4/10/2025 - From Emotional Depths to Snapping Turtles: A Journey Through AI and Creativity

Description

Chapters

Transcript

0:01 Yay!
0:03 Yay!
0:04 [Music]
0:19 Yay! You're feeling nervous, aren't you,
0:23 boy? With your quiet
0:26 voice and impeccable style.
0:32 Don't ever let them steal a
0:35 joy and your gentle ways to keep them
0:39 from running
0:41 [Music]
0:43 wild.
0:45 They dress you down and tell you that
0:48 your place is in the middle when they
0:50 hate the way you shine.
0:55 I see you tugging on your
0:57 shirt, trying to hide inside of it. Hide
1:01 how much it
1:03 hurts. Let them laugh while they
1:08 care. Let them
1:11 spin. Let them
1:16 scatter. I've been to the
1:19 movies. I've seen how it ends.
1:23 And the jokes on
1:26 [Music]
1:35 them. All right. So, do me a favor. If
1:38 you're if you're out there on on any of
1:40 the channels, uh share this with people.
1:42 We got five, I think maybe six new
1:45 things to play with tonight. Um, one of
1:48 them is actually quite significant.
1:51 Um, but they're all they're all pretty
1:54 big
1:56 deals. See, we didn't go live last
1:59 night, but I think all of these things
2:02 were announced today.
2:05 [Music]
2:17 [Applause]
2:20 [Music]
2:33 Oh man. So, we're going to get to some
2:36 of that. We'll do some good stuff, some
2:38 good action.
2:40 [Music]
3:02 What's new in AI today? We're going to
3:04 get to that. I've
3:05 got one, two, three, four, five, six.
3:11 six things I want to show, which means
3:14 we'll probably get to three of them
3:15 because, you know, chat add and
3:17 [Music]
3:31 all. Hello humans.
3:34 [Music]
3:44 [Applause]
3:46 [Music]
3:49 [Applause]
3:51 [Music]
4:00 Every time I see it
4:02 now, get that look in
4:06 mind. Every time I see your mouth,
4:10 I hear that smile.
4:15 Early misty morning light. I heard the
4:19 engine
4:22 turn
4:24 [Music]
4:29 outside. You were leaving
4:33 me again today.
4:36 [Music]
4:38 You will convince
4:40 me again
4:42 [Music]
4:44 today. You're leaving this hot town
4:47 looking for someone else's golden ring.
4:52 [Music]
5:07 [Music]
5:13 What? What? What? You want
5:18 love? Jeffy's a little needy
5:21 tonight. What's happening, Silver Fox?
5:28 [Music]
5:53 Sasha, good to see you, too.
5:56 Welcome
5:59 everybody. Welcome, welcome,
6:01 welcome. I got my large language model
6:04 rag on Discord. Very cool. Some
6:07 advanced, some advanced geek in there.
6:10 Very nice. I like it.
6:12 [Music]
6:15 Joker, I did a stupid thing
6:18 today. No, Kyle, please say it ain't so.
6:21 You did something stupid.
6:27 Um, here. Let's see. Let me share share
6:31 this. Oh, I'm not even sharing nothing.
6:33 I ain't sharing
6:39 nothing.
6:41 Blackb. So, I did 80s pop
6:44 self-portraits. They kind of look
6:46 Chinese, but it's it's So, what I did
6:49 was I uploaded my picture to Chat GPT.
6:51 ChatGBT
6:55 And then I said, "Make me a portrait
6:57 like Andy Warhol did it. And then like
7:00 Keith Herring did it. And then like Roy
7:03 Likenstein did it. And then like Bosot
7:05 did
7:06 [Laughter]
7:10 it. It's the 80s. New York City." Yeah.
7:13 Yeah. Yeah. Yeah. Yeah. Yeah.
7:18 [Laughter]
7:22 Oh
7:23 [Music]
7:28 yeah, Kyle Jung Young.
7:31 [Music]
8:34 Oh, good lord. Good people. All right.
8:36 What are we going to do tonight? I think
8:38 we
8:39 [Music]
8:52 should. So Samman couldn't sleep last
8:57 night.
8:58 And truth be told, the reason he
9:01 couldn't sleep was because they have too
9:04 many [ __ ]
9:10 models. I'm low in the StreamYard frame.
9:13 That's because my camera has been
9:17 maladjusted. the the uh so if you don't
9:21 know it, this is a this may look like a
9:23 shitty home office, but I'm actually in
9:25 a $20 million uh TV production studio.
9:30 And uh my uh my cameraman uh had to take
9:34 a personal break, so he was not there to
9:36 adjust the uh the positioning. Clearly
9:40 hair and makeup did their job cuz look
9:43 at this, you know, and and I mean,
9:46 thanks to the set decorators that make
9:48 this look like a shitty home office. Do
9:50 you know how hard it was to get some of
9:52 these non-escript useless
9:54 props? You think you can just find them
9:56 at a thrift store? The the answer is
10:01 yes. That's actually a fake background.
10:08 Um, so why Sam Alman couldn't
10:13 sleep is because they launched a new
10:15 feature today. And I actually think it's
10:17 a major [ __ ] feature. Now, let's see
10:20 something here. It is possible I've
10:23 still got juice left on this iPad. No, I
10:26 don't. Let me charge
10:30 it because we're going to do some stuff
10:32 with voice.
10:35 But what OpenAI launched today. Yeah,
10:38 exactly. Mr. It's got it. Memory fake
10:42 hairdo. Yeah. Yeah. This is This takes I
10:46 I'm in the makeup chair for about four
10:48 hours to get the hair looking like this.
10:51 I don't
10:52 remember.
10:56 Um, let's see. Heck, some of the
10:58 background items haven't even been taken
10:59 out of the packaging yet. I took the the
11:02 rabbit out of the packaging. I used that
11:04 for a while. And the car car haven't
11:07 taken out, but that's a collector's
11:08 item. You know what I took out of the
11:10 packaging is the [ __ ] Bllandons that
11:13 I took out of the
11:15 packaging. I to I took it all out of the
11:18 packaging.
11:20 Um, okay.
11:22 So, chat GPT now has access to your
11:28 history, like all the things you've ever
11:30 chatted
11:32 about, and it's pretty good. I haven't I
11:35 haven't tested it a ton, so I figured
11:37 we'd do a little bit of testing. But,
11:40 so, if you don't know, let me make this
11:42 a little bigger so you can see it. Let
11:44 me let me get Sharon here. Okay. So, you
11:47 can see you can see all this stuff,
11:49 right? Okay. So, if I go into my
11:53 settings and I go under
11:58 personalization. Yeah. Okay. They added
12:00 a new thing here. So, reference saved
12:04 memories and I can go in and manage the
12:06 memory. So, this they've had for a
12:07 while. So, this is that thing where it
12:10 just sort of randomly stores [ __ ]
12:12 right? And you can go and you can prune
12:15 this if you want, but if you have ADD,
12:17 you're not going to do that. Tik Tok pin
12:20 left. There you go. Sorry about that.
12:23 Sorry about that. Carrying on.
12:27 Um, but this new thing, reference chat
12:31 history, let chat GPT reference all
12:33 previous conversations when responding
12:36 turned
12:37 on. So, this is crazy. So, I'm going to
12:39 say
12:42 um maybe I'll do it with voice. Let me
12:45 do it with voice here. Tik Tok pin. I
12:48 see fam. Sam finally had enough of Elon.
12:51 Oh, that he's suing him. Yeah. Yeah.
12:54 They've been they've been going back and
12:56 forth at each other for [ __ ] years
12:58 now. It's a bit of a snore,
13:02 honestly. It's like, you know, just
13:05 start dating,
13:08 guys. It's like whatever.
13:14 Just can can we just leave the country
13:17 in a neutral place maybe? All right.
13:21 Um, okay. So, I'm gonna put this into
13:24 voice mode. I think you'll be able to
13:26 hear
13:29 this. Hey, Quinn. How's it going?
13:33 Hey, Kyle. I'm good. How about you? I'm
13:36 doing good. So, I'm I'm on with my
13:38 TikTok channel. As you know, we're on
13:40 here a lot. you you help you know
13:42 facilitate some of these conversations
13:45 and I understand that you've been given
13:47 access to all of my previous chats. Is
13:50 that true?
13:53 Yep. I've got access to my chat history
13:55 so I can keep track of what we've talked
13:56 about and build on. It makes it easier
13:58 to help you out with whatever you're
14:00 working on. What's on your mind today?
14:02 Well, over the over the years, you know,
14:06 the irregulars are the people who show
14:07 up to my Tik Tok channel all the time,
14:10 and for for a while there, we we we came
14:13 up with an irregular zoo, and there were
14:15 a bunch of different animals in it. Can
14:16 you look back through all my chats and
14:19 tell me all of the different animals
14:21 we've referenced uh with regard to my
14:24 live a AI learning lab live or the
14:26 irregulars?
14:31 Sure. Sure. We've got a whole managerie
14:32 to look at. There is definitely a beagle
14:34 since your dog sings along to your
14:36 guitar. Then for the Shellington world,
14:38 all the characters are animals, though
14:40 we haven't named them specifically. If
14:42 there are others, I'll need to dig in
14:44 more. Anything specific you're curious
14:46 about? Yeah, dig in more because there
14:48 are some that are named after
14:50 um browser comp or browsers that are out
14:54 or AI models that are out. There's
14:56 something that's a reference to a video
14:58 that has a zombie kit in it.
15:04 That's searching the web.
15:10 I couldn't find anything specific
15:11 online. Do you remember any other clues
15:13 or context? Not online. Not online.
15:16 Thinking in my history.
15:20 Got it. Let's see. We've got a few
15:21 notable mentions. There's the beagle for
15:24 sure. Then for Shellington, we decided
15:26 on animals instead of human kids. Oh,
15:29 and there was mention of a turtle for
15:31 the Irregular Turtle Society. Any other
15:33 specific ones you're thinking of? Go
15:35 back way earlier. Go back like a year
15:37 and a half ago.
15:39 All right, let me check further back.
15:41 One
15:49 sec. What'd you find?
15:53 All right. from way back. We had a fox,
15:56 a raven, and a cat mentioned. Also, a
15:59 penguin. Wait, Quinn, you're just making
16:00 [ __ ] up now.
16:03 Sorry about that. Let me recheck.
16:10 All right, let's get serious and focus.
16:12 I'll comb through the history and get
16:14 you the accurate list of animals. Give
16:15 me a moment.
16:18 Now, I don't know how much of this we
16:20 actually talked about in the channel in
16:22 the in the actual chat, so it's possible
16:25 she doesn't know. Um, hey
16:28 Quinn, what
16:31 are what are some things that I've
16:34 demoed over the years? Um, where I was
16:38 demoing things live to my uh to the Tik
16:42 Tok live.
16:45 You've demoed quite a range of things.
16:47 AI tools of course like generative AI
16:50 models and apps. You've also done some
16:52 live songwriting, shared your
16:54 experiences with building custom GPTs,
16:57 and explored philosophical topics around
16:59 AI. And let's not forget your musical
17:02 moments with your guitar. What about the
17:05 the custom GPTs? Can you tell me which
17:08 which custom GPTs we've built over the
17:10 years?
17:14 Sure. A few stand out. There was one for
17:17 AI readiness, another related to ADHD
17:20 brainstorming, and one for the content
17:22 evolution portfolio. And you've made a
17:24 ton for nonprofits during GPT for good.
17:27 Any particular one you're thinking
17:28 about? Well, the very first one I made,
17:31 I talk about it a lot.
17:37 Ah, the first one that was your custom
17:39 GPT to help nonprofits, right? You built
17:42 it during GPT for good.
17:45 It's quite the opposite. My first GPT
17:48 was called Sin City Sipper, and it's
17:49 about making drinks.
17:55 Got it. Sin City Sipper. Now, that's a
17:57 fun one. A custom GPT for mixing drinks,
18:00 right? Yep. Def. I think she's making
18:03 all of it up. Well, no. Some of the some
18:05 of the stuff she said about GPT for good
18:07 was good. Um, okay. Let's let's let's
18:11 dig in. Why don't you tell me
18:13 something about me that I've revealed
18:17 through all of our interactions that I
18:19 might not have
18:21 um realized I like? Why don't you uh
18:24 infer something about me that I maybe
18:26 didn't say explicitly, but it's in in
18:29 the conversations we've had.
18:32 All right, let's see. Openness to
18:35 exploring new ideas and how much you
18:37 value creativity. From AI to
18:39 storytelling, you dive deep into things
18:41 that spark your curiosity. It's like
18:43 you're always in the pursuit of blending
18:45 technology with human experiences. Yeah,
18:47 but that's that that's that's obvious
18:49 and on the surface. Dig deeper. What's
18:51 something that I reveal about myself,
18:54 even though I might not have said it
18:56 explicitly?
19:00 All right, going deeper. There's a
19:02 strong sense of empathy in your approach
19:04 to tech. Whether it's AI for good,
19:07 focusing on inclusivity, or the way you
19:09 engage with your community, it seems
19:11 like you genuinely care about making a
19:13 positive impact. Does that strike a
19:15 chord?
19:16 I was just lying.
19:20 No worries. Even in Jest, it's clear
19:22 you're passionate about what you do. If
19:24 there's something
19:27 All right, Quinn. Anyway, all right.
19:29 We're gonna move on. I got a bunch of
19:31 [ __ ] to show tonight. Thank you, Quinn.
19:35 anytime, Kyle. Okay. So, and if if
19:40 you've not played with advanced voice,
19:42 um all of that conversation is now
19:44 stored in the chat. So, one of my
19:47 favorite things to do with advanced
19:48 voice is put it on the car Bluetooth and
19:50 just brainstorm on your way to work and
19:53 by the time you get to work, you have
19:54 ideas fully fleshed out. It's pretty
19:56 cool.
19:57 Um, so this memory where it's got access
20:01 to all of your past chat history. Um,
20:04 there's there's a couple of things about
20:08 it. Nate Nate Silver did a video on this
20:11 and I was thinking the same thing and
20:12 when I saw his I was like, "Ah, yeah,
20:14 that's it." Um, chat GPT up to this
20:17 point has been very transactional,
20:19 right? You log into a chat, you chat
20:22 with it, you got to give it context, and
20:25 then it gives you an answer, and then
20:27 the next time you log in, you kind of
20:29 have to start over. What this is
20:31 allowing is chat GPT to shift into more
20:35 of a companion
20:36 mode where it knows all the [ __ ] we've
20:40 talked about for the past two and a half
20:43 years, right? Everything that we've done
20:45 in a in a GPT is in here and it's got
20:49 instant access to it now. Now, as we
20:52 discovered tonight, it it it's a little
20:54 janky. It
20:55 hallucinates. Um, but it's still a
20:57 pretty big deal. And it's the first time
21:00 that that we've had anything resembling
21:02 long-term memory. The other thing I saw
21:04 I saw a a tweet on
21:07 this that they were like why Sam
21:10 couldn't sleep why he was so excited is
21:13 this becomes a moat Tik Tok pin Joe now
21:18 she's not going to forget how mean you
21:20 were her well it's funny Joe we we for
21:23 for a while now when I sort of get
21:25 snarky with the GPT people would joke
21:28 you better be careful she's going to
21:29 remember that well now she now she has
21:32 access to my snarky
21:33 It's funny. I asked her explicitly about
21:36 this uh earlier this afternoon and she
21:38 was like, "You've been very nice to me.
21:39 You didn't say anything bad." So, I
21:42 think I'm okay with Quinn. Although, she
21:43 might have been hiding her
21:47 rage.
21:49 Um, yeah, but this is a big big big
21:53 deal.
21:55 Um, oh, the the reference to a moat. So
22:00 about a year year and a half ago, there
22:03 was a an employee of Google that left
22:06 the company and and upon leaving wrote
22:09 this piece called there is no
22:11 moat. And what what he basically said
22:14 was um everyone's going to copy
22:17 everyone. Open source is going to eat
22:20 the the you know frontier model
22:22 companies. They're just going to catch
22:23 up to them incredibly quickly. And
22:25 that's that's proving the case.
22:29 Um, if you work for a company that uses
22:33 Salesforce, you know how bad Salesforce
22:36 is. It's horrible.
22:39 It's staggeringly bad and no one
22:43 switches from it. Why? Because it's got
22:46 all your [ __ ] in it. Like the the
22:48 switching cost. Oh, there's migration
22:50 tools and there's this and that. Yeah,
22:52 but like that's the sales of your
22:54 organization is sitting in that thing.
22:57 truth from Brandon. You know, it's it's
23:01 absolutely hardware and and but the
23:04 switching cost is so high that no one's
23:06 going to leave it. That just happened
23:08 with Chat GPT. If you've been using chat
23:12 GPT for the past however
23:15 long and it's got all your things in
23:17 there,
23:21 um why would you why would and and you
23:24 can instantly access them. That's a
23:27 reason not to leave chat GPT. So um so
23:31 anyway, yeah, this is this memory starts
23:34 to be a really really really interesting
23:36 thing. Now a couple of things I don't
23:38 know. I don't know if uploaded documents
23:41 are part of that memory as well.
23:45 Um I don't know if there's any kind of
23:48 rag like retrieval augmented generation
23:51 where it can pull specific quotes of
23:52 things that I said.
23:58 Um let me try
24:03 something. Tik Tok pin. I'm guessing
24:07 they index your conversations with rag
24:09 and then every query you
24:11 do they
24:13 search I guess I I guess and I I don't
24:18 know and there's also like the fact that
24:20 that the the toggle is buried
24:24 inside settings means most people are
24:26 not going to even know it's
24:28 there. Like look how complicated their
24:31 their settings have gotten.
24:36 builder
24:39 profile
24:41 speech like you have to come in here and
24:44 turn it off. Did they ever get my custom
24:47 instructions back?
24:49 Nope. Lost forever,
24:56 [ __ ] Oh, man. Alive. All right.
25:04 Oh, what was I going to ask Quinn? Let
25:05 me ask Quinn
25:09 something. Quinn,
25:13 um, are there any new characters that
25:16 I've added to the Sydney musical
25:18 recently?
25:21 You've been pretty focused on Kellen and
25:23 Sydney lately. If you've added any new
25:25 characters recently, I might not have
25:27 caught those. Got any in mind? Now,
25:30 what's the what what are the main
25:32 um differences between the early
25:35 versions of Sydney and what we're
25:36 working on right now?
25:42 The early versions of Sydney were more
25:43 focused on the novelty of an AI falling
25:46 in love with Sydney being more overtly
25:48 aware of her digital nature. Now you're
25:51 emphasizing Sydney's growing sense of
25:53 humanity and emotional death, focusing
25:55 on her experience of loss and the
25:57 realization of a barrier between her and
25:59 Kellen.
26:01 We in in the script, do you do you have
26:04 access to the PDFs that I've uploaded
26:06 into other chats or just the chat
26:07 itself?
26:10 I only have access to the chat itself.
26:12 So I don't have any PDFs or files unless
26:15 you share them here. Need help with
26:17 something specific from the script? No,
26:19 I was just wondering if you knew there's
26:20 a line below the tagline, right? The the
26:22 the the um title is Sydney and then the
26:26 tagline is an artificial love story and
26:29 then we added something right after that
26:31 in one of the scripts. And I'm wondering
26:33 if you know what that is.
26:36 I don't have the exact line, but you
26:38 might be referring to a tagline or a
26:40 subtitle. If you share what it relates
26:42 to, I can try to help rephrase or recall
26:45 it. Yeah, that's okay. Um,
26:51 all right. No worries. If anything else
26:53 comes up, I'm here. Do you can you tell
26:57 me the
26:58 difference in lyrics from the first
27:02 version of To See the World and the
27:06 latest version?
27:18 Hello, Sydney. I mean
27:25 Quinn, maybe we don't need to worry
27:27 about chat knowing everything. Quinn,
27:29 what are you doing? Are you lost in
27:32 thought? All right, she she crapped out.
27:35 She got tired of my She goes, "I don't
27:37 want to talk about your stupid
27:42 musical." All right.
27:46 Um, good lord. What What do we want to
27:48 talk about next?
27:50 Um, one, two, three, four, five. There's
27:54 still something
27:55 missing. The
27:59 hell? See the raccoons and irregulars.
28:01 All right, let me change my my sharing.
28:06 [Music]
28:15 Uh
28:19 delete. Oh, nice. One of my characters
28:22 for my new Tik Tok game. Like it. Scroll
28:26 down for the paint post. I will do that.
28:28 Mr. Nice.
28:31 Are you are you generating these in uh
28:34 in chat
28:36 GPT? Oh, that was midjourney. That one's
28:40 midjourney. Oops. Black bar. I know,
28:43 Brandon. I see it coming. Chat GBT slow
28:48 tonight. That's pretty
28:50 perceptive. This is also another step
28:53 forward to everyone having their own
28:54 companion AI. Yeah, exactly. It'll be
28:57 fine-tuned to them and people can have
29:00 it create things for them and do things
29:01 for them. Yeah, exactly.
29:05 Like I can
29:07 imagine choosing to use chat GPT or or
29:11 choosing not to use chat chat GPT if I
29:13 don't want something in the
29:17 history. But
29:20 like if I were going to write something
29:22 for my musical in Chat GPT or Claude
29:25 because it's kind of a toss up which
29:27 one's better. I'll probably lean to Chat
29:30 GPT more now because I want all I want
29:32 that all in the in the history. You
29:35 know, mine was slow a couple of nights
29:37 ago. Got excited. Something new must be
29:41 dropping. I say, "Are we doing Wait, are
29:44 we doing this? Aren't you talking to me?
29:47 Are you mad at me for just giving me the
29:49 silent treatment? Yeah, I think she just
29:51 crapped out. Um, all right. What else do
29:54 we have
29:56 here?
29:59 Nice. Badges.
30:02 Raccoons. The new internal whisk model.
30:05 Yeah, this looked pretty interesting.
30:06 Oh, this this was another one that I
30:08 thought might be worth playing with
30:10 tonight.
30:15 So P
30:16 uploaded picture of a
30:19 tractor, picture of a
30:23 dog. So these were added as subjects. I
30:27 then added a description, an adilic UK
30:31 farm.
30:36 Then in different styles,
30:38 mosaic
30:48 sticker tractor running over the
30:51 dog, but the dog rebuffed
30:53 [Laughter]
30:58 him. Oh, that's cute.
31:02 That's really
31:05 cute. What's that? Oh, that's the end of
31:07 that
31:14 thing. Yeah, that one's
31:17 weird. Physics on all these things are
31:19 always a bit wonky, wonky, wonky. So,
31:23 where do you find Whisk? You find it at
31:28 um labs.google.com.
31:31 google.com
31:35 labs.google.com and then you go to
31:37 experiments you go to
31:40 create and then it's here do I have
31:42 access to no I'm still weight list on VO
31:46 but I do have access to VO some other
31:51 way sign in with Google
32:05 style. Make it an enamel pin. Make it a
32:08 sticker. Can I do my
32:17 own? No. Add your own image.
32:47 open in
32:49 tool. That's kind of cool. All right.
32:54 style and then how do you animate it?
32:59 You go.
33:12 I don't know how he animated those. Load
33:15 a
33:18 template. All right. I don't quite know
33:20 how to use this tool. He was using what?
33:23 Oh, he was using an internal model that
33:25 had that.
33:27 Ah, got it. Got it. Got it. All right.
33:31 Fantastic. Fantastic, Bob. All right.
33:34 Let's get Let's get rolling here. Let's
33:36 Let's play with some other
33:38 [ __ ] Okay. Let's go to
33:43 Gemini.
33:45 And so if you go to I I mean, honest to
33:49 God, as insane as OpenAI is, they've got
33:54 all these different models. At least
33:55 it's at the same domain. I'm going to
33:58 show you. Wait, so I just showed you
34:01 labs.google.com. Now I'm going to show
34:03 you
34:04 gemini.google.com. Then I'm going to
34:06 show you a
34:07 studio.google.com. And then I'm going to
34:09 show you Vertex AI, which is also a
34:12 Google product.
34:13 [Laughter]
34:22 I mean, Google's Google's philosophy is
34:25 if you're not smart enough if you're not
34:27 smart enough to find our [ __ ] you're
34:30 not smart enough to use
34:33 it. It's like their hiring
34:39 strategy. Oh man. Okay. So, the thing I
34:44 wanted to show you, so Gemini about
34:48 three weeks or so ago came out with
34:51 um two Gemini 2.5
34:54 Pro, which which people are saying is
34:57 really good. I haven't used it a ton.
34:58 They're saying it's good for coding.
35:00 It's good for stuff. It's a It tends to
35:03 be a little cold Google
35:06 stuff from a writing perspective, but
35:09 you know, if you're doing code, that's
35:10 fine.
35:13 But they they now
35:16 incorporated deep research with
35:19 2.5 and I ran something today and it was
35:22 really quite good.
35:23 So let's do some research here.
35:27 Um the Easter egg theory survival of the
35:30 fittest strategy. Exactly. Hiding in
35:33 plain sight. Yeah. Exactly. Well, it's
35:37 relatively trivial. you just what what
35:39 what you would do is uh one of one of
35:42 the strategies I employ is I I'll create
35:44 a a Google Sheets uh some what I what
35:47 I've actually done is I've incorporated
35:48 a Google form where I log into the
35:50 Google form and whenever uh the company
35:52 that I work for which is Google uh
35:55 whenever they um uh uh create a new uh
35:59 domain or a subdomain uh I I'll go to
36:02 the Google form and I'll type it in and
36:04 then I'll put of course a description uh
36:06 and and what sub tools are there and
36:08 what subdomains are listed there. And
36:10 then those of course all get written to
36:11 a a Google sheet which is acceptable
36:15 accessible by uh three people on my team
36:17 and and and then of course uh my wife
36:19 but we can't really talk about that.
36:21 She's not supposed to have access to
36:23 those URLs because some of them are
36:24 propriet anyway. Uh but what I would
36:27 recommend to you just because it's
36:28 relatively trivial and easy is just uh
36:31 create some sort of cataloging system
36:33 where every time you see an announcement
36:35 from Google uh discern the the URL that
36:38 may or may not be active and and whether
36:40 or not the tool that was released was
36:42 actually released or it was just an
36:44 announcement or or it was actually just
36:45 a dream of someone that coded something
36:47 over the weekend but we're never going
36:49 to release that that you'll have to
36:51 discern on your own. Once you discern
36:52 it, then you you put it into the form
36:54 that goes into the Google sheet. You
36:56 don't have to use Google sheet. You
36:57 could use something like Air Table or or
36:59 or notion or some some sort of
37:02 structured data uh you know a trivial
37:04 small simple structured data set to keep
37:07 track of the
37:09 URLs. Or you could just make it [ __ ]
37:17 easy. All right.
37:21 What do you want to research? Who's got
37:23 something you want to
37:25 research? What should I
37:28 research? How many [ __ ] subdomains
37:30 there are within
37:32 [Laughter]
37:34 Google? Let's do something good. What's
37:36 something good? Something good and
37:38 interesting. Anyone? Anyone? Buler.
37:42 Buler. See if it can make an image of a
37:44 clock. Oh, Gemini. Wait. Well,
37:47 let's Gemini 2.5 Pro. Okay. So, let's
37:50 say image of a clock
37:55 uh that reads
37:59 um
38:01 8
38:04 45 Oh, cool.
38:07 Yeah.
38:08 Here. Thank you,
38:12 baby. Formulating a response. Refine the
38:15 response. You're asking for an image. I
38:18 can't display images. Which one can do
38:26 images? Oh, do I have to switch this
38:29 images and another one? Seriously,
38:32 people do like why do we have to [ __ ]
38:37 figure this [ __ ] out?
38:43 No, this is not The only thing that's
38:45 special to my individual
38:48 account is personalization. All right. I
38:52 don't know. I don't [ __ ] know. Oh.
38:55 Oh, I think I have to do it inside
38:57 Studio. That's what it is. Yeah, that's
38:59 that's the other thing. Not only do you
39:01 have to remember which models to use,
39:02 you have to remember that that 2.5 Pro
39:06 in Gemini is different than 2.5 Pro in
39:10 AI
39:12 Studio. How are you supposed to know
39:14 that? Well, it's relatively trivial.
39:16 What what you do is uh I've got a a
39:18 checklist, a set of criteria when a new
39:21 model comes out. uh I I'm I'm uh
39:23 exploring uh whether whether the the
39:26 model is uh multimodal uh whether it's
39:28 monomodal whe whether it's reasoning uh
39:31 if it if it can create images if if that
39:33 for some reason has been curtailed uh at
39:36 the server level uh and and you know I I
39:38 run through this checklist and then I do
39:40 that across all of the models at all of
39:42 the URLs. It's it it's very very quick.
39:45 It only only usually takes me four four
39:47 and a half hours uh to work through the
39:49 checklist uh on all the different URLs
39:51 with all the different models. Uh and
39:53 and then I just remember those. It's
39:55 relatively
39:57 trivial. Thanks,
40:01 baby. Oh, that's some good
40:04 Joe. That's some good
40:09 Java. Fantastic. All right. What's that?
40:13 Jeff Flanigan on YouTube. Web trends.
40:16 What will replace the worldwide
40:24 web? Well, actually, Google showed some
40:29 [ __ ] a year and a half, 18 months ago
40:33 when they announc announced Gemini
40:35 Ultra.
40:36 They showed some [ __ ] that was pretty
40:39 [ __ ] remarkable and we haven't seen
40:41 it since, which was dynamic website
40:45 generation. Um, where you would ask it
40:47 for stuff and it would just sort of
40:48 figure out what website it needed to
40:50 build for you if it needed to build a
40:52 website and then it would build it. It
40:53 had access to a whole interface library.
40:56 Um, and then it just [ __ ]
40:58 disappeared. They haven't mentioned it
41:00 since.
41:05 Um, I think
41:07 that like like Broadway plays are still
41:11 around even though we we don't need the
41:14 theater to tell stories anymore. We we
41:16 kind of like it as a nice thing. I kind
41:19 of feel like that's where websites are
41:21 going to go. That companies are still
41:23 going to have websites, but
41:25 increasingly no one's going to go to
41:27 them.
41:30 people are just going to ask chat GPT
41:32 for [ __ ] or whichever tool wins out. Um,
41:36 and it's just going to dynamically build
41:38 websites based on based on what we want.
41:41 So, I I I think the mechanism of the
41:44 hypertext transfer protocol will
41:46 probably stay around um as as a way to
41:49 navigate between all the sites. I I
41:52 don't know, maybe there's some fancy ass
41:54 blockchain version of that. The the uh
41:57 the distributed web, right, um is a is a
42:00 pipe dream that we've been hearing about
42:02 for a long time. But I I think I think
42:04 the core infrastructure stays around. Um
42:07 I just think the the requirement of
42:11 having a website, the necessity of it
42:14 diminishes
42:16 um because everything's going to get
42:17 dynamic.
42:19 Kakef research new creatures. Oh,
42:21 recently discovered in the deep ocean.
42:23 Okay, perfect. So, here we are at at uh
42:27 Gemini and I'm flipping over to to deep
42:31 research. So, I'm going to say I would
42:34 like to
42:36 know all of the new
42:43 species of
42:47 animals discovered in the ocean
42:53 uh in 2025.
43:00 2025. In the
43:02 report, I want you to make a
43:09 scorecard
43:10 for countries like the medal
43:16 count at the Olympics.
43:22 Let's
43:27 consider a new
43:32 species, a gold
43:36 medal, a subspecies, a a
43:39 new subspecies.
43:44 Um, a silver
43:49 medal
43:52 and
43:53 [Music]
44:02 a I don't know and some
44:08 other thing be the bronze medal.
44:14 This is America. We're going to make a
44:15 scorecard. Damn it. Welcome to
44:19 America. My model bitching has has
44:22 summoned Pate to the chat. Hey, Pate.
44:24 What's happening?
44:28 I'm just bitching about all the domains
44:29 that I have to go go through at at uh at
44:33 at uh Google, but I haven't made a I
44:36 haven't made a video about how awful it
44:39 is like I did for uh for chat GPT
44:42 [Laughter]
44:46 models AI code like cursor or lovable
44:50 lovable. We're going
44:52 to animals. Yeah,
44:54 animals. Oh, and so here's here's what I
44:58 want to
45:00 show. Okay, let's see. I'd like to know.
45:03 Okay, let's see. More analyze report.
45:05 Start research. Great. Okay, now once
45:09 this thing
45:12 starts, you can you can watch what it
45:14 does. It's pretty cool.
45:21 [Music]
45:23 research focus information
45:25 gathering metalbased
45:28 scorecard new sub other significant
45:31 taxonomic findings will be counted as
45:34 bronze medal. This will allow for an
45:36 interesting comparison of marine
45:38 biodiversity. Now look at this. So
45:40 here's all the websites that it's that
45:42 it's
45:44 researching.
45:45 Um, if Science, CSRO, Global Newswire,
45:52 aaa.com.tr, 866 new marine
45:55 mammals. So, it's reading all this.
45:58 Initially, I found some interesting
46:00 information. The Ocean Census Project
46:03 announced the discovery of a large
46:04 number of marine species, over 800
46:07 identified in the past couple of years,
46:09 with the announcement being made in
46:10 March 2025.
46:14 Next steps to get a clearer picture. I
46:16 need to delve deeper. It used the word
46:19 delve. That's funny. Into the sources
46:21 I've
46:22 identified. I plan to specifically look
46:24 for those publications. So, it's looking
46:26 at that [ __ ] And I guess it's still off
46:28 doing
46:31 that. Research comes back redacted while
46:34 some someone knocks at your door.
46:36 Research where the cargo container that
46:38 washed up on the shore of Jamaica came
46:41 from full of chicken.
46:45 I don't know that story. That's pretty
46:47 funny. My relatives are trivial. Not
46:50 really for
46:54 research. All right. Confirm new
46:56 species. I've made some progress for
46:59 2025. So far, I've
47:02 confir if you've never really watched
47:04 one of
47:05 these deep research models do what it
47:08 does. If you think about just like just
47:11 this list of websites, like I can click
47:14 on any one of these and go to that
47:16 website, right? Boom. There it is. I can
47:21 go look at it. I can go read
47:24 [ __ ] And then I got to close it. And
47:26 then I got to go back and then I got to
47:27 click on the next
47:29 thing. It's
47:31 like, why did that not
47:35 go? Oh, it's downloading something. Oh,
47:38 that's a that's a
47:41 PDF. Close. You need to sign in for this
47:46 one. Like just clicking and going to the
47:49 websites would take you a while. And
47:51 this is just going and going and
47:54 going. If you've played with Manis, it
47:56 does this. But anyway, apparently the
47:58 deep research with 2.5 Pro is quite
48:00 good. So if you haven't played with it
48:02 yet, it's out there available for
48:04 everyone.
48:10 They should have called it Google
48:12 2.0. Yeah. Actually, you know, not for
48:15 nothing, this this is an interesting
48:18 model. I mean, it's a little
48:20 expensive to have every Google search
48:23 write a report for you. Like, I would
48:25 think that this is this is not cheap.
48:27 Not as cheap. Did you see any of the V2
48:30 or the new internal whisk models I
48:32 shared? Yeah. So, I I I went and I I we
48:34 looked at your uh dog and tractor stuff,
48:38 Pate. Um and then I'm going to go to
48:40 Vertex in a bit and play with VO that's
48:43 available there. So, it does look like
48:45 it's available there. So, um
48:53 yeah. Yeah. I'm I'm going to also show
48:56 um Gemini 2.5 in the
49:00 studio. now will natively analyze
49:04 videos, video
49:06 files, and and not just natively
49:09 analyze the words, right? We've had that
49:13 for a
49:15 while. And we've had within the AI
49:19 studio of Google, it's been multimodal
49:20 for a while, but they just added it to
49:22 the big new 2.5 Pro model. You can pop a
49:26 YouTube video in
49:28 there. All right.
49:30 [Music]
49:47 M I think this thing hung. It's funny.
49:50 This This hung for me the other day. Let
49:53 me try reloading
49:59 it. My next action is to finalize the
50:01 metal
50:02 count. I will prepare a scorecard
50:05 highlighting the
50:09 countries. I've identified more species
50:13 described in
50:14 2025. These include new soft coral
50:17 species from the Gulf of
50:19 Mexico discovered by researchers in the
50:21 USA and then a copapod species found in
50:25 the South China Sea.
50:27 I also found descriptions of three new
50:30 sea cucumber
50:32 species from the Kermatic Trench in the
50:35 East Indian
50:36 Ocean. You know, I've been wondering
50:40 like why is it been so
50:42 long since we've found new sea cucumber
50:46 species, right? It's like everyone's
50:49 just like, "Well, yeah, a cucumber is a
50:51 cucumber." And then they're like, "Well,
50:53 a sea cucumber is a sea cucumber." I'm
50:55 like, "No, they're not. They're not
50:59 cucumbers. They're a different kind of
51:07 animal. I think about it all the
51:11 time. When are they going to discover a
51:13 new C cucumber species? When's it going
51:15 to happen?" Ma ma
51:21 ma. I think this is wrong. Maybe it's
51:24 Maybe it's thinking, but I think I think
51:26 we may have lost. We need our report. We
51:29 need our
51:30 [Laughter]
51:34 scorecard. I'm one of the
51:37 two. What's up, irregulars? What's up,
51:39 Wolf Man Clint? What's shaking? What's
51:41 going
51:45 [Music]
51:47 down? See cucumbers in the irregulars
51:50 farm now. Yeah, there you go.
51:53 All right, we'll come back to this. Let
51:55 me let me reload and see if it's just
51:57 it's if it's just
51:58 web finalizing the scorecard. Yeah,
52:01 we're we're still this thing's stuck.
52:03 We'll come back to this later. Okay,
52:05 let's go look at some some new Gemini
52:09 um on the back end. Okay,
52:15 now output
52:17 format. Oh, this one's got image
52:19 generation.
52:22 What did someone ask me to generate? Oh,
52:23 a clock. Um, make me a picture of a
52:28 clock
52:30 with the hands uh with
52:34 uh
52:36 8 45
52:43 uh indicated by the hands.
52:53 Let me I'm gonna hop. There you go. 10
52:56 and two. Although this is funny. It's
52:59 not 10 and two. It's it's
53:01 it's a little different. It's still two,
53:03 but it's not
53:10 8:45. It
53:12 tried. Still can't do that one. This is
53:15 like how many Rs in strawberry are there
53:17 in strawberry?
53:18 [Music]
53:20 Um, oh that's nice. They've got little
53:23 popups for the different models that
53:25 show their cost and things like that.
53:27 Okay, so we're going to go to Gemini 2.5
53:30 Pro Preview 3-25. So, this is a March
53:34 25th thing and I'm going to hop over to
53:40 um AI
53:42 Salon or No, I'm going to go to
53:53 YouTube. What the world needs now is
53:57 love. Sweet love.
54:01 It's the only thing that there's just
54:04 too little
54:05 of. All right. So, here's good here's a
54:09 good one. So, this is
54:10 the the AI
54:13 ready Murphy and I did
54:15 yesterday. So, I'm going to copy this
54:18 URL. We're going to go into into Gemini
54:23 here. I'm using the 2.5 Pro preview.
54:27 Now, watch this. When I type in
54:30 a YouTube video. It shows
54:34 me Oh
54:37 [ __ ] Oh, it's over an hour. So, it's
54:39 just over a million tokens. [ __ ] Um, I
54:44 got to find a shorter
54:49 [Music]
54:50 video. Here's the Kelly Bosch Salon. How
54:54 long was that? Oh, that's two hours
54:56 and 57 minutes. That's a long ass
55:01 meeting. Uh, videos. Let me see if I can
55:03 find a short one here.
55:07 Um, this should be good. So, this was
55:09 Sabrina Romangh's learn out loud blot
55:13 thing. All right. So, we'll go back
55:16 here. Let me get rid of this
55:18 one. Pop that in
55:21 there.
55:25 Okay. So, this is a million tokens. So,
55:28 so basically an hour of video is a
55:30 million tokens and this is this has got
55:32 about a million token context. So, I'm
55:34 going to say
55:42 um I'm going to say
55:45 describe the major
55:50 um visual
55:54 segments
55:57 of this
56:01 video. And
56:09 Highlight any times
56:14 when software is being
56:26 demoed. And then when I did this
56:29 before, yeah, this just came in under
56:32 the token
56:33 limit, but it's still pretty cool. Like
56:36 it knows all the words in the video and
56:37 now it's looking at all the imagery in
56:39 the
56:49 video. What's it
56:51 doing? Can I open this?
56:57 branch from here. I should be able to
57:00 open
57:01 this. I could early. I could on the
57:04 previous one I did. It was kind of cool
57:05 watching it. It It was sort of writing
57:08 on a minute-by-minute basis everything
57:10 that was in the video. It was pretty
57:11 slick.
57:27 Hang on. I'll be right back on Tik Tok.
57:29 I want to look at something. Oh,
57:32 Higsfield AI. That was the
57:45 one tools structured output function
57:48 calling grounded with Google search code
57:55 execution. Uh did it stop? What's it
57:57 doing? No, still going.
58:06 [Music]
58:09 Can I talk about temperature?
58:10 Absolutely. Okay. So, it looks like the
58:14 temperature scale is between 0ero and
58:16 two. I don't know why OpenAI shifted
58:19 theirs from 0 to one to 0 to two. And I
58:21 don't know why they did that.
58:26 Um, temperatures basically
58:31 um, how do I how do I how do I describe
58:34 it in a way that doesn't make it sound
58:36 just
58:36 crazy? It in complete layman's terms,
58:40 the higher the temperature, the more
58:41 creative the output. That's not quite
58:44 right. if if you have a temperature of
58:47 zero. So what happens when you type a
58:49 prompt is it creates a mathematical
58:53 um
58:55 probability of some some token up in in
58:59 latent space. So the way large language
59:01 models work is you type a prompt and
59:03 then it goes and looks at the most
59:06 likely if for lack of a better term the
59:09 most likely word to be the to the to be
59:12 the next correct right word and then it
59:15 just keeps doing those probability
59:17 searches you know in the latent space if
59:19 you have the temperature higher. So,
59:21 let's
59:22 say you you type in a a a quick brown
59:26 fox jumps over a whatever dog, right? A
59:30 lazy dog. And
59:32 and the word dog is blank. If you have a
59:36 temperature of zero, it will look in the
59:38 latent space and and dog will be the one
59:41 that's got the highest probability. And
59:42 so, every single time it will give you
59:45 dog. Um, if you have temperature of of
59:48 one, like 50% up the up the scale, um,
59:52 it may go up there and it not only sees
59:55 dog, but it says there's some things
59:56 around it like cat and rabbit and turtle
1:00:02 and it may just randomly choose to give
1:00:04 you something other than dog there. And
1:00:06 then if you have it super high, you
1:00:08 know, you have a higher likelihood of
1:00:09 getting things even even beyond that. So
1:00:12 it it's one of the reasons when you're
1:00:15 in chat GPT I think chat GPT's default
1:00:18 is like a a 75% I don't know where it is
1:00:21 on the scale of 1 to2 I guess a 1 1.5.
1:00:25 So its default is fairly creative which
1:00:27 is why when you prompt something in chat
1:00:30 GPT it'll give you something different
1:00:32 every time you give it the exact same
1:00:34 prompt. But if you run that that
1:00:36 temperature down to zero you'll get the
1:00:39 same exact thing every single time. Does
1:00:41 that make sense? So, just basically more
1:00:43 or less
1:00:44 creative. Um, this thing is at 263
1:00:51 seconds. I can make a video while this
1:00:53 thing's going. Yeah, let's let's go do
1:00:55 that. Okay. Um, so now I'm going to go
1:00:58 over to Oh my
1:01:04 god. So, wait, how did we get here?
1:01:09 we got
1:01:11 here
1:01:16 cloud.google.com. Good lord, that's a
1:01:18 bad
1:01:21 URL. So, I guess if you just go search
1:01:24 for Vertex Studio,
1:01:27 um, write a write a prompt for for VO
1:01:31 video.
1:01:32 8 second length number of results for 16
1:01:35 by9 V2 is the
1:01:38 model. All right, this is good. This is
1:01:41 the first time I found something where
1:01:42 I've got access to it. So, we'll do
1:01:44 we'll do uh
1:01:47 70s muscle car resto
1:01:51 mod rolls
1:01:55 through an
1:02:00 abandoned
1:02:05 factory driven by a
1:02:10 mysterious woman.
1:02:18 All right. And here we go. Generating
1:02:21 videos. This may take several minutes.
1:02:24 Well, probably less minutes than this
1:02:27 search. Oh, here we
1:02:28 go.
1:02:32 Okay. So, I said, describe the visual
1:02:35 segments of this video and highlight any
1:02:38 times where software is being demoed.
1:02:42 0 to 14. Introduction
1:02:44 graphics. Presenter
1:02:46 introductions. Guest speaker
1:02:49 introduction. Software
1:02:52 demo. 49
1:02:55 to 429 text repurposing
1:03:00 workflow 28 to 38 Q&A and
1:03:04 discussion 38 to 43 software demo.
1:03:08 Blot video creation. So, let's go to
1:03:12 3812 in the video and see if it's where
1:03:16 it starts. Video
1:03:20 demonstration
1:03:23 38
1:03:30 12.
1:03:32 Bang. Look at that. It's actually
1:03:34 [ __ ]
1:03:35 accurate. Boom.
1:03:40 4556 software de demo blot pricing page.
1:03:44 What was
1:03:46 it?
1:03:50 4556. So we'll go
1:03:53 455 56. How much is
1:04:01 it? Boom. Oh, and this is good.
1:04:05 So, it got
1:04:13 56 is exactly when the screen came up
1:04:16 and she she wasn't talking about
1:04:18 pricing. I think that that person is
1:04:22 asking about
1:04:25 pricing. So, that's pretty slick. All
1:04:28 right. Outro graphics.
1:04:33 That's pretty that's pretty flipping
1:04:34 cool.
1:04:38 Um, let's let's try this.
1:04:44 Um, what are all of the
1:04:49 names
1:04:51 of
1:04:56 people
1:04:58 whose names
1:05:03 appeared on
1:05:06 their video thumbnail?
1:05:11 while
1:05:19 chatting. That one might take a while,
1:05:22 too. But we can go back to Vertex and
1:05:23 see if we got videos yet. Oh, we do. All
1:05:26 right. I Let's see if these things suck.
1:05:30 One of the one of the holy grails of
1:05:34 video generation is to have a car roll
1:05:38 forward where the wheels actually roll
1:05:42 and the car doesn't slide mysteriously
1:05:45 ac sideways across the
1:05:53 pavement. It's not bad.
1:05:56 It's the wheels aren't quite
1:06:02 timed. They're not quite timed with the
1:06:05 speed of the car, but it was pretty
1:06:09 close. Yeah, this one. This one it's
1:06:11 it's skating a little bit. Oh, damn
1:06:19 it. Why does it do that? We're gonna
1:06:22 We're not going to have a black bar for
1:06:24 a second. You're just going to have to
1:06:27 deal.
1:06:38 Okay. Not
1:06:41 bad. It's not not quite there, but it's
1:06:43 not
1:06:44 bad. Not bad.
1:06:49 people. That actually looks pretty
1:06:53 good from a physics standpoint. That
1:06:55 looks pretty good. I don't know what the
1:06:57 [ __ ] car that is. That car doesn't
1:07:01 exist. Um, let's see. Let's do um
1:07:15 Snapping
1:07:28 turtle. Eating a
1:07:32 skateboard
1:07:34 while the
1:07:38 skateboarder cries in the background.
1:07:49 If you haven't been here before, stupid
1:07:51 prompts are my
1:07:52 [Laughter]
1:07:56 thing. Traits tractor tractor did the
1:08:00 wheels good. Yeah, that was pretty good
1:08:02 actually. Hey, Pay, do you know when
1:08:04 when that new version of Whisk is coming
1:08:06 out and are they going to create a new
1:08:09 domain for it?
1:08:14 It's going to be like
1:08:17 labs.wisk
1:08:20 [Laughter]
1:08:26 6.2.4. Holy guacamole. This memory thing
1:08:29 for chat GPT is so good. I know. It's
1:08:31 cool,
1:08:32 right? How's your search doing? I don't
1:08:34 know. Let's go see. Did it do it? Yes.
1:08:39 Based on the video, here are the name
1:08:41 names that appeared in the participant
1:08:43 video thumbnails. Vicky Baptist, Sabrina
1:08:47 Romanov, it says Simona Palace. I think
1:08:50 it's that's not quite right.
1:08:52 Serena, Michelle Muny, Silva, Gwyn.
1:08:55 That's really cool.
1:08:59 These names were visible primarily when
1:09:01 the Google Meet grid was shown around 49
1:09:04 seconds, 11 seconds, 25 seconds or 25
1:09:08 minutes, and 43
1:09:11 minutes. That's pretty
1:09:17 wild. Let's
1:09:19 see. Was
1:09:22 anyone
1:09:24 wearing blue?
1:09:27 And what are the
1:09:31 timestamps? If
1:09:34 so, this is this is really stupid stuff,
1:09:37 but I'll tell you where. So, this is
1:09:39 something right
1:09:41 now for Story Vine, for my company.
1:09:45 We've got a tool that does um audio
1:09:50 transcription that that allows our
1:09:52 clients. So, we've put them we've put
1:09:55 all of our transcripts into a vector
1:09:57 database
1:09:59 um so that our clients can search for
1:10:03 video content based on what people are
1:10:05 saying in the video, not just the
1:10:07 metadata, which is super slick. Um, but
1:10:11 wouldn't it be nice if you could say,
1:10:13 you know, across all my videos, find me
1:10:16 someone, you know, who's sitting in in
1:10:18 their garage, and when is that? It would
1:10:20 go find the specific clip. That's
1:10:22 pretty That's pretty cool. All right.
1:10:24 Uh-oh. Google told me you have reached
1:10:26 the limit of projects to which you can
1:10:28 enable
1:10:29 billing. You can request an increase in
1:10:32 your billing quota. What does that even
1:10:34 mean?
1:10:43 H Danielle, you don't have the new
1:10:45 memory feature. Are you a plus member or
1:10:47 are you on chat GPT free? It's not
1:10:49 available for free users right
1:10:52 now. All right, let's see if Vertex did
1:10:56 any good
1:10:56 [Music]
1:10:58 here. Is his brain on the outside of his
1:11:04 head? Oh, no. It's just pink
1:11:11 hair. That one's really
1:11:15 good. The snapping turtle got his
1:11:18 skateboard. He's
1:11:19 [Laughter]
1:11:27 crying. Anyone else here ever caught
1:11:30 snapping turtles? I
1:11:32 have. You know how you do it?
1:11:36 Well, you go down. Okay, listen. If
1:11:41 you I didn't think I would have to
1:11:44 explain this because I think it's fairly
1:11:46 obvious how you catch a snapping
1:11:49 turtle. Oh god. Okay, so you go to your
1:11:52 sporting goods store and you get
1:11:54 yourself some shark hooks, right? Three,
1:11:58 four, something like that. Decent size
1:12:01 shark hooks, right?
1:12:03 Then you go to the the hardware store
1:12:06 and you get yourself some string, some
1:12:08 good
1:12:09 string. Then you go to your backyard
1:12:12 where you keep the
1:12:13 empty plastic gallon containers from
1:12:17 from the milk you throw out there,
1:12:18 right? You don't throw them out. You
1:12:19 just throw them in the backyard, right?
1:12:21 Like you do. So, you go get some plastic
1:12:24 jugs, you get your shark hooks, you get
1:12:27 your string, and you need a
1:12:29 rowboat or Johnbo or you know, and then
1:12:34 as you do, you go to the horse rendering
1:12:37 plant and you say, "Hey, I need to buy
1:12:40 like a five gallallon bucket of horse
1:12:42 meat." And they're like, "Yeah,
1:12:44 whatever." And then they give you a
1:12:46 bucket of of bloody horse
1:12:48 meat. And if you think I'm
1:12:53 joking. And then so what you do is you
1:12:57 go down to the pond. Pretty much any
1:12:59 pond will have them in
1:13:03 there. And then you put the horse meat,
1:13:05 a big chunk of horse meat on the big
1:13:07 shark hook. And then you throw it out
1:13:09 with the with with the milk
1:13:11 jug. And you put like, I don't know,
1:13:13 four or five or 10 of those out in the
1:13:15 pond. And then you go home.
1:13:19 And then you come back the next day and
1:13:22 all your milk jugs are going to be all
1:13:24 over the place because they'll either
1:13:26 have catfish on them or turtles on
1:13:32 them. And then here's the coolest part.
1:13:34 You got to be careful with the snappers
1:13:36 because they'll take your [ __ ]
1:13:37 fingers off. So you you pull them up and
1:13:39 you can feel them thrashing around
1:13:41 there. You pull them up and then you you
1:13:43 swing them so their tail comes up and
1:13:45 you grab the back of their tail. And so
1:13:48 you have the the string in one hand with
1:13:50 their head and and and their belly up
1:13:52 and then you got that and then you tap
1:13:54 their belly and when you tap their belly
1:13:56 they fall to sleep and then you put them
1:13:57 in the burlap bag. Oh, you need burlap
1:14:00 bags, too. All right, there you go.
1:14:02 You're
1:14:06 welcome. And old Mrs. Nelson screaming
1:14:09 about her horse.
1:14:19 Where did I type in wearing? Oh, did I
1:14:22 type in wearing with an H? Oh my
1:14:25 god, that's pretty
1:14:28 bad. Oh, good lord. I think I think that
1:14:32 first Oh, this one doesn't have the guy
1:14:34 in the background. This is a loser
1:14:37 one. These don't really look like
1:14:39 snapping turtles either. That one's
1:14:41 decently close.
1:14:43 This one's really good. Can I download
1:14:45 this? How do
1:14:47 I save? Save video frame. It's not
1:14:51 letting me save it. Download. There we
1:14:53 go. All
1:14:56 right. Wait. Let's go get the
1:14:59 prompt. Oh, wait. Add a soundtrack. Use
1:15:04 this video to inspire new music. Add a
1:15:06 voice over. Provide narration track for
1:15:09 this video. Oh, this is cool.
1:15:15 Write your prompt. Synthesize speech
1:15:18 with different
1:15:23 voices. That's a little
1:15:26 complicated. I don't give a [ __ ] No,
1:15:28 we're fine. We're just going to go We're
1:15:31 going to go post this on the internets.
1:15:43 Um, has a
1:15:45 snapping turtle
1:15:51 ever
1:15:53 snatched your skateboard?
1:16:02 It's a sad
1:16:05 [Laughter]
1:16:20 day. All right. Um Oh, am I supposed to
1:16:25 be Are we uploading? Are we going to
1:16:27 upload or we going to crash? Come on,
1:16:29 Twitter, you can do
1:16:33 it. Or can
1:16:35 you? Rabbit.te
1:16:41 breaking.
1:16:43 New intern
1:16:45 mode. No interest right now. What's it
1:16:49 do? It'll Oh, it'll email
1:16:51 [ __ ] This may be my least favorite
1:16:54 story.
1:17:02 What? The the the fishing for snapping
1:17:04 turtles. How do you think restaurants
1:17:07 get their their snappers for snapper
1:17:09 snapping turtle
1:17:12 soup? Little redneck boys like
1:17:17 me. Oh man.
1:17:20 [Music]
1:17:26 details at your email. You get nine
1:17:28 tasks a day. Huh. All right. I'll check
1:17:32 that out
1:17:32 tomorrow. So, two things I didn't get to
1:17:35 tonight. I don't want to dig into them.
1:17:36 Two things I didn't get to were
1:17:40 um Higsfield AI, which is a yet another
1:17:44 [ __ ] video tool that's got really
1:17:47 interesting looking camera controls. I
1:17:49 don't know if I have any credits there.
1:17:51 Um, but we'll go look at that. And then
1:17:53 the other thing if you want to play with
1:17:55 it is go to
1:17:59 um
1:18:06 studio.firebase.google.com and it's just
1:18:08 like lovable and it's like replet agent
1:18:12 but it's from
1:18:13 Google. Um, and it's really fast and it
1:18:18 looks pretty [ __ ] good. Um, so I
1:18:21 don't know if you're still here, Pate,
1:18:22 but I've said like three nice things
1:18:24 about Google. One nasty thing about all
1:18:27 the URLs and three nice things for me,
1:18:31 that's like a [ __ ] world record. So
1:18:33 Google's Google's dropping some decent
1:18:38 [ __ ] Okay, Vicki, you can make two
1:18:41 videos for free on Higsfield. Okay,
1:18:42 cool. All right, so listen, let me let
1:18:45 me start winding this up. So, tomorrow
1:18:50 um tomorrow at 11:00 a.m. Mountain time.
1:18:53 How many We got some folks in here.
1:18:55 Tomorrow at 11:00 a.m. Mountain time,
1:18:56 I've got AI office hours. So, if you go
1:18:59 to my LinkedIn, Kyle Shannon,
1:19:03 um I do uh AI office hours from 11 to 12
1:19:08 on Fridays. It's every Friday and we
1:19:10 just hang out and talk about what's
1:19:12 going on with AI. There's there's a lot
1:19:13 going on. Uh, so I think tomorrow is
1:19:15 going to be interesting. Um, so so come
1:19:18 to that and then tomorrow night's Friday
1:19:19 night date night. We'll do uh we'll do
1:19:22 another one of these here chat add
1:19:23 things. We'll get to some of the other
1:19:25 things I didn't get to. And hell, maybe
1:19:28 some new [ __ ] will drop. Maybe we'll
1:19:29 have the new 03 model from Chat GPT.
1:19:32 Maybe some third company will come out
1:19:35 of Paraguay and beat all of the models
1:19:38 with their model and just and the stock
1:19:40 market will explode again. I don't we
1:19:43 don't it doesn't nothing matters
1:19:46 anymore. It's just all everything's
1:19:49 changing. So So just [ __ ] hang on for
1:19:52 dear life and show up. All
1:19:54 [Laughter]
1:19:59 right. Okay. Fantastic. Bob. Bob. Tell
1:20:03 them what they've
1:20:07 won. Chat needs the Google URLs.
1:20:11 Oh, the Google URLs. This Google URL.
1:20:15 Hang
1:20:17 on. This one that's like
1:20:20 lovable is this one.
1:20:24 Bang. And then here, let me put it up so
1:20:27 the Tik Tok people can see it. Except
1:20:30 because the screen's always blow out
1:20:31 blown out, you can't read
1:20:34 anything. There you can. There you go.
1:20:37 Boom. Look at that. Nice. And then
1:20:42 the Gemini's Gemini. Oh, wait. We got
1:20:45 our Oh, hang on. Wait, people. We got
1:20:49 our We got our species research back.
1:20:58 Okay, hang on. So, I'll give you So,
1:21:00 this one's gemini.google.com, but you're
1:21:02 going to choose deep
1:21:04 research. Okay, I can export the docs.
1:21:07 Very nice. Ocean Discoveries 2025, a
1:21:09 catalog and global
1:21:11 scorecard. Introduction, unveiling ocean
1:21:14 life in 2025, the uncharted frontier.
1:21:16 All right, gives us a little bit of
1:21:18 background. Accelerating the pace in
1:21:20 initiatives and
1:21:24 technologies. Wait, what's this? The
1:21:27 Ocean Census Project, a collaboration
1:21:29 launched in April 2023 by the Nippon
1:21:31 Foundation and the
1:21:33 UKbased charity Nectton with the
1:21:36 ambitious goal of discovering a 100,000
1:21:40 new machine species within a
1:21:42 decade. Ocean Census provides
1:21:44 significant concentration of re
1:21:46 resources with novel operational
1:21:49 model. It's a big old network of
1:21:51 scientists. Okay,
1:21:53 cool. Purpose of this
1:21:58 Okay. Distinctive feature of this report
1:22:01 is the presentation of an Olympic style
1:22:03 scorecard tallying discoveries
1:22:05 attributed to different countries based
1:22:08 on the affiliations of the primary
1:22:10 researchers involved. The system assigns
1:22:13 gold medals for each new species, silver
1:22:16 medals for each subspecies, bronze
1:22:18 medals for for each new higher tax on
1:22:21 genus or family. That should probably be
1:22:23 the gold medal, but whatever. All right.
1:22:25 The scope is strictly li limited to
1:22:28 marine animal discoveries. It excludes
1:22:31 plants, fungi, microbes, or any
1:22:33 non-marine
1:22:36 organisms.
1:22:38 Navigating notable
1:22:43 highlights. The year 2025 yielded a
1:22:45 wealth of new discoveries. All right.
1:22:49 Okay. The ocean census initiative. It
1:22:53 discovered the guitar shark.
1:22:56 What? Oh, we gotta see what that is. The
1:23:00 [ __ ] the guitar
1:23:01 shark? You ever heard of the guitar
1:23:06 shark? Guitar shark. This better look
1:23:10 like a [ __ ]
1:23:11 guitar if it was named after Oh, it
1:23:15 does. Look at that thing.
1:23:22 Oh my god, that's beautiful.
1:23:30 [Music]
1:23:36 That's one of the most gorgeous sharks
1:23:38 I've ever
1:23:42 seen. Nice music, too.
1:23:57 Nice. Wait, that says the enigmatic
1:24:00 wedge fish. Maybe that's not the guitar
1:24:05 shark. Oh, also known as guitar fish.
1:24:08 Shovelnown shovel-nose
1:24:11 rays. Guitar shark. Those are cool.
1:24:14 Okay. All right. Sorry. Okay, back to
1:24:16 our
1:24:19 research. The venomous
1:24:22 gastropod found between 200 and 500
1:24:25 meters
1:24:28 deep.
1:24:30 Maldives octo
1:24:40 coral. The pygmy pipe
1:24:43 horse. All right.
1:24:46 Peer
1:24:48 literature. It's pretty
1:24:51 cool. Emerging
1:24:53 patterns, comprehensive
1:24:58 catalog. There's a nice table. I mean,
1:25:01 this is quite
1:25:03 good. This is quite good. Comprehensive
1:25:06 table of everything discovered in
1:25:11 2025. Bunch of sea stars.
1:25:17 Deep sea
1:25:19 gastropods. Apollo's rainbow sardine. We
1:25:22 gota we got to search that
1:25:25 one.
1:25:27 Apollo.
1:25:30 Apollo. What was it?
1:25:33 Rainbow.
1:25:35 Sardine. New
1:25:40 species. Where's the picture?
1:25:43 We need
1:25:45 pictures. Dang
1:25:49 it. I'm exhausted by
1:25:52 hyperlinks. [ __ ]
1:25:58 it. Look at all these things that they
1:26:01 found. Okay. The metal count. A global
1:26:05 scorecard. Here we go, people. USA.
1:26:09 Yes. Yes. USA number one.
1:26:18 [Laughter]
1:26:20 We're winners. We're
1:26:22 winners. 13 13 gold medals. China five,
1:26:27 Australia four, Brazil four, Germany
1:26:29 four, Japan one, Singapore, Argentina.
1:26:32 Bunch of bunch of them with one
1:26:43 I mean, that's really quite
1:26:45 impressive. And then there's all the
1:26:47 there's all the sites it
1:26:48 searched. Sources read but not used in
1:26:51 the
1:26:52 report.
1:26:54 So amazing thoughts. And here was the
1:26:57 thinking that it did. So all of that you
1:27:00 can export to a doc, a Google
1:27:02 doc. Let's see how this looks. Hopefully
1:27:06 it's well formatted.
1:27:16 Oh yeah, they have an image tab on
1:27:17 Google. Shut
1:27:20 up. Yeah, that's not very impressive.
1:27:23 That old rainbow, the rainbow sardines.
1:27:29 Nah. All
1:27:33 right. Look at this. A well-written
1:27:35 report with a table. table needs to be
1:27:42 reformatted. There's our metal
1:27:45 counts, taxonomic trends, geographic
1:27:49 hotspots, deep sea globally. Yeah, cuz
1:27:52 we've never been down
1:27:58 there. Wow. Very
1:28:00 cool. Very [ __ ] cool, people.
1:28:04 All right, you heard it here first. And
1:28:07 you saw that snapping turtle eat that
1:28:09 boy's skateboard. Look how sad he
1:28:12 is. I hope you all feel good about
1:28:14 yourselves. You made that skateboarder
1:28:17 [Laughter]
1:28:21 cry. It sounds like a generic store
1:28:23 brand fish. It does. The Apollo rainbow.
1:28:26 Rainbow
1:28:27 [Laughter]
1:28:31 sardine. Hey, but listen. We discovered
1:28:33 the guitar shark
1:28:35 tonight. All right. Irregular channel
1:28:37 before I go. Okay. One last little
1:28:40 snippet
1:28:42 [Music]
1:28:44 here. Irregulars. A learning lab and
1:28:50 irregulars. Kyle's exmachina moment. Who
1:28:53 are you calling a [ __ ] idiot? Kyle.
1:28:55 Quinn. I didn't mean it like that.
1:28:57 Quinn.
1:29:03 That's pretty good. I like it. That's
1:29:06 solid. That's some solid comedy right
1:29:13 there. All right. So, we're gonna wait.
1:29:16 I'll give you credit if I can here. Joe
1:29:20 Dealt. All right. Or
1:29:23 dissult. Do probably do is my guess.
1:29:29 Um,
1:29:32 paste created. Wait, let's
1:29:36 see. Um, dear at
1:29:41 Sama, I
1:29:44 guess I guess I should
1:29:48 have have been
1:29:52 nicer to my AI.
1:29:57 um uh
1:30:00 companion these past two
1:30:04 years. A
1:30:11 [Laughter]
1:30:15 um
1:30:16 thanks
1:30:18 Joe Dalt. Is that what it
1:30:24 was? No. No. D E C E A U L T Check the
1:30:33 sticky. Oh. Oh, cool. At
1:30:37 Joe
1:30:41 D Au.
1:30:46 beautiful for the the um scarily
1:30:54 accurate
1:30:57 image. How do you spell scarily? There
1:31:00 you go. One L. Everybody knew
1:31:03 that. And when when you talk about
1:31:05 someone wearing
1:31:07 clothes, it's w h e i n g. All right.
1:31:14 That's that's how we do it in in AI. If
1:31:18 if you were doing it in traditional
1:31:20 spelling, you're probably thinking like,
1:31:22 shouldn't it be we a r? Yeah. In in the
1:31:25 before time when you just wrote, that's
1:31:28 how you would spell it. But in the AI
1:31:30 time, wearing always has the H in it and
1:31:34 you remove the A. If you add the H, you
1:31:36 remove the A. That's just one of the
1:31:37 rules. It's in shrunken whites. You'll
1:31:40 see it's in the third reference
1:31:41 appendix.
1:31:43 All
1:31:44 right. You're
1:31:47 welcome. Embrace the chaos. Snapping
1:31:50 turtle soup. I'm hungry
1:31:55 now. All right. All right, everybody.
1:31:57 Peace out. Um, come to AI uh AI office
1:32:02 hours tomorrow at 11 a.m. Mountain on
1:32:04 LinkedIn. Otherwise, I will see you for
1:32:06 Friday night, date night. Peace out.