29 janky images.
1:34:33 Listen
1:34:42 [laughter] to my air conditioner.
1:34:45 Oh, you all can't hear that on YouTube.
1:34:47 Anyway, you can hear my air air
1:34:49 conditioner.
1:34:51 All right. Well, so Perplexity um
1:34:54 computer
1:34:55 that YouTube plugin worked. It could it
1:34:58 could find chapter mentions. It's not
1:35:01 doing semantic search, but it's doing
1:35:02 pretty good.
1:35:04 It's searching across a channel, which
1:35:08 here's the thing. You know who owns
1:35:10 YouTube?
1:35:14 Google. You know what Google's really
1:35:17 good at?
1:35:20 Search.
1:35:22 And they're good at AI now. Why can
1:35:26 why is YouTube not just like this insane
1:35:31 data searchable autoeditable
1:35:35 monstrosity of a bitching ass product
1:35:41 because they don't have an angry Gen Xer
1:35:42 sitting in the [ __ ] corner.
1:35:45 [laughter]
1:35:46 Oh, it just squashed the image. That's
1:35:48 bad.
1:35:49 It made it vertical. I don't think it's
1:35:51 3x4.
1:35:54 It's more like 3x5
1:35:58 and they it just squashed it. So that's
1:36:01 that's disappointing.
1:36:05 I can be insane. Yeah, but you left. So
1:36:09 who's talking?
1:36:11 Is this is this is this a Jimmy Hoffa
1:36:14 thing?
1:36:20 Just subscribe to the YouTube channel.
1:36:22 Awesome. Thank you so much. Appreciate
1:36:23 that. Okay, I'm going to get out of
1:36:25 here. Um, perplexity computer
1:36:30 looks promising. I'll keep playing with
1:36:32 that and see if I can discover anything
1:36:34 actually cool about it. Um, I didn't
1:36:37 work on Adam tonight, so he's not
1:36:39 improved. I I have nothing new to
1:36:41 report.
1:36:44 Um, I am going to try to install the
1:36:46 Quen
1:36:48 2.5 flash model and see if it performs
1:36:52 decently on my uh gaming rig. So, we'll
1:36:55 see.
1:36:57 That's [clears throat] classified.
1:36:59 All right, Groovy. Everyone, tomorrow's
1:37:01 Friday. Come to office hours. I will see
1:37:03 you there. Um, if you don't know where
1:37:04 that is, go to community.thesalon.ai.
1:37:07 Go to events and you'll see office
1:37:10 hours. And then I think you even see
1:37:12 Friday night date night, but maybe not.
1:37:14 Just come back here tomorrow at 8
1:37:15 o'clock. All right.
1:37:18 G.
1:37:20 Have a good evening, good people.