SPEAKER_00: The action item's obvious. If I were to say right now, action item, each week, Vinny shows up with more than just his opinions, but he brings two or three actual demos of AI products, that would put it as an action item. Yes, Vinny?

SPEAKER_02: That's correct, Jason.

SPEAKER_04: Okay. This Week in Startups is brought to you by Masterclass. Learn from the world's best minds

SPEAKER_05: anytime, anywhere, and at your own pace. Get 15% off an annual membership to Masterclass at masterclass.com slash startups. Merge. Let your developers get back to their core product. Merge is a single API to add hundreds of integrations to your app. Integrate up to three customers for free today at merge.dev slash twist and lemon.io. Need to speed up your product development without draining your budget? Hire vetted engineers from Europe startup at lemon.io. Go to lemon.io slash twist to get 15% off for the first four weeks.

SPEAKER_00: All right, everybody. Welcome to this Week in Startups. It's our AI roundtable with Sonny Madra and Vinny Lingam. If we look a little different right now, it's because we're trying

SPEAKER_08: out Vinny's super cool new product, weightroom.com. Tell us, uh, welcome back to the show, Sonny and Vinny. Good to be back.

SPEAKER_10: Good to be back, Jason. Thanks. Thanks for joining me on weight room.

SPEAKER_08: Yeah. So it's weightroom.com. This is your product. Um, I see three video windows looks similar. Uh, it's in a browser. It looks similar to zoom with some slight differences. Tell us,

SPEAKER_13: uh, what we're experiencing here and how AI works into all this.

SPEAKER_15: So if you just click on the right hand side, you obviously see people, but go to the summer, summer AI, summer, summer, summer, summarize, exactly. Yeah. Summer AI. Summer. Yeah. And so you can then see all the, all the stuff, which like we've been talking about before we started the recording, um, action item, uh, Michael to investigate audio and video stems and signs to Michael. Michael's my co-founder is in the room here as well. It's listening and it gives insights there that you have. You can scroll up and see all the different stuff we've spoke about. And it's really cool because not like we're going to have a slack integration button here. So you will push these

SPEAKER_27: things to slack and you can push. This is crazy. Look at this. It says catch up the team engaged in a

SPEAKER_00: casual conversation about video quality, beauty filters, touch up my parents and recording methods during a virtual meeting. That's an accurate description of my first five minutes on here. So if somebody was coming in, they can catch up really quickly, but also this insight here, the action items obvious. If I were to say right now, action item each week, Vinny shows up with more information, which isn't just his opinions, but he brings two or three actual demos of AI products that would put it as an action item. Yes, Vinny. That's correct. Jason. Okay. What's the lag between me telling you an action item and actually getting done and then also some AI summarizing it.

SPEAKER_15: It's two questions. So why don't we get my, uh, my co-founder Michael to chime in quickly and this is the charming feature. So he gets 45 seconds to tell us. Oh, okay. So it takes about two and a

SPEAKER_34: half minutes. Obviously you can't summarize every second cause that doesn't make any sense, but it's

SPEAKER_37: approximately two and a half minutes of conversation, uh, that we summarize every, in chunks as they go along. We also, the stories also get better over time because we keep feeding that context as the conversation's happening to the, to the LLM. And so it's more accurate and just, you know, improve as it

SPEAKER_40: goes along. What's more difficult, Michael? Um, you know, the new state of AI and working with these

SPEAKER_27: language models or dealing with an insane dictator, like Vinny.

SPEAKER_45: Oh, sorry. You're out of time. You're out of time. That's the end of your 45 seconds. Sorry, Michael.

SPEAKER_00: Uh, that's actually pretty cool though. So I'm looking at it right now while we were having that chime in feature. Um, so this is kind of cool for a live audience. Uh, I could have people, I could allow people 45 seconds and I can hover over that end key and just end them anytime I want.

SPEAKER_52: Um, and you get multiple people chiming in as well and you can have, yeah, we, we've had hundreds of

SPEAKER_15: thousand people in the room. So it's like, kind of like Twitter spaces, but with video. So you could do a video like the three of us and then people, you know, you can add more co-hosts or you can just

SPEAKER_52: have people chime in for a few seconds.

SPEAKER_00: Sure. Great. Um, what I like about it though, is while we were talking, it did an insight. And again, it's like, it's like a DVR. It's like every two and a half minutes, it chunks the transcript and

SPEAKER_60: tries to make sense of it. And here it says insight weight room.com can help teams catch up faster with

SPEAKER_63: meeting summaries and action items, et cetera, except, except it's a transcribed weight room

SPEAKER_15: incorrectly, uh, W E I G H T versus W I T room. Now, um, the, the, the cool thing is that now on the back end, we're going to train that. And so it'll now recognize that when you refer to weight limits, actually means weight room. So, so it's going to get smarter over time. And when we move

SPEAKER_65: from, you know, like GPT four to 4.5 or five, as we look, my action item came in. Um, so let me

SPEAKER_69: do another action item here. Action item. Allow each some AI item, whether it's a catch up action item or insight to have a threaded comment thread and port all of those or pipe all of those into a Google sheet, notion page or slack room. So we have them for all time. That's an action item.

SPEAKER_72: Action item, Michael. Um, so you, you, you know, I, I think this is great for a dictatorship.

SPEAKER_00: This kind of software really shows the power of, uh, dictators. No, Michael cannot chime in. No.

SPEAKER_77: Okay. I'm not Michael. You had your shot. That's enough. I want another 45 seconds.

SPEAKER_80: So that's the beauty of this. So anyways, I think the purpose of this demo, we always just show you

SPEAKER_15: like what we're doing and you, and so we're, we're actually focusing right now is on, on VCs funny enough. So because VCs have a very strong power in the industry right now, if you're a startup and you get a link from a VC, you'd want to pitch them on, um, you know, on your startup, you don't care whether it's a zoom link, um, a Microsoft meetings link or a wait room link, you'll use whatever they give you to join the call. So that's the first thing. Secondly, is VCs tend to be a little lazier. They don't want to be updating their Salesforce every single time they, they, they chat to a startup. We will have a Salesforce integration where every time you have a meeting, it updates when you spoke to them and then also put in, uh, you know, information, you know, insights from the meeting in there, into the notes, et cetera. And the reason this is interesting is that we are building custom prompts around our LLM and we'll have an LLM in future, probably, but for now custom prompts on, on, on, on GPT, uh, on open AI and those custom prompts understand the VCs business intricately. So as the, as the startup is, uh, pitching, giving things, giving ideas, it'll be able to alert you to potential conflicts with other startups in your portfolio. It'll be able to tell you, uh, you're in, maybe you should connect this guy with that, with, with another founder in the portfolio. So we're going to make it very customized. And the summer AI tab right now is common between all three of us, but you can have a private, you can have a private tab as well, where it gives you private insights. Like, Hey, Sunny's company is in conflict with this other portfolio company. Be careful not to tell them about this project that the company is working on.

SPEAKER_83: Be careful with that. Be careful with that feature.

SPEAKER_15: Well, well, no, so we have to train it, right? So once the model's well-trained, and it's going to give insights to the, to the VC or ever the host and private insights.

SPEAKER_28: So that that's, I think that's powerful.

SPEAKER_00: That's a really interesting idea. Okay. So now you've been hearing all this and you're looking at it. Uh, what are your thoughts on this 1.0 of real time meeting transcriptions, action items? I know some other people like otter, and there's a couple of other people

SPEAKER_89: who've had some, you know, uh, touched on some of these ideas, but this is pretty much the best execution I've ever seen.

SPEAKER_92: Yeah. I mean, I think in your action items just showed up in the right here. Um, look, I really liked this, you know, it's almost, there's like two aspects. One, there's the real time capture of information. That's like, you know, generally been a problem. And I think, you know, Vinny has it done a really nice, like integrated natively in the product. A lot of those tools, you have to add them as like extensions or like a new person into a zoom meeting, which obviously, you know, changes the dynamic of the meetings. I really liked that. I think the other thing that's cool here is the feature that Vinny just talked about is like the back channel, right? So whenever you're running a meeting, you're always have some kind of back channel running, either it's, you know, within a private chat or a WhatsApp or I message. And so to, and most of that is just for context of like, Oh, Hey, like, let's talk about this next or something. And so I think like combining that could be really, really powerful. Um, you know, these actions are coming in really

SPEAKER_60: great. Like, I think it's, it's, you know, it's in the other items, which was pipe, uh, AI items, like catch up action items, insights into Google sheets, notion, and of course, lack, which was the

SPEAKER_00: exact. So in order for these kinds of technologies to work, you need to have some percentage confidence, I think, right? So any, like, yeah, I don't mind if weight room is spelled wrong, you know, that people could hear that wrong, like a human hears that wrong all the time. I'd say calm, and they think it's Palm, you know, you know, I might say Apple, they might think it's Snapple. It doesn't really matter if it gets a word wrong once in a while, but if it gets the, the gist of it, I can edit it and, and work on it. But when you have a great product like this, or a great

SPEAKER_69: starting point, Vinny, I always tell founders this, um, you can tell when something's really got some potential, something does have a lot of potential if users when using it immediately come up with five or six ideas. Now that doesn't mean that it's the perfect roadmap. It could be custom software for them. But at least it means sunny that people are inspired. So what do you think you add to this

SPEAKER_60: next? What are the features you add next to weight room? If you were on the board, or you're an investor, I don't know if you are an investor. What are you telling Vinny? Hey, dummy, you missed this,

SPEAKER_92: this and this? Well, look, I think the the opportunity here is when zoom took off, you know, you know, during the pandemic, and at the start, it was really low friction, what happened is, as it's, you know, kind of gone enterprise, it has lost some of that low friction stuff. But now, there was reasons for people were showing up in random zooms and all that. So I'd say, go back to low friction, Vinny, keep the cost super low. I really like this idea of shortening the meetings. And so take your you're already kind of doing that. And so and then integrating with the other tools, I think that's the biggest, biggest, you know, feature here is integrating

SPEAKER_60: into other, you know, workflows that we have to connect the API to if this then that and Zapier,

SPEAKER_108: and you've got up. Yeah, we need Zapier very soon as well. Tons of tons of information.

SPEAKER_110: And by the way, it's free. So like, we're not even charging for it right now. It's free,

SPEAKER_15: because you know, we want the models to learn and train. And so the more conversations we have, the better we're getting at doing this. So

SPEAKER_08: the good news for building services like this, I remember there was a Sequoia company

SPEAKER_69: that made video conferencing as a service. And I think all like AWS and Google, they all provide some sort of video relay as a service now. Yeah. Yep. Yep. Yep. Which one, which ones have you tried? Or which one do you use? Any thoughts on the back end here? And how effective?

SPEAKER_118: I haven't used I haven't used any of those.

SPEAKER_120: Yeah. So did you? You didn't have to write the video back end though, right?

SPEAKER_73: No, no, no, no, no. We use Agora. So we use Agora on our back end. But we, you know, we like

SPEAKER_15: the number of companies doing it, and Agora has been doing it for quite a while and their public company and we've been working and they, you know, we kind of use them, I guess, when Clubhouse was blowing up, and they were using Agora. And we said, Okay, well, we should try as well, because they got some crazy amount of scale. And by the way, Clubhouse went the same VC route as well. They went to the VCs first, both the community and then it just blew up. And we started using Agora. And the product has improved dramatically over the past two years. I mean, we, you know, Agorah's product. Yeah. Agorah's product. Yeah, they just keep adding features. Yeah. Yeah. And the founder came from Webex. And the other founder, co-founders, like, I think it was a co-founder of Zoom as well. So, you know, these guys understand their business and we're very, like, we're very happy to, you know, use their platform. And it's going well, I mean, look through the quality we, so we're running this stream at 640 by 360. We can jack it up to whatever, right? So it's pretty, it's pretty easy for us to go up. Because while it's free right now, this is live in production, by the way. So it's just free. So we're running low res, because why should we pay, you know, the high risk, we're not charging people, we're still in product development phase. But as the product improves, and we get, I think we're gonna, we can go up to 4k. So, you know, so we, that could be an upgrade package, right? So you pay for weight room, the AI is built in, you want 4k, you pay a little bit more, you want integrations into certain, you know, into Salesforce, there's a fee for that. So there's ways to make money.

SPEAKER_126: The great news is, you could also have a, a third tab here, which is transcript,

SPEAKER_128: and you can annotate the transcript. So you could have another one, just the, you know, running transcript, and you can annotate it. Highlights out. At the end of the meeting,

SPEAKER_15: at the end of the meeting, you're going to get a meeting minutes of the entire meeting. So they get emailed to you, and then you can forward it around, or you can send it to people

SPEAKER_27: in the company. So we do already, even here, when you have an insight or a catch up or an action item, it tells you the timestamp of it. And so that is super helpful. If you wanted to jump to the timestamp,

SPEAKER_135: you could,

SPEAKER_15: And it's a range. And it's a range. Like the last one you came through, 10.36 to 10.38. So for the past two minutes, we've been speaking about Agora's experience at Clubhouse and Founder's background.

SPEAKER_28: Like, so it tells you the time, like, it's not a single timestamp. It's the range of when the conversation happens.

SPEAKER_140: Yeah, it's great.

SPEAKER_141: One of the reasons why criticism can feel obnoxiously aggressive is that sometimes people use criticism to sort of dominate or assert superiority. And that is not helpful criticism. So state your intention to be helpful.

SPEAKER_144: That was Radical Candor, author and friend of this podcast, Kim Scott. She just did her masterclass

SPEAKER_146: session called Tackle the Hard Conversations with Radical Candor. A great concept, a great philosophy of running a business. And if you're a business leader, hey, you're going to learn so much on masterclass. There are amazing lessons from Bob Iger on leadership, Chris Voss on negotiations, Alexis Ohanian, another friend of the pod on startup investing, and so much more. These are the legends in their specific verticals in their crafts and masterclass will teach you what they know. And you're going to learn on your own time paying for an unlimited masterclass subscription. It's a total no brainer. Listen, how many times in this life do you get to hear from this level of virtuoso? Very rare, very rare that you get to. And that was just one, you know, insight from Kim in 15 seconds. Imagine how much you're going to learn in 10 minutes, maybe in two hours. Why don't you set yourself up with a masterclass subscription? Here's your call to action. Super simple, get unlimited access to every class. And right now as a twist listener, you're gonna get 15% off when you go to masterclass.com slash startups. That's masterclass.com slash startups

SPEAKER_150: for 15% off an annual membership masterclass.com slash startups.

SPEAKER_00: And I guess you could pick more or less. So that would be kind of cool too, is if I could have three settings, like more details, less details, because you know, and then also, how does it know who everybody is?

SPEAKER_69: Yeah, exactly, exactly. So that's kind of interesting. For our producer, producer Nick, you know, when you have guests on the show, or people are speaking, you get four different quick times from them, and you can kind of line them up. But when you send a raw zoom meeting, it doesn't know who's who, right, you have to train it on who's who, if using a service like Temi or any other script or whatever, you have to like, assign who is who, but here you don't, you know, 100% who is who, because they dialed into the meeting. So that's kind of awesome as well, you could put the percentage that people spoke, people who didn't speak, you could do what, what is the one for salespeople gong. And there's a bunch of competitors to gong, you could put like people, you know, people who ask questions, people who are inquisitive, people who are definitive, you can kind of, you know, add those kind of characteristics to it. So when I use Grammarly, Grammarly has a feature where it will tell you, I don't know if you use Grammarly, Sonny? Yeah, I do. Do you either of you pay for Grammarly? I do. I pay for it too. This is very interesting. And I pay for all my teams to use it. And when they don't have it on, like on a video conference, I go nuts, because I'm just like, look at these spelling errors, look at these grammar errors. Like, I'm a writer, like, I have errors, like, if you're going fast, and you're typing, but it has AI built into it now to make things longer, shorter, etc. But it also just has tone detection. So it's like, this is definitive.

SPEAKER_73: Well, that's the thing. So with AI, you're gonna have so in the future, within three to five years,

SPEAKER_15: we're gonna have the ability to read Sonny's face, your face, my face, look for like, expression points. See, oh, Jason's angry right now. Sonny's happy, etc.

SPEAKER_24: And my bullshit face. I think you pretty cool stuff on especially large groups. But like,

SPEAKER_15: you know, exacerbated what you were saying about the summaries. So Michael's actually working on daily digest right now. So you'll be able to get a data of all your meetings, daily stats, summaries. And oh, the other cool thing, Jason, you could say something in this meeting, like, hey, Nick, ping me next week, or even say, Nick, I'll give you I'll send this doc to you by Sunday. And on Saturday, you'll get an email from the system saying, Hey, you promised Nick in this meeting. Reminders. Reminders. So because it's intelligently understanding what the conversation is about, we're gonna make it we're gonna make it way more integrated into your life.

SPEAKER_173: It's interesting if you combine this, Sonny, with something like Slack, or Asana, or Notion,

SPEAKER_69: or Coda, where people are doing their task management, etc. You could have these conversations, know the project we're working on, and have the OKRs, or whatever, and have the context of

SPEAKER_92: that. Yeah. Yeah, I think what's interesting here, Vinny, like, you know, when you've been evolving this product, and it's weight room, you know, from its first incarnate incarnation, but it really is interesting now that you could think about this as a AI native, um, with meeting tool, rather than sort of the weight, rather than the weight room concept. And so that's super powerful, because it's doing all the work is what we see these AIs doing for us. And so I don't know, maybe you have to even consider the name. We have we have we've had some conversations

SPEAKER_183: about that. That's definitely what's great weight room is great. Jason says it's nice.

SPEAKER_185: So I think I think it's a domain name, it'll add all this other features. It sticks to the sort of

SPEAKER_15: roots we have, we can consider it. We do have like nation notion integration as well. You'll be able to ask summer like, hey, summer, what are the sales for the past three months in the company. And if it's in notion, it'll look it up and give it to you right there. So while you're having a conversation, having a meeting, you can actually without interrupting, like you interrupting the flow, opening up a new tab and whatever else be able to pull the information out.

SPEAKER_188: Assistant who's doing work for you. It's like having Jarvis as an eventors meeting, you know,

SPEAKER_189: I tried getting Jarvis.com. I was thinking about that.

SPEAKER_00: All right, enough. This has been great. There's your 20 minute commercial for weight room, but there's a great job on the AI features. And let's go back to zoom and finish the episode. All right, Sonny, you've been great at getting a bunch of really interesting demos every week here

SPEAKER_40: for our AI roundtable on this week in startups over the past week. What did you play with? What did you

SPEAKER_92: find most interesting? Yeah, so I'm gonna touch on some things that we saw in different places. I think a couple of weeks ago. So I'm going to just pull up a video here and we tried the Google

SPEAKER_198: music generator. So now we have, which was called music LLM. Now we have music gen from Facebook and meta. And so very similar. Interestingly, you know, there is someone has taken it and

SPEAKER_199: themselves and hosted it at hugging face. Cause you know, we've talked about what hugging face is doing for the AI community as well, which is great. And so very similar, um, to the, the Google one we

SPEAKER_198: looked at to Facebook one and, uh, you know, it's hosted here. So it takes a little bit longer to generate. So I pre-generated, but you can see the examples down below here, where if you want an 80s

SPEAKER_199: driving pop song with drums in the background or, uh, I like that 90s. Yeah. You want to try this one? Yeah, sure. We're going to have to wait a second here. This does take some time, but, um, so this

SPEAKER_198: particular, um, service is generating, um, again, music samples, I think really, really powerful for

SPEAKER_199: creators or, you know, folks that need intro outro music. Um, maybe for this pod, we can generate one of these now. Cause, uh, we can do something slightly different than, um, uh, what, what's already there in, uh, the, uh, the default this we can start up J Cal. Okay. Let me ask you a question about this

SPEAKER_27: processing speed that's going on here. If you had, um, if you were doing this on your local computer

SPEAKER_00: and you had some great GPUs, would it go faster or is this like using a lot of cloud?

SPEAKER_198: No, this, because it's a free service is not using a lot of infrastructure. So you're probably getting like shared GPU services. So if you had a really great GPU locally, um, you would, you would definitely

SPEAKER_199: get like a faster return here. It's only generating about 12 seconds as well. So,

SPEAKER_207: yeah, let's hear it. Yeah. Feels like a Miami vice episode, Miami vice, a two part series this

SPEAKER_208: Sunday night on NBC.

SPEAKER_209: So I was going to talk in the episode. Okay. Okay.

SPEAKER_00: It's pretty great. I have to say that. Like if it's doing this today, it's going to be doing so

SPEAKER_135: much more. I was, you know, when we went to see Kygo, uh, me, me and Sonny went to see, is it Kygo? Is that how you pronounce his name? I think so. Yeah. Yeah. I decided I would like to be a DJ, uh, and do a DJ set. Um, so I'm going to, I want to do like a, a J Cal AI DJ set, but I need to, I'm going to need help. I want a completely AI generated original DJ set.

SPEAKER_27: So I'll put that on the list of things that, uh, somebody can help me with to get themselves a general

SPEAKER_215: admission ticket. I need, I need a producer or somebody to prompt engineer. I don't know what

SPEAKER_203: I need. Exactly. Um, I think you need a prompt engineer, a producer that is also good at prompt

SPEAKER_27: engineering. I mean, in the future, you'll have like prompt engineers and then, uh, yeah, dropping

SPEAKER_40: some crimes vocals from the AI there and, uh, yeah, just get a producer, get blood pop to polish it off.

SPEAKER_221: So done and done. Yeah. But, uh, so, you know, look what, why is this interesting? We're just

SPEAKER_92: seeing now and like, well, let's maybe pause this one for a second, but you know, Facebook's work there in and around the open source is great because they're making these models available for everyone to use and pick them up working obviously very closely with hugging face. And so I think it's great, great for the ecosystem. What's going to be great about this, you know,

SPEAKER_00: Vinny is if you look at the bottom 50% of music usage, um, it doesn't require, you know, all that much, um, quality. It doesn't require that much uniqueness. If we were doing something where we needed a little bit of music for a short film or commercial background for an ad read in a podcast,

SPEAKER_69: you know, this will let you, instead of using stock music for your YouTube video, create something unique for you. But the truth is those people would never have hired an actual musician in the world to do this. They would never have spent a thousand dollars on it. They would have bought stock music for 10 bucks. And so here, instead of buying stock music for 10 bucks, you can do it for free, and make something truly unique. And that's just something that dawned on me, Vinny, is that we're going to be able to make unique stuff that's better than clip art and royalty free.

SPEAKER_15: Yeah. So I'm actually investing in a, a music startup, um, that I can't take a little more stealthy, but they're basically, their, their thesis is that, um, music gets democratized. Like I can't play an instrument, right? So I'm, I'm terrible. I can't, I can't play music. I love listening to music. I can't play it. And so they're trying to make it simpler for people to create this music and using AI as part of the whole thing. I, and, and I think, you know, I think we're going to see a lot more of those types of startups in the market where you're basically enabling the mass market to be more participatory in the music industry, creating music. Now, obviously you're going to, you know, it's, it's one of the things where music itself to learn like how to read music and sheet music and, and how to play instruments is, it takes years and lots of like, uh, you know, muscle memory, et cetera. Now I think there's, there's definitely an opportunity for like what rock band did, right? Rock band crushed it. When they came out, they, um, you know, kids were playing rock band all time on their playstations and drumming and whatever else. Uh, and I think you're going to see more of that. I think you're going to see more startups leveraging AI to, to basically make it really, really easy to create your own music. And, and this is a good example of it already.

SPEAKER_00: Well, you can see where this is going to go, Vinny, uh, Sonny, very simply, uh, there'll be one of these for drums, one of these for guitars, one of these for vocals, one of these for piano.

SPEAKER_69: And, you know, you'll be able to specialize in a particular instrument, you know, and go deep on that, right? Like, cause you just think about guitar. I saw you had like a guitar riff. You could start really, uh, you know, an AI could go very deep just in guitar, guitar, whether it's classical or rock or heavy metal, you know, uh, folk music, et cetera. Nobody's done that yet, right?

SPEAKER_241: Yeah. No, I mean, I think, and that's, that's a really good insight, Jekyll, in terms of like,

SPEAKER_92: where we're probably going to go overall with these models is we'll start with really large ones that are

SPEAKER_199: generic and then we'll work our way into like, you know, I guess like almost application specific. And so a really great one for guitars and a really great one for synthesizers. And so, yeah, I, I, and then people will specialize them in the way they do with real instruments, right? They'll understand how to work with them to get them, you know, like as a, as a musician can understand how to work with a guitar to get it to generate, you know, incredible sounds.

SPEAKER_00: We'll see the same thing here. Yeah. And then the next step Vinnie is you say, here's an artist

SPEAKER_69: and here's their catalog. Um, here's their views and comments on YouTube. You know, you see, you take an artist, you know, that has an incredible library, say Bob Dylan, just feed in the entire Bob Dylan.com lyric site, every album, every bootleg concert. Uh, and then you say, come up with a new song for Bob Dylan. That combines, you know, this news story, you know, something happening in the world with, you know, and write me a protest song or give me some ideas for a protest song. I did this the other night where I was trying to come up with names for a new conference. I was brainstorming after angel someone so well, I had an idea for a new conference and I was brainstorming with chat

SPEAKER_00: GPT. And I was like, come up with a name for this new podcast and this new conference. And it has to have this word in it. And it came back with three incredible names that I would have never come up with. And this was in the first 10 minutes of brainstorming with chat GPT. So the brainstorming

SPEAKER_69: forgives, uh, your accuracy, hallucinations in brainstorming are good. You know, kind of like, that's why people smoke weed and I'm not advocating smoking weed, uh, for kids listening, but that's why some people will, you know, take a gummy or smoke some weed and then come up with brainstorming ideas.

SPEAKER_00: I kind of feel like we're, we're also going to see something in that vein where you take the body of work plus these new tools and then come up with, uh, either unbelievably uncannily accurate

SPEAKER_15: tracks or new sounds. Yeah. It's also, it's also hard to tell. I mean, I've seen people posting clips of, Hey, here's a collab between these two musicians. And then they go, ha kidding. I just created that using, you know, AI. And so like, how do you know anymore when you were like, it's going to be really, really difficult to, um, yeah, like I'm, I'm, I'm a bit concerned with identity as it applies to audio. Cause you can, you can't see, right? Like, you know, visually you can see when it is a fake bot on the screen. Like you can, yeah, that's not real, but with audio, it's so perfect and so crisp and so clear right now you can mimic people's voices. You can create songs. You can like, I, I, I, I'm over

SPEAKER_167: certain we're going to get a number one hit that's fake at some point. Totally.

SPEAKER_251: Closing big enterprise deals is never easy. The last thing you want to do is slow down your sales team because of a lack of integrations in your product. And you know, B2B buyers, they expect integrations. You're a startup founder, you're building B2B software. They expect people management tools to work seamlessly with the payroll provider, to work with your CRM, to work with the accounting software. All this stuff has to be integrated in 2023 and going forward. And if it doesn't, it's a huge problem. You're probably going to lose the customer, right? But integrations, as you know, they're brutal to implement and they take a long time and you have to maintain them. What if there was a better solution? Well, there is, you need to use merge, which makes app integrations seamless. Merge is the leading unified API that allows you to launch your integrations in days, not quarters. Merge is going to unlock new revenue opportunities and make your customers happier. Merge offers hundreds of integrations across seven important categories. HRIS, that's Human Resource Information Systems, ATS, that's Application Tracking Systems, accounting, CRM, ticketing, you know, all those marketing automation, of course, and you know, table stakes, file storage. Merge has unlimited integrations, and they charge based on how many of your customers use these integrations. You only pay for what you use, or I should say what your customers use. So they're going to give you three linked accounts for free today at merge.dev.twist. Again, you get three linked accounts for free already at merge.dev.twist. Check it out. It is a game changer.

SPEAKER_256: I have a guy on YouTube I follow. He's called Laszlo Bernie. It's only got 34,000 subscribers,

SPEAKER_00: but check this out. This first one is the Sultan Sultans of Swing by Dire Straits, and if it was written by Pink Floyd. So this guy understands the style so well that he can, you know, play the Sultans of Swing as if it was Pink Floyd, you know, Dire Straits to Pink Floyd. And then you can also do a Pink Floyd song as Dire Straits. Dire Straits. So this is Comfortably Numb, famous Pink Floyd song,

SPEAKER_260: coming at you. K-Rock 97.5 New York City, Long Island coming to you. Jones Beach 87 degrees and overcast. We might get some thunderstorms while you're rocking the beach. Stick with us, get some AA batteries for your CD player. I also, I think if I tweeted out like a couple of weeks ago,

SPEAKER_15: these guys hired this actor who looks just like Paul McCartney and, and they wrote this like parody song and he actually played the song, but like they use AI for the voice. So it sounded just like

SPEAKER_263: Paul McCartney. They use AI for the actual song, but there's an actor. So now you think what, that

SPEAKER_63: sounds like him, it looks like him. It's not, it's not an AI, it confused me because I thought it was

SPEAKER_24: an AI vote. I thought like, I thought it was Paul McCartney for real, but they made, it made it looks that it was kind of weird. It's crazy. Well, it's kind of like the deep fake, uh, Tom Cruise.

SPEAKER_268: There we go. That's the one. Oh my Lord. That is.

SPEAKER_92: Okay. Let's do another demo. All right. Sure. Okay. Um, so just continuing here, another area, which I think is, is going to get really, really exciting, you know, ties into

SPEAKER_198: a bunch of, uh, spatial stuff as we should call it now. Cause we don't use the M word anymore. Um, is, uh, you know, 3d M word being metaverse. Metaverse. Yes.

SPEAKER_279: Spatial. Now that there's actually a real product. Yes. Made by a real company.

SPEAKER_92: Awesome. There's always been real products. Um, so what you can see here and you know, if you've ever been around, uh, Luma labs, Luma labs. Okay. Here we go.

SPEAKER_288: Version 1.2 alpha. Okay. Yep, exactly. And what they have here, and I, I don't have access,

SPEAKER_198: I'm on the wait list, but they have some, you know, prebuilt, um, 3d models. And so here, I'm going to just pull up a, um, a highly detailed sculpture of a squirrel wearing a purple hoodie. And, you know, this is a proper 3d model and you can imagine this is incredibly valuable for folks that are building things in the spatial verse and or video games, uh, anything along those lines. And this, um, if you've been in around like 3d generation, this was a very, very, you know,

SPEAKER_92: kind of time consuming aspect. Yeah. And expensive. And so now, and you know, for years, the way that this was mostly handled is there was these huge libraries of open source models that people would start with and then, you know, go from there. But now the ability to kind of generate

SPEAKER_198: these from a prompt is going to be incredible for, for the industry. And this is for video games as

SPEAKER_69: well. This is for Pixar movies. This is for any 3d modeling, uh, could be games, could be casual games,

SPEAKER_00: uh, could be a movie or a TV show. So you can imagine immediately somebody taking a storyboard or sorry, somebody taking a screenplay and you could literally upload a screenplay to this product with a little bit of glue, maybe use chat GPT or barred as the general AI in this model and then tell it to use this one to make the characters, but you take it and just say, Hey, you know, make pulp fiction, uh, but with cute furry animals or miniature animals. And it just takes every character

SPEAKER_69: makes them into puppets or whatever you choose. And, uh, now you've got a storyboard and you just

SPEAKER_00: literally take scenes and say, make me the scene from inglorious bastards. Uh, and, um,

SPEAKER_69: then just give it prompts to change the characters around until, you know, Quentin Tarantino feels a certain inspiration. This is really going to be amazing for ideation from directors and screenplay

SPEAKER_92: writers. And they want, and by the way, they want to ban it. And on music, um, you know, a friend of mine who, um, is, uh, like kind of does a podcast as a rapper, you know, he, he was saying what people don't understand with these music generated, uh, AIs, like what we were talking about earlier is this is how music is pitched, right? So today, if you want to pitch something to a famous artist, you may write it and then you may lay it down on a track and then get it over. But to put it in their voice is going to be hugely beneficial for the creation industry, because that's how, um, you know, that's the, and again, like that's the general process. And I think, you know, Jacob, you and I have talked about maybe having Rob Goldberg on to help us here because he spent a lot of time in around music, but, um, that's the process, right? You come up with either the lyrics or the beat. So I write this great song, I'm Sia, I write

SPEAKER_00: Chandelier and instead of, and then I just say, Hey, here's me singing it. But then you say, take the top 10 pop artists, uh, who are trending right now and put it in their voices and send them a one minute sample to their phones. Yes. And whoever wants this song can have it for a million bucks.

SPEAKER_203: Yeah. And wow, that's powerful. Yeah. And so that, I thought that was a really interesting

SPEAKER_92: way of looking at it because, you know, already today, you know, it's already happening. I may come up with a, like a beat or I may come up with the lyrics and then I'm recording it and sending it over. And then the artist has to think about how they think about, you know, how it would sound in their voice. Now you can take it one step further. And then the artist says, Oh, great. I want to take this on. So for creators, I really thought that was a interesting way for them to look at it.

SPEAKER_310: Hmm. And that applies here. If you, you know, if you've got a great game idea before getting it funded, you could build something that's like a prototype and then, you know, go off, get your

SPEAKER_199: funding and build out the whole, you know, the whole game. All right. And, uh, as we were talking

SPEAKER_135: about before, here is the, uh, AI Steve jobs in a J Cal, uh, and, uh, shout out to, uh, Edward Brower,

SPEAKER_316: Brower. I have, uh, I have one demo I prepared for us. Great. Awesome. You'll go next. One, one demo and one other thing. It's already a demo, but I want to show you guys and tell you what I think.

SPEAKER_319: Oh my God. Thank you. Here we go. Oh, I saw this one. Yeah. You know how they think,

SPEAKER_321: but this is Apple one Oh one. It just didn't make sense to do a VR headset until now when Apple releases something. It's because the preconditions for releasing something insanely great have been assembled. Really. That is the secret formula. Apple has made mistakes and you have to try things,

SPEAKER_323: but Apple tends to get the big things right. Now. I think that brings us to Siri. Apple had an incredible lead and it seems to have vanished. Now Siri feels antiquated. Isn't Apple dropping the

SPEAKER_321: ball at this point? You know, sometimes when you are the first, that means that going back and redoing the architecture when there are big breakthroughs takes more time. I'm like,

David Friedberg: oddly compelled to keep listening, not for myself, but for Steve jobs. I look like some weird hybrid of

SPEAKER_329: me and Christopher Walken. Yeah. So Steve jobs, this reality headset. My mind go cuckoo. Yeah.

SPEAKER_332: I'm just saying Edward Norton, you look a little bit like Edward Norton.

SPEAKER_331: It's a little Ed Norton too. There's a little Ed Norton going on there. Um, but not, not too,

SPEAKER_335: not too shabby jobs. I bought Apple stock in 1979. Yeah. Wow. Um, they're better off doing it with the 3d models.

SPEAKER_00: Yes. But I mean, the, the thing I love about that version, I think that the, the copy was written by

SPEAKER_341: the individual, not by AI is that it's a, it's familiar to have me interview a founder, which is cool. Yep. But if you put the right text in his and you give him the right script, you know, and I listen, respect to the family and like, we're not making a mockery of anything here. It just does bring back this, uh, and maybe there's some like period of time before doing these things that make sense, but it does make me really miss Steve jobs being in the industry and the incredible force. He was, I don't mean to get emotional or anything, but like, he's just such a great force to have in the industry to comment on stuff and to push the industry. And, and it was like such a non-conformist and an incredible thinker that just kind of miss hearing him pontificate. I could listen to the guy for hours. I have listened to the guy for hours.

SPEAKER_150: All right. You got a great idea for your tech startup and Hey, listen, you're going to change the world. I know you're going to do it, but there's a problem. You don't have engineers, right? You need engineers. You need those developers, but it's hard to find them. And it's certainly hard to

SPEAKER_146: find them quickly. So if you're trying to reduce your burn rate and you're in a tight spot and you need to get engineering talent right now, well, what you want to do is imagine there was a partner who could provide you with more than a thousand on-demand developers. These are really good developers. They're vetted, they're experienced results oriented, and they're passionate about helping you grow your startup. And they're available at competitive rates. Does this sound too good to be true? Well, you need to head to lemon.io right now. Startups choose lemon.io because they only offer handpicked developers with three or more years experience. And they got to have strong portfolios. Only 1% of candidates who apply get in. And if something goes wrong, no problem. Lemon.io will get you a replacement as soon as possible. Many of our launch founders have worked with lemon.io and they've had great experiences. So here's your call to action to learn more. Go to lemon.io slash twist and find your perfect developer or tech team in 48 hours or less. And twist listeners get 15% off their first four weeks. So stop burning money, hire developers and hire developers smarter by visiting lemon.io slash twist.

SPEAKER_352: Vinny, your feelings? Yeah, I feel the same way. Um, I mean, I think it's a good segue to talk about the vision pro. I mean, like, this is this is the product. Um, I think that none of us have used it

SPEAKER_00: yet. Yeah. They gave select 30 minute demos to specific, uh, I don't want to say friendly journalists,

SPEAKER_240: but you know, they, they're very savvy about who they give access to these things. Uh, and he's

SPEAKER_354: brown. He did a good one on YouTube on that. I saw that awesome. So, um, so a little bit of the PR

SPEAKER_00: game that Apple plays. They'll give early access to folks who they know are super favorable to Apple. That doesn't mean they're going to get a glowing review, but I think they know how to use access to

SPEAKER_69: early products as a way to shape reviews, uh, which I guess is also, I love in PR.

SPEAKER_15: I'd like to talk about this more like from a strategic perspective, because I think a lot of people came out criticizing the price point. Okay. Like who's going to say, and I want to make it clear, like this is, this is not, this is the first version of something, which is game changing. You're going to get probably maybe even a million people buying it. And version two is going to be cheaper and better. Version three is going to be cheaper and better. And that's not Apple iterates. Apple watch version one was sucked. Version eight now is amazing, right?

SPEAKER_27: You know what? I should say it's not, that's not a based comment. It's a basic comment. Let me restate. Um, the based comment would be, um, this thing just killed

SPEAKER_240: the Oculus, whatever Rift seven. Yeah. What did you think? I think it does. I think it kills.

SPEAKER_105: No, it doesn't. Not this version. Not this version. I think, I think there was a couple of issues,

SPEAKER_92: right? If you, um, if you take a step back, you look at almost every scenario they showed it in, it's highly, you know, sort of single player mode. And I think one of the things that we've seen now

SPEAKER_198: is, um, you know, people maybe coming out of COVID or whatever it happens to be returned to the office, people want to be in multiplayer mode. And so every, you know, that, that was something that really

SPEAKER_92: jumped out at me. Um, I think another, another thing that really was, um, uh, unique in that, uh, you know, they have this mode where it shows your eyes. Um, and so they're sort of contemplating

SPEAKER_198: you wearing it and then being around people that aren't wearing it. And I, I couldn't get my head around like what that use case would be, Jacob. You're in your office. Yeah. So you're in your office and you're working and then I come up to you and you're staring at me. You're in your house,

SPEAKER_00: you're using it while cooking was like one example and you kind of have some transparency.

SPEAKER_370: Here's the ready player one, dude. Uh, you know, okay. So the reason why this just killed the

SPEAKER_256: entire metaverse and meta is done is because of apps. Uh, when you see this demo, what this is

SPEAKER_69: going to do, I don't think it, when I say it's, it just killed it. Um, the reason I think it killed it is because if you're an app developer, this is going to be 95% of your focus. You're going to deprecate everything you're doing on the Oculus, uh, or whatever they're calling it. What are they

SPEAKER_52: they call it now? The meta quest. Um, I'm not sure I, I'm not sure. I fully agree with you there. I think they're two different markets they're going after. I think that, that, that, um, Facebook

SPEAKER_115: is definitely focusing on the gaming side of things a lot more than productivity and communication.

SPEAKER_15: Exactly. Um, I think Apple is going to be, you know, it's a different market and the price point guys, it has to get down to sub a thousand dollars, of course, the mass market, but that's going to take five years. In those five years. Well, I don't think so. I don't think we get from three and a half

SPEAKER_13: thousand. I think they'll have multiple models. Remember the iPhone, they came out with that

SPEAKER_15: Essie, what was it called? I, I, I, that happens when the, basically you end of life in the certain design, you have excess capacity, you can produce stuff. That's what happened here.

SPEAKER_174: I think it's three, three years minimum, probably five most likely.

SPEAKER_380: People pay, but 1500 bucks, but for their iPhones now, a thousand to 1500 bucks. They boiled that

SPEAKER_384: frog. It's been 16 years, Jason, like at a thousand dollars, this thing's flying off the shelves.

SPEAKER_15: At a thousand bucks. What it used to be. But I don't think they're going to go from 3,500 down to a thousand in two or three years in two cycles. What do you think?

SPEAKER_385: Yes. I think they'll get the price down. I'll, I'll take that side. I think they can get the price

SPEAKER_388: down. By when, by when, when does it go? 999. Yeah. That's the price point. When it, when it gets

SPEAKER_92: under a thousand, 995. Let's just use like the iPhone curve. So the iPhone really took off in that third generation of iPhone. And so I think, yeah, three, three years, three to four years.

SPEAKER_393: Yeah. That's what I would say between three and five. Yeah. It will. Yeah. So I think we're all

SPEAKER_394: agreeing on a timeframe here. Three to five years. Yeah, exactly. Okay. So we have the time frame.

SPEAKER_15: But, but so meta now the gauntlet's been thrown down, so they have to respond. What are they going to respond with? The meta is going to be like the Android to Apple in this space, right? So it's going to be the cheaper, they're going to try and make it even cheaper, 500 bucks, 200 bucks, improve the quality, et cetera, et cetera. And they're going to compete with Apple. And when it's down to 995, the quest is going to be 495 for whatever it is similar today, which is I think 800.

SPEAKER_63: And they'll probably get it down. I don't think you're going to get, I don't think you're going to get

SPEAKER_92: everyone moving to Apple. Can I ask a different question on the pricing? So J. Cal, you know,

SPEAKER_198: in the videos that we were just showing there, which show it using it as like an infinite desktop, that's the cost of like those studio displays, right? Those Apple studio displays are they're

SPEAKER_199: about exactly right. They're like three to 5,000 bucks. Luxury product for them. Is that Mac pro

SPEAKER_256: tower and those luxury wide screens? I'm using one right now. That's exactly my point. You're you're

SPEAKER_405: how much did you pay for that insanity? $5,000? I can't remember. No, I don't know. It's at least

SPEAKER_407: five brands. No, the studio display is 2000. Okay. So just so you know, you could have bought the

SPEAKER_135: equivalent Dell monitor and it would have been bigger and just as crisp for half the price

David Friedberg: and then put a thousand dollar Dell computer on the back of it and had a Mac and a Dell.

SPEAKER_198: I think this use case that we're showing in the video here, which is like an infinite desktop, I think is the right way to think about it. I think this is an interesting one for which could really take off quickly in a trade off with the studio display. And so that's I find

SPEAKER_256: a little bit fascinating. I just also would love to have this when I'm working. Yeah, in a hotel room

SPEAKER_69: on the road in an Airbnb on a plane to be able to pull up your entire setup and to see through it if the flight attendant goes by and you want to ask him for coke zero or something, you could still do it through your headset. Um, and uh, this is the minority report future I've always wanted. So I, I love this and I hate VR, but I can see myself using this. I don't use VR. I, I, I, I, I bought a couple of these things. I don't like being, I wouldn't use this every developer on the planet is now looking at this gun. You know what? I have an app in the app store. Uh, every developer with an app in the

SPEAKER_00: app store is saying, oh, what can I, what can I, what can I, what can I charge for this? So if you're notion, if you're grammarly, if you're weight room, whatever product you are service, whatever game com.com fit bod, if you're fit bod and one of our investments, you're looking at this going, you know, I can charge 50 bucks a year for people wearing this while they work out and I can work on their technique. And when they just look around the gym, it uses visual AI to create an inventory of all your equipment. And then after it creates an inventory of your equipment, it creates you a workout based on you having, you know, dumbbells there or kettlebells or whatever. And then it directs you to the proper kettlebell of the proper weight and says, pick this one up. And it glows it red, you know, like this is your personal trainer, you know, AI personal trainer in this goggles. Now, people in the gym would sound ridiculous wearing this, but if it got you a better workout,

SPEAKER_73: it'd be worth it. So people glossed over this point that I'm going to make now in the, in the Apple announcement, they showed the integrations with zoom, Microsoft meetings, wait room potentially as

SPEAKER_15: well. They basically created this, the, the, an SDK for video conferencing platforms to plug directly into this and we're going to be the one of the first to do it. It's going to be amazing. And I think that, I think this is like, they've really thought through how do you, you know, how do you put the, how do you use these, um, vision glasses, whatever you want to call it to do remote work. To me, it looks like, as Sonny said, a single player mode product. It's like you're at home, it sits next to your bed or wherever it is, and you use it when you buy yourself. But when you go out

SPEAKER_136: and play multi, but play with your friends, or you go to the office, you don't really need it.

SPEAKER_341: Uh, you know, I gotta think like, you know, virtual white board services like Miro, you, you pop this up, you got three people in the office, you know, in Palo Alto, and then you got two people, you know, Ukraine and two people in Uruguay and one person in Australia, and you're all just going through the product roadmap and building stuff in real time. I could see a remote team all deciding to wear these 10 hours a day and build together and be on the same, what do they call those Kanban boards? Is that the right term? Kanban. Yeah. Kanban. Um, yeah, Kanban. Um, uh, you know, like a visual

SPEAKER_425: workflow. Yeah. And then you have summer AI in the background. Well, you know, and then people could

SPEAKER_341: put on a track and say, Hey, we, you, like when I worked in an office, we had a Sonos and people would put tracks on the Sonos. It was kind of a fun way to share music with each other. Uh, some people would put their noise canceling headphones on and not participate in it. It was like a fun thing to do on a Friday, but you can imagine people like walking up to the Kanban board, picking an item, a story, and then walking away with it and put it on their task. And then me being able to walk over to their cube or whatever it is and watch them working in real time writing code. And like, there's just 10 desktops in a space. And I walk around like the plant manager at a factory and I walk up to the designer's desk. I see them working. This would be a way to navigate, um, the big tension between remote workers and the management management is scared that you're working on two different jobs and they can't see you. Yeah. Workers are like, trust us. We're gonna do some side projects probably, but yeah, during the day, but you know, but seriously, trust us. Um, and then imagine you, you're, you know, during these four hour blocks, we're having two four hour blocks, one in the morning, one in the afternoon, I could walk around and see everybody working and what they're working on and watch them work. It's kind of cool. Uh,

SPEAKER_27: and you know, it doesn't have to be creep or anything. You could do the same thing. You walk over to my desk. And if you walk over, it just chimes me that you're there, you know, uh, you know,

SPEAKER_69: it's kind of like the, uh, you know, Zuckerberg or, you know, walks around the design team. Hey, what are you working on? Just a casual meet and greet. You could just set it for maybe,

SPEAKER_27: Hey, every afternoon for three hours, we're going to do a coworking session together. And you just have a virtual coworking session. What do you guys think?

SPEAKER_316: I think it's great. And I, I'm looking forward to that world. Uh, I think this is a much needed product. I'm excited by it because I, I work from home. So like I am, I am in single player mode all the time. This is, I, a lot of people do, by the way, like, I think it's a big enough market.

SPEAKER_435: All right, Vinny, you gotta go. Thanks for coming on. Congrats on, uh,

SPEAKER_436: your work on weight room.com. Go check it out. Check out the weight room. All right.

SPEAKER_439: Uh, let's rip through these, uh, send you a couple more demos for me. Yeah, I do. Um, we'll save some as well. Cause you know, uh, let me show you, uh, this next one here.

Chamath Palihapitiya: I saw you do a repli. I saw you writing some code. Yeah, yeah, yeah. I'm always doing credit for some, uh, uh, repli copilot code. No, no, no. That's all mine. Uh,

SPEAKER_92: I mean, I didn't use a copilot. I thought this was interesting. So this is, uh, using control net to, uh, take the, you know, QR codes, but make them artistic. So obviously we've seen these a lot these days because they're, you know, people use them for

SPEAKER_198: menus and, you know, I, I actually, you know, I like the integration that Apple and everyone has done where you can just open up your camera and do it. You don't need anything kind of special, but this idea of turning it into a, um, and I think on this one right here, you can pull this one up, uh, if you want, and you can just put it on your screen. If you want to pull your phone up and it'll, it'll open the link up. And so, um, the idea that you can make a QR code artistic and having the AI do it, I thought was really, really bright and just, um, you know, could open up a lot of new possibilities of embedding more information inside a scene or inside of an advertisement, um, and make it sort of native. And so all of these ones that I'm just scrolling through here,

SPEAKER_199: uh, can all trigger your iPhone. Um, or yeah, any, any, I guess can be any phone.

SPEAKER_256: Um, these are stunning. What a great idea.

SPEAKER_218: Yeah. They're really, really awesome.

SPEAKER_256: This is all done with control control control. That is a stable diffusion model. Correct? Correct. Part of stable diffusion. Um, it's, it's basically prompts for you, you give it a reference image. Yes. Say, this is what I want to start with. I believe that's the difference, right?

SPEAKER_198: Control. Yeah. Yeah. You give it your image and then your QR code, and then it comes up with this, um, which is, you know, I, I was really blown away by these. And so anyways, I, I thought this was

SPEAKER_453: really cool. Yeah. And maybe we can do some of these at the table for people who understand what

SPEAKER_27: stable diffusion is. Maybe you could explain the, when people say stable diffusion, it's a little bit of confusion because there's an open source project and a company. Maybe you could explain

SPEAKER_105: what that is architecturally on a business basis. Uh, yeah, I'll just kind of go down the basics.

SPEAKER_241: Like, so when we talk about open AI, their models are like, you know, GPT X, right? It's a one, two, three, four. That's how they name it. That's their naming convention. That's their naming convention.

SPEAKER_198: And then they have a different one for DALI, which is their image generator. So stable diffusion has a model called stability that is their image generator. Um, and so stable diffusion is the company and like, that would be open AI and they have a model called stability, stability AI. That is their,

SPEAKER_341: um, that's their model that they use. Yeah. And so you can use this open source project to make your own, you can fork it and make your own versions and compete against stable diffusion, a company or the startup companies called stability AI, I guess.

SPEAKER_241: Um, companies called stability AI and their model is called stable diffusion, right? So this is super

SPEAKER_384: confusion, confusing. So this for people who don't have the reference point here, if you listen to this

SPEAKER_279: weekend service, probably do wordpress.org. It's an open source project, a bunch of people for free

SPEAKER_341: contribute to the code. There's wordpress.com, which is a for profit multi-billion dollar company run by Matt Mullenweg that provides hosting services and services around that open source project. So you could take stable diffusion. You can make your own products and services out of it. You're well within your rights. There are some rights where you have to give back, but, um, yeah, this is a great

SPEAKER_384: use of, um, yeah, stable diffusion. Yeah. Crazy. All right. You got one more for me. Hit me with one more.

SPEAKER_92: Yeah. Yeah. So here, you know, I just wanted to bring this up because I think a lot of people are tuning in and then, you know, I think a lot of kudos here to, to replet and then, you know, a couple of projects here, you know, specifically laying chain. Um, and so what I wanted to show here is, um,

SPEAKER_198: one replet makes it quite easy to get into a development environment and, and, and build something. You don't have to configure anything on your machine anymore. Everything, everything's done on a cloud. It's like a hosted, it's a hosted IDE, uh, that also has, you know, access to a shell and everything else it needs and they make it easy to deploy. So what I did was, is I created my own, um, chat bot here that sort of has three functions. It can do searches out in the internet. It can calculate things and it can run some Python code. Uh, and you can see here, it's only about, you know, 60 lines of code for, for this. And so it's, it really is incredible. And I can deploy. And if I go here, uh, we'll just refresh this. This is, you know, I've deployed it out to here. And so we can ask

SPEAKER_466: something on what is the weather in San Francisco, uh, let's go today. And, uh, what you'll see is

SPEAKER_199: it'll just kind of go through the process and realize it needs to, um, and it's here. So you had

SPEAKER_198: a thought, do I need to use a tool? Yes. Uh, I need that's current search. And then, so it, it, it inputs

SPEAKER_308: this into whether in San Francisco today, it gets that back. Obviously I'm just showing you the innards here, but you'll, if I didn't have that. And it's using LLM chain, uh, alongside alongside

SPEAKER_198: a service that provides, um, uh, search engine results API. So SERP API is a service. And so it allows the LLM to go to the internet, to get this information, get the summary and then provide it back. So if we didn't, so there's three steps. It first figures out, it needs to use a tool.

SPEAKER_92: Then it looks for the weather. It gets that back from a search engine reply. I then says, do I need to do anything else? No, it has the answer. And then it basically returns that back. And so the, what I wanted to show here is, you know, for folks that are not, uh, and, or have not been developing for a long time, the barrier to entry is quite low to build these bots. And you know, there's lots of tools. So the tools I used in that particular code that I showed you here

SPEAKER_198: were very simplistic, uh, you know, Lang chain tool for current search or calculator in a, in a,

SPEAKER_92: in a Python runner. Um, I really want to encourage folks to, to go and try it because between replet between these open source projects, you can get going quite quickly and you can create something

SPEAKER_198: that's pretty remarkable that runs. Um, and you can, you can even have it deployed out to the

SPEAKER_240: internet on a URL, as you see here. Yeah. Great. Awesome. Uh, well done. All right. Uh, I think this is enough product. You should be inspired to go do more friends of the show. Um, and, uh, yeah, the, I'm really excited about the apple vision pro I'll buy the first one. I think the, when

SPEAKER_69: it's combined with AI, but that's gonna be one of the nice things about it. Um, is it's gonna land right as a lot of these AI development kits kind of hit, uh, some level of, you know, critical mass and or fidelity, let's say fidelity. And so it being able to look, look at your refrigerator and

SPEAKER_00: identify everything and just tell you the inventory of your refrigerator as you open and close your eyes. I really think people are underestimating how great this would be. Just literally, you're gonna

SPEAKER_341: be able to put this on, open each cabinet, look at each shelf, move stuff around. Um, and then look at the, each of the boxes and it will make you a list of everything in your cupboards. Then you're gonna be able to tell it as you're looking at stuff, there'll be an interface, a chat GPT interface for Uber eats or Uber grocery or, uh, Instacart or good eggs. And you'll literally be looking at your pasta shelf and say, you know what? Uh, add some Barilla capellini, uh, to the checklist. And I'm

SPEAKER_385: like, dang great. It's a great use case. I hadn't thought about that, but I think it'll be an excellent

SPEAKER_69: use case. Imagine walking around your house, just looking at every single, uh, you know, uh, appliance and then having a catalog every appliance. I'm always thinking about estate management now, because I have two houses and you start to think about these things. Yep. Just walking around,

SPEAKER_341: looking at everything, the back and the front of everything, getting every model and then telling

SPEAKER_00: you, Hey, you know, your Sonos is at a date. This is a seven year old Sonos. Here's the new ones. This is the cost difference. Would you like me to get you a quote for this? Right? So you take something like thumbtack, which we're investors in and imagine thumbtack. You just walk around your home and it's like, Hey, you know, the handle's broken here and you just speak out loud. What's wrong? I think, I think this deck needs to be sanded. I think this needs to happen. Just boom goes out and

SPEAKER_479: gets your quotes. Wow. Yeah. Yeah. It being your assistant on the side. And this is where Siri could

SPEAKER_69: actually make an incredible recovery. Imagine Syria was able to do that and they made a Siri connected

SPEAKER_105: to a chat GPT four level. I take the other side. I think you'll just, I think you said it, right? Um, the apps will do it, right? And so there'll be incredible apps that you'll load up. It doesn't

SPEAKER_198: even have to be native within, uh, you know, sort of the operating system of the headset. So you'll just have an app that's there for inventory app for what should I do? What can I eat?

SPEAKER_93: I think you'll have all these kinds of use cases. That's pretty cool.

SPEAKER_27: Yeah. Uh, all right. Listen, amazing job. Really appreciate you taking the time to come on again and we'll see you all next time on this week in startups. Bye. Bye.