0:00In the cold open, an app Jordan Metzner built in Google AI Studio looks at him through the webcam and correctly guesses his costume: a vampire.
2:35Jordan Metzner says his Mirror Mirror on the Wall app uses real-time speech to text, sends it to Gemini, then reads the reply with text to speech. It was slow and named Microsoft instead of Nvidia as the most valuable company.
5:16Jordan Metzner says Google's vibe coding tool at ai.studio/build lets you prompt an idea or pick from 16-plus starter ideas, such as a Nano Banana app or a conversational voice app, with Gemini already integrated.
6:44Jordan Metzner prompts a 'guess my costume' app, and Google AI Studio builds it using the webcam and Gemini, suggesting extra features like a difficulty setting and a timer. It identifies his vampire costume.
8:22On Sam Nadler's suggestion, Jordan Metzner one-shots a trick-or-treat route planner that uses Google Maps. For an eight-year-old in Beverly Hills it recommends a one-hour route in the Flats with safety tips.
13:04Jordan Metzner calls it a step function improvement because Gemini is built in, so you do not need a back end, an API choice or a key to add AI. A Nashville run recommends Sylvan Park, which Sam Nadler says is spot on.
15:28Sam Nadler says Nvidia, the first company past $4 trillion, crossed $5 trillion after announcing a deal with Nokia, because demand for its chips is insatiable. Jordan Metzner calls Nvidia the number one shovel dealer of AI.
18:07Sam Nadler says Elon Musk launched Grokipedia as an AI-powered rival to Wikipedia. Jordan Metzner says Grok needs its own source of record, and Grokipedia will need strong SEO to rank on Google.
20:12Jordan Metzner says anything written by humans is biased, including Wikipedia. With Grokipedia, the bias sits in the AI and the people who built it, so it would show uniformly across all articles rather than in single pieces.
Google AI Studio Build: 3 Halloween Apps Vibe Coded Live: the key questions from this episode, answered with quotes and timestamps.
Jordan Metzner [0:00]: A vampire. Alright. So here's what it says. The spirits have spoken. Woo hoo. What a spiky sight.
Sam Nadler [0:07]: Can you talk with the fake teeth in?
Jordan Metzner [0:10]: No. Not that much. Sure.
Sam Nadler [0:25]: Hey, everyone, and welcome to Built this week, episode 19 special Halloween episode. And Built This Week is the podcast where we share what we're building, how we're building it, and what it means for the world of AI and startups. I'm Bad Bunny I mean, Sam Nadler, cofounder here at Ryz Labs. And each and every week, I'm joined by my friend and business partner, the vampire himself, Jordan Metzner. What's up, Jordan? Can you talk with the fake teeth in?
Jordan Metzner [1:04]: No. Not that much. Yo, Sam. Happy to be back. Oh, wig's falling off. It's a huge mess over here in Vampireland. But, yeah, happy to be back for another Halloween special spooky episode. We got a lot of spooky treats today.
Sam Nadler [1:22]: Yeah. And how does AI impact the life of a vampire?
Jordan Metzner [1:25]: It keeps me up all night.
Sam Nadler [1:29]: Awesome. Well, before I get into the docket, please don't forget to like and subscribe. We have new episodes out every Friday on your favorite podcast platforms, whether that's YouTube, Spotify, or Apple Podcasts. And we have a fun docket today. We're gonna go over another, Halloween type product that we built called Mirror on the Wall. Who built the smartest AI of all? It's just a fun kinda quick build just to highlight the fun we're having, this week. We're also gonna cover the Google tool to, vibe code. And then lastly, some news, some really fun and interesting news every week. Grokipedia, Nvidia becoming a $5,000,000,000,000 company, and that's about it. Any thoughts before we dive into the mirror on the wall product we built?
Jordan Metzner [2:21]: Yeah. So we made some Halloween fun treats last week, but I thought we'd escalate a little bit more as we drop this episode on October 31, Halloween. So shall we jump right into it? Yeah. Show me. Okay. Cool. So this is built using Google's new AI build studio and we'll talk about that just a little bit more. But let me jump into the mirror on the wall. Let's see how this goes. So I'm going to make my chat bar a little smaller here and allow my camera. There we go. Okay. Cool. Alright. Mirror on the wall. Who is the smartest of them all? Okay. So there's my mirror on the wall. You can see every time I say something, it's kinda pulling in my words in real time. But let me just refresh so we can do another query. Maybe we can ask it something that Mirror on the wall, who is the most valuable company on the stock market today? Okay. So it didn't get it right. It said it was Microsoft here. It looks like it says that it's Microsoft instead of Nvidia. But, and it's a little slow and laggy. But you can imagine this inside like a little mirror. All of this is powered by Gemini.
Jordan Metzner [4:23]: It's using real time speech to text and then sending it out to Gemini and then Gemini is replying and then it's doing some text to speech as well. What do you think of my spooky mirror?
Sam Nadler [4:35]: I think it's cute. I know you live on a street where there's a lot of Halloween activity. I'm curious if you could set up a little webcam and a monitor and a microphone and have some kids do some mirror, interactions with this product. But, yeah, it's a fun, quick build.
Jordan Metzner [4:59]: That sounds like a cool idea. I don't know, like, I gotta think about it a little bit, but I guess you're right. It sounds pretty feasible, so I'm trying to think, like, a monitor, a camera, and some speakers maybe, and we could pull it off. Yeah.
Sam Nadler [5:16]: Okay, Jordan. Thanks for the mirror demo. I think for the tool of the week, we're gonna cover, the vibe coding tool, the Google builds tool within the Google AI Studio. We, covered it as news last week, but in the past week, we've both been using it. There's some really fun features that I think are improvements from other similar tools. But, yeah, why don't you show us some of the unique things about this tool within that Google has recently launched?
Jordan Metzner [5:50]: Yeah. So we've talked about a bunch of vibe coding tools in the past from Bolt.new to Lovable, v0.dev, Replit, and all different ones in between. Amazon has one as well, and we've just seen a bunch of them come out. And Google finally launched their own. It's at ai.studio/build if you wanna check out the website. And you can simply prompt your idea, and it'll start building. It has Gemini already integrated, or you can pick one of their, looks like 16 plus ideas, including, a Nano Banana powered app, a conversational voice app, animating images, all different cool stuff that you can do with Gemini. I built the mirror on the wall with this app. But why don't we do something just because you know, we're we're doing Halloween. So why don't we do like a okay. Let's do a guess my costume app where you use the webcam and you guess who I am, who or what I am dressed as for Halloween. Does that sound cool? Alright. So all you gotta do is hit build and Gemini will do its thing.
Jordan Metzner [7:13]: So if you think about this app in general, it's gonna use, some JavaScript on the front. Oh, let me show you some other cool features. Here you can add a difficulty setting, so you can make it more of a game. Okay. Cool. We can add that as a future feature. And then it'll give you some other features, like, you wanna incorporate a timer to increase the excitement? Players will have to guess a costume, so you can add that. Okay. So here's the guess the costume AI. I'm gonna put my vampire teeth in.
Sam Nadler [7:39]: Fixing the wig. Okay. Summoning spirits to guess.
Jordan Metzner [7:45]: A vampire. Alright. So here's what it says. The spirits have spoken. Woo hoo. What a spiky sight. With those gleaming things and the glowy silver locks and that magnificent ruffled shirt, you've sunk your teeth straight into the classic elegance of a vampire. I bet you're ready to charm the blood out of everyone at this Halloween party tonight. Okay. So alright. So, yeah, this was all built, obviously, using the Google AI Studio. Here, you can see kind of what it did. It uses the webcam. It uses Gemini. And, yeah, they made it a little like Halloween style.
Sam Nadler [8:22]: Can we try just in the essence of creating something a little bit different, maybe try a different prompt just to here, I'll give it to you. What about what if we had a route optimization for trick or treating, and we paste in a either a neighborhood or a screenshot of map coordinates and with the goals being to cover as many residential homes as possible. And, we can put in our parameters for total number of miles or kilometers walk, total radius, etcetera, and just see, what we can do. It's slightly more practical. Maybe your, business is looking to optimize a route for deliveries or something like this. In this case, it's, optimize, number of houses covered or something like that. And, it allows any, user in The US to choose their city, choose their neighborhood, select the parameters, maybe how old they are, five to 18 or whatever the different parameters are, and it can tell you, okay. Well, you should hit the Westwood area of Los Angeles or something like that.
Jordan Metzner [9:42]: Alright. So I wrote, like, you Know, please create a route tool that optimizes my route for trick or treating to cover as many good homes as possible to do a one hour trick or treat in any city in The USA by number of houses covered, allow any user to pick their own city, how old, what other parameters they want, and tell it where to go and what's best. Does that sound good? Any other changes? Okay. Cool. So maybe we can build that. Alright. Well, a good thing that building on top of the Google stack is obviously that they have Google Maps. Part of this tool, like I mentioned, they give you features. So it says here, integrate real time weather data. But, yeah, it said integrate route, excuse me, integrate weather just in case that's an impact. It also suggesting to allow users to create profiles and save their preferences of past routes. So that's kind of a cool idea as well. Maybe we'd add that in the future. And let's see.
Jordan Metzner [10:33]: It even recommends using let's see what other feature it had. Predicting which houses are likely to give out the best candy based on historical data and user reported information. So, maybe you want to take some notes on which houses gave you the best candy last year and then you can
Sam Nadler [10:52]: Yeah. Those full candy bar houses, you wanna revisit those every year.
Jordan Metzner [10:58]: Oh, yeah. Alright. So it is building right now, so there's not that much to see, but just give it just a few seconds here. Alright. So let's do like Beverly Hills, California. Let's say eight years old, quality candy five out of 10, spookiness five out of 10, find my route. Alright. So it's consulting the spirits of the sweetest spots. Routes are generated by suggestions, always trick or treat with caution and adult supervision. And again, this is a one shot, know, we haven't done any follow-up here yet. So let's see how this goes. Alright. The Flats of Beverly Hills recommended neighborhood start on Crescent Boulevard. The spooky stroll for a one hour route, start here, walk north, go here. Then it says, why is this good? The Flats of Beverly Hills is renowned for excellent trick or treating environment. It boosts the highest density of well maintained single family homes that are typically generous with candy and perfectly aligning for at least a five out of 10 candy quality experience. The streets are wide, well lit, and generally flat, making it safe and easy to navigate for children. Many residents embrace Halloween. Okay?
Jordan Metzner [12:09]: And then it's just given us some safety tips as well. So, yeah, Trick or Treat Planner. Now we've built multiple apps in this episode, and you just see how easy that is. And as I mentioned before, recommended route optimization based on weather and any other things we could add here.
Sam Nadler [12:28]: And if you wanna make a suggestion, you just prompt it down here as a, an iterative, and maybe add a note taking feature. So as you visit these homes, you can record who had licorice and who gave out the full Snickers bars.
Jordan Metzner [12:50]: That's right. And you can also click some of the suggested features like AI features, which it'll just generate some, add a current location button, improve the route optimization visualization, and, yeah, just gonna go do it.
Sam Nadler [13:03]: So, just to wrap this up, from a quick glance, it seems really similar to the other tools out there. What is it about this tool in particular have you really liked? Is it just the suggestions as you're building that that kinda really helps push the product along quickly from a v zero to a v one, or is overall, do you think there's something foundationally better?
Jordan Metzner [13:30]: Oh, no. There's a step function improvement here, and the step function improvement is that Gemini is already integrated. If you were to go back to like a Bolt or a Lovable or any of those and you wanted to add AI into your app, you need to, usually build some type of back end or something to call those APIs. You need to figure out which API you wanna use, you need to get the key, you need to give it to one of these types of tools. And, the fact that Google's tool has Gemini built in means that you can rely on Gemini to be doing the AI inside your application. We made some quick changes here. We'll run it one more time. Let's get a sixth level on quality. Maybe we can try something like Nashville. I know you live in Nashville, Tennessee, and let's just find a spooky route.
Sam Nadler [14:16]: Yeah. Amazing. So we have a
Jordan Metzner [14:19]: Map. A map? Yep. This is Nashville. It recommends to go to it gives you kind of the whole breakdown. It tells you why Sylvan Park is a great area for
Sam Nadler [14:29]: My old neighborhood. Yeah. It's like a very residential. Houses are pretty close together. So if you scroll up, I used to live on Dakota, which is part of the route, I think. Yeah. Right up there, you cross Dakota. So, yeah, it's a great area and very close to where I am going to take my kids and where we go every year.
Jordan Metzner [14:54]: So this is pretty much spot on. Okay. Great. So, I think we talked about it last week, but I just wanna wrap up. The new Google vibe coding tool is obviously pretty fun to use, really cool to play with, highly recommend to use it. And, yeah, I think, know, just wrapping up, we did, two, three Halloween tools in, just minutes and, for pretty much no cost almost, you know. So these are great throwaway ideas, super fun, and, yeah, I hope the Gemini team likes it too.
Sam Nadler [15:27]: Alright. Let's transition to news. I think both of these topics are and news as of today. And number one is NVIDIA continues to break records. It was, I believe, the first company to cross over 4,000,000,000,000 total valuation. As of this morning, crossed to 5,000,000,000,000, in valuation. And huge news yesterday, NVIDIA had, a big presentation. They announced this deal with Nokia. But, I think it while it's shocking, it all just continues to make sense. The need for NVIDIA chips is insatiable.
Jordan Metzner [16:10]: Every day we're seeing, this massive company continue to grow significantly. It says intraday, we're at five percent up today. I think it was up two and a half this morning or 3% already. So yesterday was up five, today up five, yeah, three and a quarter this morning. So, this thing just can't stop going. And, Nvidia is, they're the number one shovel dealer, and they make the best shovels, and everybody's buying their shovels. Yeah. You mentioned, the Nokia deal. Obviously, there was a great shot with Jensen and Alex Karp from Palantir. And, yeah, I think, we're gonna continue to see NVIDIA break records, launch new chips, and, continue to power basically this AI evolution or revolution that we're going through.
Sam Nadler [17:03]: Super exciting. I've said it before, and I'll say it again. I'm a big believer. I still I think in NVIDIA could go to 10,000,000,000,000. I mean, it's not that crazy with if you think if AI isn't a bubble and NVIDIA continues to be pretty much the only source for the chips and the most efficient chips, then, it's gonna keep growing. So, I'll probably eat my words as don't this is not financial or stock advice, but, it's just a bit an amazing story. And, for me, I'm not a savvy investor, but learning more about Jensen and, seeing more of his interviews makes me more bullish on the company.
Jordan Metzner [17:56]: Yeah. I mean, think, NVIDIA is at the center of this AI revolution, and it's not gonna slow down. Let's move on to Elon Musk, and then, we're almost at the end of this episode here.
Sam Nadler [18:07]: Cool. So next story is, Elon launched Grokipedia, which is an AI powered rival to Wikipedia. And, yeah, I think, the immediate reaction has been that it's it has some biases. But, yeah, I think what do you think about Grokipedia?
Jordan Metzner [18:30]: Well, Wikipedia is just like has been the standard of, kind of information online since, like, the encyclopedia moved to digital format. And so kinda most people use, like, Google plus, like, Wikipedia as kind of source of record. I think, as we now have LLMs, those become a new source of record. You don't really need Wikipedia when you ask ChatGPT or, kind of any of these chat tools, some factual information because they're able to source that. And I think that's probably one of the reasons why they launched it is that, in order for Grok to be successful, they need the equivalent of Grokipedia, and so why not just, put a free website up for it? I do know that Elon said he's done it so it can be as an alternative to woke Wikipedia, he said. And I know he's made fun of, like, the guy who makes the most amount of edits on Wikipedia, etcetera. But it does make sense that, like, over time as AI gets better and better that, the encyclopedia is written more and more by, AI that can write it fast, quickly, real time. And do we even need a website of record when we have these chatbots that can generate any piece of content at any time? So I do think it's interesting.
Jordan Metzner [19:45]: I think he's kinda like poking the bear. It is very hard to rank for a lot of these terms on the top of Google, and so I think, like, in order for Grokipedia to be successful, it's gonna have to have a really good SEO. And, we'll see how much money and time he continues to invest in it, or is it just like another website that, he launched as a fun joke, maybe just the backbone of, what Grok needs to do as far as kind of, maintaining a source of data and information.
Sam Nadler [20:11]: What do you think about the you know, Wikipedia is famous because it's, like, crowdsourced information almost, and, I don't it there felt like there was a degree of truth to that. Does Grokipedia, not having that or from my understanding, not having that same kind of crowdsourced input, does it potentially hurt its ability to be that source of truth?
Jordan Metzner [20:41]: Yeah. Well, I mean, anything written by humans is gonna be biased. So, probably everything in Wikipedia has some bias in some sense, and that's kinda like the whole, like, crowdsourced idea. But I think as you know, I mean, there's a whole, like, cabal that controls Wikipedia and its changes. And, if you wanna get a Wikipedia article about you can go online and have some people that are, like, famous writers write it for you. And, there's kinda like a gray black market to everything. And so, if Grokipedia is written by AI, even, the biases lies in the AI that, generating it and the humans that built the AI, but not so much in probably, like, the content on a unitary level, like, on a piece by piece basis. So, if we overall know that, like, the Grok engine is biased towards x or y, then, like, we'll see that slant across, like, all articles uniformly rather than, one specific article, for example, that that could be taken out of context.
Sam Nadler [21:34]: Well, that wraps it for episode 19, the Halloween episode. We covered Mirror. We built some additional tools using the Google vibe coding tool, which we're we're both using pretty much exclusively in the past seven days. It's been a lot of fun. And some, huge news with NVIDIA and Elon's Wikipedia competitor. Anything to wrap up here? And if not, happy Halloween.
Jordan Metzner [22:02]: Yeah. Happy Halloween, Sam. Hope everyone has a great, fun, safe Halloween. This is the first Halloween where you can vibe code your own mirror on the wall and a bunch of costume designers and much of other ideas. But I think you gave me something cool. Maybe I'll I'll put something outside of my house or something like that. But, yeah, I think that's super cool, really fun. And hope everyone has a really great Halloween. Hope you can use my tool to find the best route to take your kids to. And anyway, I think what are we gonna see next week? More NVIDIA gains, more growth from all these hyperscalers. So yeah. So Something like that. So anyway alright, Sam. Have a happy Halloween, and to all of our listeners, I think we're almost about to hit 15,000 subscribers on YouTube. So please tell your friends, like and subscribe, and see you next week. Built this week,
