BONUS: We Built an App Live in 10 Minutes with AI (Vercel's CPO Shows How)

25:23Tom Occhino:Doesn't have to be, you know, authenticated, et cetera. So if you're building a general purpose thing for anyone, it doesn't need to connect. But I really love building personal software. It's just the thing that I built for me. And, you know, therefore, I can connect it to those things. I want to show a couple other things. Yeah, and for what it's worth, it makes connecting them really easy. I, you know, because what I've been working on, I'm not going to show it today because we're not there, but really soon. I had to set up API connections and OAuths with Google, Slack, and some other tools.

25:54And it's like, you're going to go there. You're going to set this up. You're going to go to this menu. Here are the two things I need. Please name them this in your environment variables. And it was, you know, it just walks you through it in baby steps, frankly. Not baby steps. It really wasn't that different. But it makes it work with the parts that might be intimidating to someone who's new. I've got a good question in the chat. but I saw you were going to, let's show the design feature first.

26:24Tom Occhino:Yeah, that's what I wanted to show next, actually. This color scheme, I think, is great. I love, you know, the blue and the green, but, like, we can very easily transition between various different colors here. I actually, like, there's a few other things you maybe can't see as well, but, like, if I wanted the radius, all of the radii on all of these things to be different, I can kind of adjust some of these things. But you can also do, oh, and here I can do a, let's see, let me do a light mode oh you can get very granular with your colors um i love being this is this is all built into v via v zero this is all built into v zero that's absolutely right um and you know you can select your fonts and such i think you know i would have to i'll tweak some things here and it's telling me okay you've got some unsaved changes here i'm going to reset the thing that i really like is you can actually do this like point and click edit So this is my portfolio tracker.

27:19Tom Occhino:I can change the name of that and then save. And we can make changes in real time. You can also prompt to make changes. So you can say, you know, if I wanted to like modify this, it says refine this element. Let's change positions to, let's say, equities.

27:44Tom Occhino:and now it'll prompting granular control. Yes. You can highlight the thing and you can say, fix this. Super granular changes and it'll try and do the minimal change to do that. But the other thing that's really cool here is, and let's see, it's doing this really quickly. KPI card. There we go. Found it. Updating the label. I like that it kind of tells you what it's doing. So you can kind of follow along. I sit back with a cup of coffee and I'm like damn look at me go look at me go look at me go and so it kind of updated that for me um but the other thing that's really cool so last thing on this one I'm going to click this publish button and I'm going to publish this to production and so what's happening when I click this is we are we are taking the application that was just created we're going to actually compile it into the artifacts that are needed to deploy this globally around the world for anybody to be able to access.

28:42Tom Occhino:The thing that I feel good about the fact that the visibility of this is only my team, right? So what we've done here is because I'm creating on my team, by default, you can only create public visibility to your team. But when you create with your account, you'll be able to decide, do you want it to be publicly available? Only people with the link? Is it password protected? So we really care about that visibility and the security associated with it. But what's really cool here is I've actually just published this. I will have access to it because I'm authenticated, as you just saw. This is a live website.

29:18Tom Occhino:This is at v0financialportfoliotracker-fi.versell.sh. And again, you know, lockdown to only my team, but I can share this with anyone on my team and they can see what I just built. and better yet, if I wanted to, I can give them access to fork it, if you will, or create their own version of it. So like, you know, this is such a good collaboration tool and such a great workflow. Like that thing is live now. And I can make changes to it and publish new versions. So yeah. Wow. You're totally solving the vibe coding's first problem, right? Which the first problem is, okay, I can make code. I can even make apps and they look beautiful in my UI.

29:57Now how do I actually get it out to people?

29:59Tom Occhino:Yeah, that's right. And critically, this is powered by Vercel. Remember I said the Vercel core platform, VZero is an app that sits on top. We've been building the Vercel core platform for the past decade, right? So the CICD flow and building it and deploying those artifacts at scale and giving you that URL and that thing has an SSL cert automatically configured and deployed globally. Like all that stuff just works because VZero is built on top of Vercel. So Tom says, I'm shocked this was created with one prompt. I guess I always thought vibe coding would require a lot more work. Yeah. Yeah, look, you know, I think it's getting better and better literally by the day.

30:38Tom Occhino:And the team is really working on it. I mean, even though this looks like a sophisticated UI, this is a fairly rudimentary example. We only really have one page. It's sort of a single page app. We can do a lot more. And actually, I don't know if it's time yet. I guess I'll leave it up to you guys. Like, do we want the audience to decide what we build next? Do we want to try to do another demo? Like, yeah, absolutely. We let them decide. All right, guys and gals, whoever's out there. What should we build? What should we build? Give a row some suggestions in the chat. Tom, I will ask you some of your best best practices for prompting.

31:21You mentioned that you had just shared a prompting guide for sale put out. to kind of help with this type of thing to Tom's point. And while we're waiting for ideas to come in, send in your ideas, we'll prompt them. I want to circle back to one question that was asked earlier, which I think is a really good question. Michael Jabara said, how do I MCP to my existing vendors like Tavoli or BrowserStack?

31:46Tom Occhino:Yeah, so you can actually connect MCPs directly in v0 here. so you know obviously you can you can do a lot of things you can connect design systems or you can select a design system you can ask it to generate images but here's where you would configure mcps so i can browse the existing mcps here's the ones we have you can also add a custom one if you enter the server details manually so that would be right in this little drop down here that makes sense prompt tools yep perfect perfect yeah we had another question how protected is the code i think this is something that especially a lot of vibe coders are concerned about and traditional devs are concerned vibe coders are doing which is not protecting their code so by default you know you have a personal v0 account or you have a team account anything that you create is private to that account right so you are the only one that can see it it's not until you click that share button or you click that publish button that it's available to anyone else um the thing that i really love that's really resonating, especially with larger companies, is that by default, when I have a company account, this is a little confusing because our company and our team are both called Vercel, but when I create something with the Vercel account, it's available to my colleagues if I want them, if I want it to be.

33:06Tom Occhino:It could be only me or it could be to my colleagues, but it's not open to the public internet. So they actually have to authorize and log in with Vercel. So by default, private. And then when you publish, it can be made available to anyone you choose. Yeah, and you could absolutely, I would say, you know, ask it. Do I have any client-side API keys? Do I have, you know, the kind of questions that I read. I read an article the other day that had a list of like the 10 things you should ask while you're doing this to make sure you're safe. I need to find that. We need to share that in the newsletter because people need to know this.

33:43People need to know that. But with little things like that, here are the most common vibe coding trip ups. and just go in and ask a tool. That's right. So we have a couple ideas in the chat. So these are a little vague. I wish we could get a little more specific than this, but I have a favorite. Rachel McNamara says, something that shows my trades. Randall Novel says, something that uses AI to process data and do something with it. It's like, cool. Yeah, that's a good idea. An app that organizes my decades of misguided PC file storage.

34:21Tom Occhino:I feel that one. I did too. This seems a bit easy, but what about a media organizer like an RSS feed plus a newsletter podcast? I like that one. Corey, which one did you like? I'll be honest. I'm a big fan of that one and of the decades of misguided PC file storage because I am a disaster. when it comes to search things. Tom, do you want to do tiebreaker? Which one of those do you want to do? I really like this idea of a media organizer RSS feed newsletter. I think that could also fit into the theme of the other one. So let me try and prompt. I'm going to use a tool called Whisper so that I can type with my voice here.

35:03Tom Occhino:And we'll try to give it a prompt. But then I'm also going to show you a little trick after that. So let's build a really professional-looking RSS reader that helps me organize all of the things I want to read, all of the newsletters, magazines, and podcasts. So let's see if we can insert that. Okay, so this is what I just typed. But I'm going to – there's a little trick here. I'm going to ask V0 to enhance my prompt for me because I'm a little bit lazy. And let's see what it comes up with as a suggestion. This seems like a little bit more detailed. So I'll again tell it to start cooking and we'll read kind of what it's going to be building here.

35:42Tom Occhino:So RSS reader application, newsletters, magazines, podcasts. That's what I said. Platform should support seamless addition and categorization of feeds. Thank you. Customizable interface, easy navigation, search, filtering, bookmarking. Nice. I love bookmarking. Design should emphasize usability, aesthetic, clean, modern, responsive. Oh, this is good. I like when we call out explicitly that we want it to be responsive, so it works on mobile web. Suitable for deployment as a web application, leveraging common web technologies. So it's got a design brief, created a code base. It's got some things, and it immediately started cooking.

36:19Tom Occhino:I also checked the box when we did the generation here. I said generate images. This, obviously, generating images via LLMs costs a little bit more. we've got the price there. But what it does is it makes a much more compelling sort of first experience. One of the things I didn't ask for here was to populate it with sample data. We'll see if it's smart enough to do that as it cooks, but I'm excited about this one. We actually have a specific request is to connect it to a live data source from MJ23312. So perhaps we could try that. Let's see if we can do it. So let me see if I can find... I think we should, I think it'd be cool if we connected this to the Neuron podcast, the website with the newsletter archives.

37:06We do have an RSS feed. We do have an RSS feed. Do we have one for both the newsletters and for the articles we publish by chance? This is a good question. I believe so. I believe so. Another option would be, you know, you can query the Reddit API from like Local Llama and some of these other places pretty, pretty for free. So we could do that as well, if need be.

37:32Tom Occhino:Let's see. Here's, let's see if I, popular RSS feeds. And maybe we can get a list of the most popular RSS feeds. And then obviously the neuron will be added to this list. Of course. The most popular RSS feed. No, I'm just kidding. While you're doing that, I want to circle back to a couple other things that were said in the chat. Process delivery had a more specific question, which was, can I create a web-based app for client usage, but the client can't see the code? A web-based app for client usage, but the client can't see the code. So basically the way I'm interpreting that is you like would minify perhaps and obfuscate the code that's shown on the front end.

38:32Is that what you're asking? Like basically like putting it behind, making it private. I don't know the right terminology for that, but I kind of get the idea that he's going for. I get the idea and it's way beyond me. Yeah.

38:47Tom Occhino:Yeah, I mean, I would try. I would basically ask if that's a... But how to do that, right? Like you could ask V0, say like, hey, this is what I want to do. Can we write the code in such a way that this is how it works and see what it says? Yeah, we could definitely ask it to do something like that. I think that's great. I'm multitasking here. I'm trying to also find the Neuron podcast RSS feed URL. if uh if you if you can't find it uh let me let me see if i can pull it up and share it i'm looking to uh i mean we would we could get it from youtube we could get it from spotify we could get it from apple youtube's probably the most easily uh discrepancy youtube channel rss feeds Hey, if anybody in the chat knows how to do this, has a quick tip on that, let us know.

39:49Okay, we've got two other questions. We'll see if we can answer them while multitasking. Can this use AI to generate a summary of the content in the feed from yesterday? So I'm interpreting this to mean, like, let's say, for example, you're trying to, you know, get all of the stuff that happened yesterday and then have it summarize it for you. potentially it could summarize it based on the descriptions in the rss feed but that would be a lot of fetching if it had to go and fetch everything and to like give it to you with 100 fidelity um what's your take on that tom it could absolutely do it i mean that would just be

40:24Tom Occhino:something that i would ask it for first i would connect it to all of the rss feeds that i care about um and then i would ask it to give me the summary of all of the rss feeds that i'm that i monitor and care about. So I do want to like, should we look at it here a little bit? Yeah, let's look at it. Pretty good. This looks like a pretty good thing. So, so far we have a reader, it's called it Readwise. We have all articles, unread, priority. This, does the bookmarking work? Oh, indeed it does. I love it. I can actually bookmark. It's so good. add them to my priority feed here. Let's see, technology.

41:03Tom Occhino:It actually, it looks like it did populate it with existing articles. So the Verge, TechCrunch, maybe, Harvard Business Review. Yeah, can we click out? Does it work? I mean, you know, this is a pretty solid start as far as I can tell. It sure is. I like the layout. I like the functionality. I've liked it. You can change the layout. Hold this first. Yeah, I mean, you know, we're filtering here. Let's see all types. And then let's search for the feature of work. I mean, this search works. So, like, yeah, I mean, somebody asked us to connect this to real data sources. Let's add a real couple of real data sources via the following feeds.

41:51Tom Occhino:And let me see if I have that ready to go up here. Great. this should be a good list hacker i found what i think is is our rss feed for our youtube channel oh let's go in the private chat yeah i dropped it i'll paste that one into attached in the pasted text along with this one i gotta get that from you yeah this is so fun i love doing it live and just like exploring we're just trying stuff out we're in the arena trying stuff out exactly you gotta get in there and test them the pasted text i gave it was hacker news bbc news tech crunch artist technica and npr news as well as your youtube feed so i'm actually curious to see how it mixes like rss feeds that contain articles plus youtube videos is going to have to get kind of creative.

42:54Tom Occhino:All of these are public. All of these are open. So it doesn't need any type of authentication. But I mean, that's another thing you could do. The other thing that happens is if you ask it for an app that requires a database connection, it will actually prompt you right in the chat. Like it shows you a little bit of generated UI that says like install a database here and you click it and you go through that flow where you connect to a database. So that's pretty sophisticated. and pretty cool. So look, you know. I was actually wondering about that in the last one. Like, for example, it created that pretty sophisticated CRUD database.

43:29Like, how would you connect that to the actual Superbase through that prompt?

43:35Tom Occhino:Yeah. Or whatever tool you use for the map. What we'll do after this one is I'll switch over to my personal where I have admin permission to create databases. I'll show you what that looks like. If it determines that it needs a database, it just prompts you. If you ask it for a database, it will also prompt you, and then you can go through that flow. So I'll show that next. It's adding these real RSS feeds to it. You know, it created a mock data, so now it's trying to incorporate the real stuff. So letting it cook on that. Maybe I can actually open a new tab here where I will transition over from Vercel to my personal premium subscription.

44:19Tom Occhino:and we will do this one live as well. I'm going to attempt to answer this question from Tom. How would it handle podcasts? Could it pull individual podcasts from shows you might not know about? It would probably need to have the RSS feed for a bunch of different shows to do that, but you could do something like that with a web search-like feature where it searches for podcasts based on whatever topic you're interested in and then populates them based on that. I don't know if you agree with this, Tom. I don't think it would do it automatically per se, but you could prompt it to go out and find your new stuff on a regular basis.

44:56Tom Occhino:Yeah, vZero can also use tools to go out and do things like web search or to ask an LLM for something. So it can use the AISCK under the hood or use the AI Gateway under the hood. You can have it ask the LLM for recommendations, for other sources, and then you can build a more sophisticated app. You know, I think what we're doing now because we're doing this live is we're kind of doing the one-shot first experience. But when you start tinkering with this, you literally just ask it for the thing that you're looking for. And not only will it tell you about what it can and can't do, if it can do it, it'll tell you exactly how to do it.

45:34Tom Occhino:So the beauty here is it's not necessarily about, like, what is the laundry list of features. It's about your ability to explore the capabilities by just asking it, can you do this? And it's going to say, I don't know, let's think about it together. And we discover so much new emergent capability by just asking it, like, you know what I wish this could do? And actually, Corey, I mean, before we started the live, you were talking about like, yeah, I just asked it to connect to my Gmail. And then it like told me I need this and go to this settings page and do that. So you can think of it not as like a calculator, a service that sits over there that you send requests to, which is kind of how we're using it right now.

46:12Tom Occhino:You can think of it as a sidekick that's sitting next to you and you're kind of in conversation with it. And it's like, well, let's explore. Let's check it out. Let's try it out. So we got another question real fast. What kind of versioning is baked in? You can connect this to GitHub so you can get and push at any point, right? That's exactly right. So, you know, one of the things we have here is the ability to and, you know, the button is not shown here because of the permissions that I have on this for sell accounts. That's actually one of the reasons I wanted to do it on my personal account is when you do one of these generations.

46:49Tom Occhino:So here's one that I did a couple of days ago. So I asked it for the prompt was build me an app for the magazine wired with social features, add sample data for a bunch of different articles, commenting and social features. It actually inspected the wired website and was like, Oh, here's the rough aesthetic. and it can kind of try to recreate that for you to start with. But you'll see here that not only, I mean, this is a very compelling generation. This is the most impressive one shot I've seen in a little while. But what you can do is you can connect this to a GitHub repository. So I can connect this to get to my Tomokino GitHub.

47:23Tom Occhino:It'll create a net new Git repository. And then from there, I can use any other tool. I can just go to GitHub. I can open it up in my editor and I can do other things. So from there, you know, I think that becomes a jumping off point for a lot of other stuff you want to do. Yeah. And to Tom's point, maybe versioning makes no sense in AI. No, definitely version with AI because you never know. AIs are a little bit unpredictable. They could always take things in a weird direction. Yeah. And every time it makes a change, it automatically updates your repository. That's exactly right. Yeah, so what happens here is, you know, for now, if I haven't connected it to a GitHub repo, what's happening is it's adding, you know, anytime it does new functionality, it responds to something, it just says, oh, it's v1.

48:12Tom Occhino:If I was to go and make some changes here, it would say, oh, now it's v2. And we can kind of see that in our RSS reader app. I think we've probably gotten up to v2 here, you can see. So it's versioning itself. I can go back to the original generation. This is the v1. That's stored. That's ready to go. I could even from here, I could fork this. That's what I was going to ask. Exactly. That solves my problem. There you go. See, and I can say, you know, this chat is read only, but I can create a new chat from that point in time. And somebody else that I share this with could say, oh, I want to start from here.

48:47Tom Occhino:Now, when I connect it to a GitHub repository, a Git repository, what's really cool is every time the chat does something, it becomes a commit into the code base. You can actually revert those commits or fork from those commits or branch off of them as well. So back to V2. Amazing. It's a pretty sophisticated thing here. And what I like is it grows with you. As you start to explore what it can and cannot do, you'll say like, okay, it doesn't really do this particularly well. Then you'll come back a couple weeks later and you're like, oh, my gosh. Not only does it do that, it does this other thing too that I needed.

49:26Tom Occhino:So we're iterating on this so quickly, and it's just becoming this cornerstone of how we do a lot of things. We talked about the idea that this is a communication tool. The functions, the people at Vercel even that are using this every day to make their own job better, it's not just developers. It's not just product managers. Account executives are dumping, you know, these total economic impact analyses and total cost of ownership analyses and generating apps that are much more compelling than a spreadsheet. Anyway, I could yap for days. No, that's awesome. Okay, there's two or three things I want to double back to.

50:08So the first one was the database you were going to show us on your personal account, like how that works. Yeah, that's one of the things we wanted to do here.

50:14Tom Occhino:So what we'll do is we're going to start a new chat. and the idea for the new chat, maybe let's build a simple app that will track something that needs to be in a database. Let's build like a blog with a database backing it. How does that sound? There you go. Sure. Let's build a simple blog that I can use to publish articles and back it with a real database. I'm going to, again, have V0 enhance this prompt for me because I wanted to build some better functionality. One of the questions I had was, What are the things that it's doing to enhance the prompt? If you can just kind of broadly overview.

50:51We'll also share the prompt blog that Vercel just published, which I think is really cool.

50:56Tom Occhino:Yeah, for sure. So one of the things it's doing is think, okay, when others have, you know, other sort of, I can, the LLM can go out and say, like, what are the good characteristics of a blog? Right? So I said the word blog, and it's, like, going to ask the LLM, what is a good blog look like? What are the features of a good blog? And so when I enhance the prompt, it's basically taking the very minimal information that I have given it, and it's turning it into much more helpful information. And I'm going to try to share the link to the how to prompt. Oh, no, we got it. I can share it in the chat here.

51:33Tom Occhino:Yeah. That's great. This is really cool. So what it's doing here is it's giving me an architecture plan, including a database scheme architecture. I mean, this is, I didn't ask for row level security policies. I love that it's doing the, I don't know how to pronounce this out loud, ASC2 diagrams. The ASCII art? ASCII, there we go. The ASCII art diagrams it's doing. In my head it's ASCII. It's drawing the architecture of the system for me, which is great. That's how you got internet street cred when I was a teenager. Cool. This is awesome. So what it's doing here is because it's designing a comprehensive plan, it's not yet building an app.

52:16Tom Occhino:I didn't ask it for an app. I asked it for a comprehensive plan for an app. Well, you know what I'm going to say as soon as it's done. I'm going to say, all right, can you build it? Yeah, totally.

52:31Tom Occhino:And let's see how much further it's going to go here. Emery, great. Would you like me to proceed with building this platform? I love that it breaks into phases automatically, too. Could you, like, let's say you're a pretty sophisticated engineer, or you're working with agent tools like Cloud Code or Gemini's different versions or codecs. Could you also at any point upload your code and start from there with V0? Or what would you use in that instance? Yeah, what we're working towards is a world where you can open any app in V0, and you can start from what you've already created. and so for now you know I think it works best with apps that started in v0 there's some constraints it imposes but very soon you'll be able to open any app in v0 and kind of we want to enable this sort of like fluid motion where you go back and forth from v0 to your other editor etc so whatever the best tool for the job is yeah coming back to our blog that we're building you can see this is kind of a new UI we haven't seen this yet what it's telling me is like I need need a database and I'm going to choose to install a Supabase database.

53:43Tom Occhino:I could skip here. And what it'll try to do is create some mock data and say like, oh, later when you connect a database, I'll swap out my mock data. But I'm actually going to go through the install flow so you can see what this looks like. When I click install Supabase, it takes me over to this marketplace integration. And I'll actually show you how, let's see how long it takes to provision a brand new database. So I've accepted the terms and conditions here for getting started with Supabase. I'm actually on my hobby account. We'll see if it'll make it work. Let's see. I have a Supabase free plan.

54:16Tom Occhino:I'm going to start with that. And I'm going to leave the defaults and click continue. I'll just let it stick with the defaults for naming. And now it's going to start processing. So this is creating a brand new Supabase database. I haven't logged into superbase i haven't created a superbase account i haven't done any of that um all it's doing is uh all of that for me and when i click done here um it's redirecting me back to v0 i didn't i didn't navigate in my i didn't even close the tab in my browser just brought me right back and it said okay we're con we're ready to continue building so that skip button turned into a continue building button and this is kind of interesting i could type this manually but when i click continue building, it's going to say, great, I've connected Supabase.

55:03Tom Occhino:Great, I've connected Supabase. And then you will know to say like, okay, let me go look to confirm that you did. I'm going to look for the necessary environment variables and the connection that I need in order to be able to run a migration and actually create this thing. So now it says, I've got some examples. I'm going to reload the thing. I'm going to read the layout, read the globals. Okay. I'm reading the readme for how to create Supabase. Here's my server client, my browser client. So the UI is going to guide you towards the right prompts that tell v0 what it can do next. So I just, I love that.

55:38Tom Occhino:And then here's our tasks. Oh, let's see it. Yeah, we're gonna let it cook. While that's cooking, one of the questions I had, because I've seen some concerns about, you know, let's say like a tool like Lovable, a bunch of people use it, spin up databases with Supabase and they don't protect those databases properly. How are you or how is Supabase kind of handling that for folks so that they aren't like leaving their databases totally exposed? Like what settings should they be focused on or what do you handle for them in that regard? Yeah, if you saw our comprehensive plan here, one of the things it enumerated was the RLS policies, row level security.

56:18Tom Occhino:The other thing here is that that connection to that Supabase instance is directly tied to my Vercel account. So this isn't a publicly accessible Supabase database. This is a Supabase database that is tied to the authorization provided by my Vercel account. And so not open to the Internet by default, absolutely has the right security defaults applied and sort of like as vertically integrated as you can imagine. Perfect. Nice. We have a couple other lightning questions that would be good to cover before we, because we're getting close to the hour mark here. Going back in the chat, Michael Jabara said, what is both the Vercel and the V0 pricing model?

57:00And do you support existing subs like CloudMax as LLM providers?

57:04Tom Occhino:Yeah, that's a good question. So the Vercel pricing model, the way to think about it is we're a control plane over lots of different primitives. And so, for example, if you have a static website that has lots and lots of artifacts in cash, we will kind of put those things in cash. You'll pay a small amount for storage. We will, you know, when a user requests them from your site, you'll pay a small amount for requests. But you're basically paying kind of on consumption. What we find is that our hobby plan, which is free, is extraordinarily generous for almost anyone. And then as you start to build a business that has real customers and you're scaling, our$20 pro plan, which is how you get started, actually ends up being even more generous.

57:46Tom Occhino:So you should be able to, in that$20 a month, scale to pretty massive scale. When you want to add teammates, you'll kind of like pay per teammate. And then the VZero pricing model is you're basically paying for some amount of consumption. There's no bring your own key to VZero yet. but it's one of the things we've talked about. We think it's good for all parties, right? Like we consume your tokens, we let you consume your tokens. And the way that we're thinking about that is the AI gateway from Vercel already allows you to bring your own key as a customer, as a user of the gateway. And V0 is powered by the AI gateway.

58:25Tom Occhino:One of the things that's hard about this is now we need to add more configuration and we need to add more sort of like sophistication, but it's definitely something that we're considering. The VZero premium plan, I think, costs like 20 bucks a month, but you get quite a number of generations from that. You know, like we were saying earlier, depending on the model that you choose, a million output tokens, which is quite a bit of output tokens on VZero Mini is about$2.50. And then, you know, a million input tokens is about$0.50. And we can do optimizations on that and such. But I've been kind of letting it rip here.

59:02Tom Occhino:I would say all of the everything that we've generated today on the show is probably sub a few dollars, cost a couple of dollars to generate. So if that puts it into perspective, when I think about the time saved and when I think about how, for me, I just wouldn't, if I didn't have access to the school, I just wouldn't build these things. I have all these ideas. I just wouldn't have time to build them. To me, the value that I get for the price that I pay is just very worth it. but it kind of varies and how much you use it is definitely a function of how much it costs. So we're committed to making this as accessible as possible and as generous as possible for as wide of an audience as possible.

59:38Tom Occhino:So we'll keep looking at ways to do that. Too many, or sorry, two more last ones here as we hit the hour mark. What were the differences between V0, Max, Mini, and Pro again? Max is like kind of like Claude Opus. yeah so so the the models that we use and we use a multitude of models for um prompt enhance uses one generate images uses one app scaffolding maybe uses one app generation uses another one um but the v0 max is powered by the latest uh claude opus which is uh sort of best in class for code generation i think according to the v0 team so it's basically just what model we we use underneath.

1:00:18Tom Occhino:And there's different price performance and also like latency output. So we're using the best model, which is why these generations take a little bit longer, but they end up being a bit more complete than if we were to do multiple iterations on a smaller model. Yeah. Tom, do you have to go right at 11 or do you have a couple of minutes? I have a couple of minutes. Okay, great. I appreciate that. One more question from process delivery. Are development advisory services available? If so, how does that work? I'm assuming this is from Vercel? Yeah, from Vercel, we have a technical consultants team.

1:00:53Tom Occhino:So when you become a Vercel customer, we can kind of onboard you. This really is only a thing that we offer to our sort of larger, you know, enterprise customers. I think we have 11 million Vercel users and, you know, many, many, many Vercel self-serve paying customers. And so we can't offer that at scale, but there are a ton of systems integrators that are deeply familiar with our platform and our frameworks and things like that, that can fill in the gap. Then Rachel McNamara asks, is the code generated maintainable and extensible? Yeah, you know, I mean, obviously biased. I'm speaking from my own experience.

1:01:30Tom Occhino:But in fact, I find this code to be, in many cases, better structured than the code that I would author myself. But whenever there's a preference, I say, you know, oh, I want this actually like handle save to be in its own file, because this has gotten quite long. And I want to abstract that away. You can go in and like, you can actually change that you can create new files here. So let's say I wanted to move handle save into its own file. I could make that, you know, its own thing. So it's not muddying up some of the things. So I say yes. But and, you know, everybody has slightly different preferences.

1:02:02Tom Occhino:And we give you the tools to be able to customize. I want to actually look at the blog that we've created. This generation requires that we run some scripts. And these are going to populate sample data. So I'm going to actually run this migration while we answer maybe some more questions. So it's running the scripts. We have a console here. Cool. John's list just said iPhone app. Oh, yeah. iPhone app. We do have a V0 iPhone app. And it's actually one of my favorite things because maybe I can show you guys. As we've been doing these generations, this is wild. the iPhone app has been sending me live notifications about progress.

1:02:50Tom Occhino:So the iPhone app, not only does it exist, it's connected to the experience. So I, this is what I do all the time. I have a meeting to go to, right? I fire off a generation. I lock my screen. I go to the meeting and I monitor progress on my phone and I get back to my desk and I'm like, all right, this is good. and now I get to keep cooking. So definitely check out the iPhone app. I was getting those when we started this call from this computer. Could you, I interpret that to mean, that's great. I'm glad you have an iPhone app. That's awesome. I interpret it to mean that he's like, could you make an iPhone app?

1:03:28Tom Occhino:Could you make an iPhone app with B0? Oh, I see, I see. Maybe he meant what you said, but now I'm just curious. Yes, yes. Somewhat ambiguous, the question there. But right now what we're focused on is the web. So we're focused on you can make apps that are mobile web apps that are very, very friendly on mobile. And almost every generation, especially if you ask for it, is responsive and will work on mobile. But no native code generation yet. It's on the list. We're thinking about it. Cool. Very cool. Very cool. It looks like those scripts ran. And I had another question, which was, you know, a lot of times with these builders, sometimes like the context will get really long, like with V0 and that could mess it up.

1:04:13With V0, do you ever have to reset the context window here? Like pull the code up in another? How does that work? Like start a new chat or something? Yeah, that's what I'm trying to say. Yeah.

1:04:23Tom Occhino:I mean, you know, I haven't. I keep all the context in there. And I think we're sort of speed running up against the ability for context windows to get larger and us on the V0 app team side to compress that context into only the necessary stuff. So I don't do that. But, you know, your mileage may vary. I think it's definitely something you can do. Again, like you can fork any chat at any point and then start with fresh generation and context there. And I guess my overwhelming sort of like guidance for folks is just kind of tinker. just play with it and explore the edges and Corey I know from the first time that you and I spoke to now you've just been like I haven't told you anything I haven't guided you on how to use it or what to do you just kind of have been playing with it and and that I think is it's so fun it's just building fun um and I'm not a builder that's what's fun is that uh you know I'm trying something new and uh and i'm excited and i'm learning as i go uh it's absolutely i think it's it's a gift from god but the truth is it's probably hot garbage because it's the first thing i've ever tried but uh not not your app my yeah you might want to clarify my apps are all garbage too but they're my garbage but i will get better and they will get better and i've learned so many lessons like You know, when I first started, the first thing I went out and did is I went to Claude and I went to Opus 4 or 5.

1:05:52And I was like, I got an idea for an app. I want to build this thing. I want all of these things in it. Give me something I can go feed to V0 that will help it understand. And it gave me 3 ,000 words that I threw in without realizing maybe that was a little much to start with. So I started there and I was like, man, I think maybe I bit off more than I could choose. So I went back and I was like, break this into chunks. so uh i actually know i went back and i was like is this too much to feed it one problem it's like oh yeah absolutely and i'm like well why didn't you tell me that but you've learned you've learned and i had to break it into like 25 steps and i came through one at a time one at a time one at a time and uh just like clockwork they fell right over and uh suddenly i was i was getting integrations that worked with like, you know, Google and Microsoft and these other tools.

1:06:48And it's very much, I like that you mentioned like having a sidekick. Like I feel like I have a friend who's, who's a hell of an engineer and I sat down at the table by him and I'm like, here's what I want. No, no, I don't want that. Do it this way. And then it's like, oh, well I want it green. Let's do it green. But instead of annoying an actual developer with all of your indecisiveness, you can get it out with an AI.

1:07:12Tom Occhino:These are very understanding. I love it being, you know, people say, oh, it's like having a junior engineer. It's like, I've been in leadership for a long time. It's a better engineer than me in most cases. And it doesn't get annoyed by my stupid questions. And so like, it's quite fun. You know, that's just like make development fun again. I think this enables you to focus on your idea and not focus on all of the sort of like, I don't know, nitty gritty, like wiring things up, et cetera. So my advice is tinker, play with it. Give us your feedback. We want to hear from you. What could be better?

1:07:43This was great, Tom. I'm so excited about this. I think this is exactly the tool that people need to get started and actually get their stuff out there. So thank you for taking the time to show everyone this. Seems like everyone's been really loving it. And I will be annoying you with my questions from here forward, man.

1:08:00Tom Occhino:Well, you know, I'm going to look forward to that, Corey. I'm not annoyed at all. Guys, this was fun. This was my most fun meeting of the day. Thank you so much for having me on the show. And thanks to everybody who tuned in. I hope you are able to tinker and play with V0 and Brissell, and we'd love your feedback. Absolutely. And where do they go find it while we're on here real quick? V0.app. V0.app. See, I would have said.com or.no. That's important. Well, Tom, thanks again, man. Keep it all up, and we'll do this again sometime. Sounds great. Thank you, guys. Thanks, Tom. Take care. Bye-bye.

1:08:38before we jump off here you and i cory let's just mention that tomorrow we are going live with logan kilpatrick of google deep mind uh the gemini flash had a flash release so everyone make sure to tune in at 9 a.m pst tomorrow uh i tried to include someone someone requested we're doing a survey right now and and we're asking for your feedback on how we can make the neuron better in 2026 and Someone requested include London time on the live stream, the live stream notification. So today I did. It's 5 p.m. London time. Yeah. So going global. Yeah. Anything else? Do you want to get off? Or what if I brought up my Vercel and we took those three RSS links I got and saw if we could build a neuron app right here?

1:09:29Let's do it. Let's try it. um i'll i'll just excuse if if anyone on our team needs to go they don't have to wait for us yeah absolutely no need to wait for us if you guys want we're gonna uh we're gonna we're gonna get silly here yeah let's do it i'm excited so you're so you're saying we're gonna start from scratch on on v0 and yes well well sort of scratch we're gonna we're gonna use mine uh and burn burn my credits which is fine which is fine i'm glad to let me uh give me just a second here to get things up yeah let's do it um right yeah so uh anyone else if you have questions things you want to see i know there were a couple suggestions for apps to build earlier um maybe you know if you want to follow along you can go to v0.app and try it yourself a couple things i wanted to highlight that really stood out to me during the presentation the first one is the enhanced prompt tool I think every chat app should have an enhanced prompt.

1:10:31Every one of them. It's 2025, dang it. It's time to have one. Yeah, just like it's right there. You can click it. Then people don't have to worry about prompting it the wrong way. I mean, yes, you're still going to want to give as much context as possible if you want to build something really specific, if you have features in mind. Another thing we saw Tom do that I thought was really cool is he used Whisperflow to just talk to it. so let's say you don't know i saw that i i caught that i i didn't mention it i should have no free ads whisper flow does advertise with us um so just to be clear i don't think but they've been with us a long time yeah not a free ad but it is a cool tool um uh and yeah so if you don't know the features that you want to build just ramble and chat it doesn't even have to be in v0 maybe you take it to chat dbt or something so you don't burn credits on it but just say like hey this is what i'm thinking about i wanted to have these types of things i don't know what features i need but this is the sort of aesthetic that i want and and and and then workshop it there then bring that into v0 and use the enhance prompt with all of your rambling and have it fix it for you all right we're live we're in here i want to build an app for the neuron that brings all of our sites together.

1:12:05You might notice typos in the chat. AI will figure it out. Yeah, I don't even care about typos anymore. I just assume there's somebody coming behind me cleaning up my messy, messy typing. The blog, newsletter archives, and our YouTube channel slash podcast should all be included.

1:12:40let's go from there here are the RSS feeds I have one from Beehive as well I'll send that to you in the chat okay excellent pull it out that'll pull in the actual newsletters yeah it should okay newsletters and I have the blog Neuron blog and

1:13:11brand aggregation. I like that. Lumen. Neuron YouTube. We're going to see if this will work. We might be taking, okay. First, we're going to go down here to the nifty button. Oh, I have 22 credits left. We're good. Enhance my prompt. Is it one credit? How does the credits work? Actually, is it one credit per request? uh no uh essentially you'll see that like it shows you what they cost here right so like you know you pay your monthly fee and then you go buy more credits and your credits are just a dollar credit you know they're comparable and uh you know if you know i want to use this and you're like i'm going to need a few more million tokens you can just go right up here and come down you see gifted credits, monthly credits, purchase credits, and right here you can buy more credits.

1:14:05Buy more. So I have the$20 subscription, but I did buy an additional$30 already. Lumen, we'll have to have you on and do that at some point. Lumen says, happy to hop in and share screen and co-brainstorm through this. Hashtag just saying. Another time perhaps. I think that's a good idea. We'll have to technically figure out how to do that. It should be easy, but we won't burden our tech team behind the scenes with trying to do something new today on the fly. And let's see what she does here. Show full message. Okay, I got those in here. Oh, this is jammed. No, that was supposed to be a new line.

1:14:50Oh, this is too. Maybe that's just how it's appearing. Okay. Did you hit the prompt enhance or is this just raw? I did prompt enhance, but then I went back and put the RSS feeds in again. Okay, smart. Yeah, because I can see it not figuring that out with the links. Let's see what happens here. This is fun.

1:15:13Is the neuron across its channels available via RSS? Maybe. We're going to find out if these work. Yeah. We don't know if it's publicly available. I can answer this one too from Tom. Sorry, not a dumb question at all. If you don't know how many tokens you're requesting, what I do, what you should know is that it will just pop up and be like, hey, you're out of tokens. Would you like to re-up, grab some more? uh the truth is i i've been doing some really really big generations and you know a million tokens goes a long way and i spent longer than i should have on on vo max or it probably would have been cheaper uh so because there are definitely things that you know you might want to do on a smaller one as opposed to you know just like building an agent you don't always need to use the latest greatest frontier model for every step of the tasks i'm tasked you can use something smaller and uh and if you don't know maybe like your first attempt to use the the smallest one and see how well it does yeah it'd probably be my the mini yeah i don't know what i'm gonna do is go back because i realized that i changed this and when i started having trouble with with the app i've been working on because like i've been working on one here for a week and a half two weeks well A couple of weeks.

1:16:42And, and it's been going really well, but then I hit a, hit a hurdle in the last two days that I haven't figured out yet. And I'm going to go back and branch off of it later and see if that helps. But I'm very much trying to do it without asking for help. Like I want to, I want to see, can I do this? I like it. I like the challenge. Also something that if, if you're someone who's using these tools and you're not a traditional engineer, I would also, whether it's asking V0, So, you know, mini or taking the code. I think you can download it. I'm pretty sure. Well, once it's on GitHub, you can ask questions on chat.

1:17:20Okay. Step one. We'll see what happens. Yep. That's it. By golly, George, there is today's neuron. All right. That's a good start. Yep. This is looking good. It's looking really good. Yeah. Let's go over here to blog. All of the neuron in one place. Okay. so right now we're not getting the articles traditional engineer okay so here's what we are getting

1:17:49Tom Occhino:traditional engineer as they call it and the videos didn't come in okay okay so what we're going to do is we're going to come over here and say good work just in case the models uprise and one day come to eat us it tokens well spent appears the The blog feed didn't come in. Here is the RSS link.

1:18:20I think those are the two of the RSS links that wedged together. Yeah, that might have thrown it off. That makes sense. That'll happen. Definitely going to go to design in a few and change our colors over to neuron colors. We'll change them. I don't know. We're not going to change it to tan in the background. Please don't make me change it to tan, Grant. Corey wants a rebrand away from tan. It's funny. I love everything about the brand, but I hate tan. So many brands have like tan now. It's funny. Tom asked about how many tokens would a hobbyist vibe coder, asking for a friend, need in a month, like for purposes of clearing the cost at family budget time?

1:19:03If you want to run on VO Max You could spend a decent amount of money Over the course of a month Are we talking over under 20 and over under 100? Under 100 over 20 Under 100 over 20 Like what I'm building is reasonably comprehensive It's truthfully bigger than what I should be trying to do For a first attempt and I went through 20 over about a week, but that was me using it every day and, and making constant changes. You know, I don't know how many versions I've done now, but it's, it's a lot, a lot of them. And okay. So here's, here's good news. The feed is inaccessible. Oh, hang on. Now I understand the full flow, maybe returning a different XML structure.

1:20:01Yeah, and I opened this one. I know it worked. I've significantly improved the parser. The parser now handles C data sections better, extracts links from multiple formats. Okay. Let's see. I can probably just copy it from... Let's see what happens when I go here. Grant. What happened? Bam. Bam. Oh, yeah. These are our articles. Yep. let's go over here and look at sam's snazzy yearbook photo

1:20:35this this this uh all of the collar newsletter did not get enough love sorry no i agree that was a good one but okay so we are pulling in our explainer articles excellent now next up let's go to videos 200 plus for daily and here's the live we're on right now hey there you go we have live section. Does it actually go to the YouTube channel? Let's try it. I'm not going to try on that one. Let me try. Oh yeah, we should promote this. Guys, tomorrow Logan Kilpatrick's going to come back on, our good friend who comes around once in a while, and we're going to talk about Gemini 3 Flash tomorrow at some time that I don't know.

1:21:229 a.m. Pacific, 5 p.m. London time. 9 a.m. Pacific, so that's 11 central. That's 12 Eastern. That'll be a fun one because Logan's always a great guest. He's good people. Guru demo. Okay, so we're also getting all of our shorts, I think, which is maybe not desirable. We can figure that out, though. Oh, so this is actually going to the YouTube channel. Let's just click out to one and test it. Yeah, it should be going directly to our videos. Perfect. so like we got cleanup work to do wow that's

1:22:05wild this is youtube inception right here is this is i kind of want to click this but i'm afraid it'll knock everybody else off and we still got a decent number of people hanging on podcast did it separate no okay okay we have a feature request so in videos it's also pulling in our shorts. What if we're like

1:22:29blog and YouTube are connected?

1:22:35Please don't pull in YouTube shorts as they look funny on the page. Let's also I don't know if we want to do these as two separate prompts, but we have a feature request. As a user, it would be nice to semantically search, filter, sort, bookmark, pin across any given term and or media format. I love that. You want to just copy that request verbatim?

1:23:07I just typed it into the box there so I'd remember when I get there. Thanks, Rachel. Appreciate you staying on. I know you mentioned you had a meeting.

1:23:17By the way, anyone, if you do need to drop, feel free. this will also be a recording and we'll share a blog sometimes it takes us a while because like i said it's cory and me who write everything so so if it takes a minute it's just because we have other higher priority things to write first but we will eventually turn all of our lives into blogs at some point oh we have oh hang on video content vibe coding with versell look at that it got rid of all the shorts we filtered out the shorts okay so i'm going to add these a couple that's good A few small requests. One, bring in more of the YouTube archive.

1:24:04Yeah, I wonder if there's a limit. But still keep the shorts out. Nice. include the following features for all posts across the site. Thumbs up in the chat if you want us to actually release this. This is useful for you. We will publish it if this is useful for you. Oh, man. this is great how cool okay so also I notice okay so that took us to video content good that's that that took us to video podcast is going to video blog appears okay so the header links are not doing anything so we gotta fix that also you should know if it catches a bug you get a little notification up here and it'll say have a bug fix and it'll also save free it'll fix its own bugs that it detects at no cost you don't spend tokens on it right how cool is tom though is he dude that guy is awesome yeah i it was great meeting him he did an amazing demo uh he was so fun just to hang out with and pretty much answered every question we threw at him i think we met at aws reinvent two three weeks ago and and had a long talk about this stuff and kind of knowing that this was a thing that grant and i really wanted to do we wanted to bring vo on because you know we're always hearing good things so we we coordinated it through their pr people and all of the stuff and got to talk with tom figured out how we wanted to do this and that we wanted to kind of just live do something that the audience sent oh hello what's happening here things are occurring grant

1:26:07now i have a complete picture i don't feel like you added features i feel like okay oh it's loading hang on it's funny the more that you do this the more that you start to get a sense when it's gaslighting you. You're like, I don't think that that's, I don't think you did that. But then as the models get better and they're faster, it'll still do the things, but you'll have your suspicions up and then you'll be surprised to see that it actually did what it said it could do in the time it took. Okay, so watch this. Walk us back to V3. Something broke. Yeah.

1:26:56I don't know if this is helpful, but Lumen AI said all YouTube archive, they're just deep links. So perhaps if we change the language, that could help. Quite possibly. Yeah, get rid of the live at the beginning. We're also going to, I hope By the first part of January Be taking these lives And publishing them as extra content On our Spotify and Apple podcast Channels as well Just to kind of help us over there Like our Spotify is pretty strong Our Apple podcast Has a way to go, so if you're on Apple And you wouldn't mind going and chasing Sound and listening to a video Subscribing, we'd really appreciate it We keep flirting with their top 200 we've been as low as high i guess as 103 but uh we just can't seem to top the crack 100 and you might make the difference we have some amazing guests coming up too like i think everyone that we've talked to in the past two to three weeks that we haven't published yet are just awesome and i can't wait for them to get out um check this out reverted to v3 v5 oh god we have so many we've done at least 15 interviews in the month of december we have more content right now than we have ability to get it out and uh some of them are so cool some are so cool i would by the way if you haven't go back and check out we had a video from aws that i liked so much that was just a short conversation with stefan armand stefano armand from inception labs and uh he is such a a nice guy he's a stanford professor but uh you know essentially they're using diffusion models as llms so uh what they're doing is uh super neat and it's a different approach and it is lightning fast uh and really cool to watch so you know that's one that i would love to see get a little more attention that hasn't yet um also grant we should tell them what we're going to do on the 30th of December.

1:29:05Yeah, for everyone who's stayed on this whole time, y 'all are the real ones. So we'll give you a sneak peek of what we're doing. On December 30th, it's a Tuesday, we are going to go live and do our predictions for 2026. So everything that we think is going to happen based on where we are in the industry, we'll be doing some research and trying to, or at least I will be trying to get it right, which is my modus operandi. I don't try to overhype things. I don't try to look like doom and gloom too much. I try to be realistic as possible. And so that's what I'm going for. If I'm hyped up and flappy about something, it's because I genuinely like it and I really am excited.

1:29:51Like with VO, it's a cool tool. Okay. Yeah. Okay, so look here. Reverb to V3, still made it to V5. So let's go over here. Something I should have done to begin with Is Neuron content aggregator Senol's 1980 Create Okay, I wanted to get it Saving these in case I want to play with them some more But I just wanted to show that We're figuring this out too It's really

1:30:26It's been a lot of fun It's challenging It can be mind-numbingly frustrating at moments, but it's doable. And it took a lot of the intimidation of it out for me. Okay. Do you want to try one more thing and change the design here? Let's change the color to we need an orange, of course. Can you give it an exact color code? Yes, you can. I sure can. You got one? Yeah. What is Neuron Orange? I can't remember. I can't either, but it's in Figma. I always just go to Figma. Because that's where you should store codes, right? I mean, color codes, yeah. That makes sense. Makes sense. Okay, there you are.

1:31:16Right? In the private chat. I like that you put it in the private chat, even though I'm about to paste it here on screen in the middle of the line. I mean, it's just what the window that... Oh, we're looking slick. Okay. We're looking slick. I wonder, can you upload images, right? Could you upload a couple images of Oli and tell it to spread this throughout? Oh, hang on one second. I did this wrong. Go back to black there. Then let's come down here to secondary, and I'm going to make it a little more. Oh, no, no, not there either. you know what's another thing that i i was trying to find the background oh i wanted to do the background in like a

1:32:07like a dark gray maybe for the moment i don't know man we could change this however we want i'm just noodling around to show how easy it is and like you know it goes more than colors colors Just where you start Because you've got fonts You can change You've got The ability to put big shadows on things Or small shadows No shadows What else is in here A radius you can put on things I like the brown Grant, what did you think of the WinApp thing I sent last night? The one that you created? The WinApp meme last night. Does anyone remember Winamp? The old school media player on Windows? Somebody in here remembers Winamp.

1:32:56I know you do. That was a flash from the past. What a sexy app. I had like tons of skins. All of these things. Can I create one or use an existing export? Brand design tokens. Nice. Call cancel. Found free time. Enjoy social interaction. on creative that's awesome i'm so glad you're here and and that you've been so participatory today participate that you participated it's a it's a lot of fun to do these and it's so much more fun when we get good back and forth yeah cory i just shared you some cats my question is could you upload the cats into the chat and ask it to put them on the website sure so i send it to you on slack okay am i in the right conversation yep okay let me download the cats here just a couple of your favorites let's see which one we want oh this one winky cat yeah i was thinking about that lumen actually like well my question that i want to ask tom and maybe i'll email him is how how easy would it be to create and mock up your entire UI in Figma and then connect to it via the MCP.

1:34:13I assume it's easy, but I wonder where would it be better to start? Would it be better to start in Figma and then bring that design over to VO? Yeah, I'm curious. Because the Figma MCP is really awesome. Basically, at this point, you can design and mock up your whole front end, and then it will create the code for you for it. We really should have Figma on and show that, actually. Yeah. Didn't we talk to Figma? Did I dream that? No, not yet. We could potentially talk to them next year. I think that's likely. Oh, yeah. Yeah, they're definitely someone we can reach out to for sure. I think that's an interesting tool that I just don't use.

1:35:07oh yeah that's great i i know like ux people love it you should you should actually test it out i know who designed like user experience and things love it and the ability to have a ton of them right together that's correct lumen so lumen said mcp are token and tool call inefficient great to set up not for ongoing use yeah i really want to do i mean at some point we'll get somebody from Anthropic on the show. But I really want to do a deep dive explainer on skills because I'm seeing more and more that skills.md files are replacing MCPs for a lot of use cases. And so it would be great to kind of explain those.

1:35:50What I really want Anthropic to do is take skills and make it a core part of the desktop and windows a uh ui of claude because it would be great if you could just spin up skills in your chats um but at this point it's not a core part of the ui yet so i i really wanted to do that yeah start in figma if you're figma native yeah yeah well i was gonna say cory one thing you could do is you could use figma make which is a chat interface and you can just ask it to essentially like design for you yeah that's a good call you can do it that way uh okay so the issue is the image was a ping and the background is too dark i said hey his mouth is black now instead of white uh with instead of the chat mobile being white with black text it's just all black and gray and it's like oh that's because it's a ping you didn't fix it You just said you fixed it unless it's still going to reload.

1:36:56No. Well, you know what? It just needs a white background because the problem is that it's see-through. It's transparent. Yeah. So that's why. Yeah, it just needs a white background. It's easily fixable. Just swap the image out. Let me. What time are we? 136? Yeah. All right. We still got. Looks like we still got 20 people on. Okay. Y 'all are awesome. Thanks for hanging in there Even more if you look at What we see through here Which I guess is including Live on X and LinkedIn as well Very cool We're going to do more of these lives These go really well And I appreciate that you all come here Again if you haven't Please take a sec to like and subscribe It really helps us out And lets us convince the powers that be To let us keep doing these and bringing in cool new guests.

1:37:49We have a lot of cool ones lined up for next year. Tell it to follow WACG guidelines for light and dark mode. Lumen OG. Oh, heck yeah.

1:38:09Yeah, that's interesting. Create a simple MCP to serve skills anywhere you want across environments. Yeah. That's cool. that makes sense you know what i haven't created an mcp yet i have used mcps others have created but uh that's a thing i'd like to like us to learn and uh i'm hesitant to yeah i'm hesitant to give people the advice to go create them because they're easy which i do believe but i'm not confident enough in my own ability to secure them or other people's ability to secure them to recommend that you just like go off and build one because you know if if for example someone was to connect to one that you built and it wasn't properly secured like a bad actor for let's say a lot of access to things so so i'm i'm nervous about uh suggesting that people go create them on their own um that's fair that's fair but you know what i bet a clawed opus 4.5 or um um Gemini 3 Pro or GPT 5.2 would probably know how to guide you to secure it the right way um but yeah so we we should definitely do it as a demo another time I'd love to have someone who has built a lot of them come on and and yeah yeah what happens if I hit subscribe oh well we want to fix that don't we oh i think the problem is the www isn't it try it without i think maybe perhaps the subscribe button just doesn't work on there yeah truth is it should be able to just ignore www by now well and it's actually just the main page it goes to isn't it it's not a slash subscribe page in the first place i don't think yeah securing services like api handlers yeah exactly yeah you just need to make sure you set it up in the right way excellent i've uh i've only ever started with an mcp server that somebody else has built and edited it and made changes to it i find that that's easier so if you find someone has created an mcp server for uh that's reputable on github for a program that you want to use that's usually a good starting point to copy that duplicate it and then work with a command line agent to basically just start editing it for you and say, hey, I want to add these features.

1:40:37I want to do this and that.

1:40:41Let's check out Vercels AI gateway. What do we have in here? We've got fast one. On it, 4.5. Disqualified. Not a lot built yet.

1:41:00okay mini max i've got glm oh that's not a sort button okay opus 4 5 gpt 5 2 gemini 3 pro preview nice glm 4 6 oh cool yeah that's a new one that came out mini max is up there too i'm not seeing three flash it's only been out 24 hours but oh there it is you are now jk man they got everything yeah they do have everything don't they yeah this is really cool this is the model list through the ai gateway so yeah yeah so like if you want your app to do you know have ai features for example you would you would run it through their ai gateway oh bam uh this is where your variables get stored if you want to add a variable you just come in here and you tell it like i want it in i want it here on the production side or wherever uh api key value if you want to add more you can add more just click add all safely and it stores it all right here securely where it's easy to access project rules you can go in here and customize those as well anything you want to do uh and settings so like you know here's the settings for this app not the settings for your whole verse your whole vo account v0 account but there's the github if we want to dome here's this domain but we can go by our own domain and just connect it right here ah yes i was that was a question i wanted to ask but but i didn't want to waste too much time with it yeah like obviously when he spun up the deployment it had a versell.sh yes url and obviously if you are creating your own website you want to connect it to your own domain yeah yeah it comes with that you can also download everything that's great download zip you can run this command in your console to add to code base come over here and you can see everything it's there's the kitty that's what it's supposed to look like in the corner and you can come through here and peek through everything see what you like what you don't and then you can obviously go in here if you know what you're doing and you can edit the code yourself yep and you know something we didn't show that i will show real quick hang on uh while we're still here let me go

1:43:40is it.com i think it's.com there we go and here's what the back end of your versell looks like where you host it like i've got one well there's this and i've got this other one tied here uh but you're able to look through you can do it at the project level as well but you can go through here and see what you're working with. See any deployments. You might want to not and scroll back. Yeah, there are definitely some of these I don't want to show.

1:44:13Ship faster with... Oh, wait, I should... Yeah. Yeah, there's actually a couple things I shouldn't show there probably. Like my keys. I'll go back and watch this later and make sure I'm safe. I don't think I shared anything fully, but but that's a good call just to be safe just to be safe yeah just to be safe not that i'm worried about any of you but you never know what kind of sloth might scroll by this video

1:44:40i'm just imagining flash from zootopia with eyes going wide i can steal those i can steal those yeah um one other thing i wanted to share that is good for i was saying this earlier if you are not a trad engineer and you want to learn what all of this code you wrote does, what I would do is I would download this and I would take it into one of your$20 a month subscriptions, whether that's ChatGPT or Claude or Gemini, and chat with it and ask it questions and have it go line by line with you on functions and ask it, what does this do? How does it work? What does this function like? And actually try to actively engage your brain and tell it, I'm a new engineer.

1:45:34I'm trying to stay in the proximal zone of difficulty where I want to learn things that are just difficult enough building on my skill set in order to help me learn a little bit more. and really try and test yourself and to not just let the AI code everything for you, but also build your understanding, not only of the application you're building, but of how all of this works. Yeah, because the more you learn, the better you'll get, the faster it will get, the less dumb things you will do. Like right now I'm doing dumb things that I don't even know I'm doing. And that's part of the problem. But I do a lot of this.

1:46:17and it does a really good job of breaking it apart chunk by chunk and being like default border color uses your token outline color uses this uh and it'll it can break it down more granularly but you can always be like you know you know uh explain this better or what i like doing is i like saying okay i get this and i understand this but i don't understand this this this and this and actually try to think through the limits of your understanding and try to repeat back to it what it just taught you. Like, okay, so you're saying that, you know, in your own words, okay, so you're saying that this works, like for instance, the single line tells you your build tool, run Tailwind's compiler, inject all the CSS Tailwind generates, base styles, components, utilities right here.

1:47:08It's not importing a static CSS file. It's like pulling a lever that starts to factory. Like, okay, so what you're saying is that it's bringing in all of the components that I need across, you know, X, Y, Z. And then you're sort of like building your understanding of it as you go. Yeah. Yeah. And then it's like, okay, what actually happens? It's doing this, names it finds in your project files, conceptually expands into something like this. You never see that output directly. It just streams into your final CSS bundle. why this is not the old not the old tailwind syntax okay see all of this I would probably be like you gotta go even further up the stack of my understanding here Eli 5 dude Eli 5 and you can be that got it crayons no compiler goblins

1:48:08that's funny oh my god imagine your website is a lego city right now your css file is just an empty table how does that have to do with legos um this line import is you dumping a giant box of lego pieces on okay all right i'm following not a finished city not instructions oh that's the classic pattern just all the bricks you might need what's in the lego box when tailwind sees that line it It gives you blocks for colors, blocks for spacing, blocks for layout, blah, blah, blah. Okay. But here's the trick. It only hands you the blocks you actually use. No unused junks. Why this isn't a normal import.

1:48:47This one means, hey, Tailwind, wake up, look at my projects, figure out what I'm using, and build the CSS for me. It's more like summoning Tailwind than importing it. Tailwind, I choose you. I love the wizard. It knows me too well. yeah that's funny cory at some point you need to share your some of your uh idiosyncras idiosyncratic um memory blank like always talk to me like xyz with people so that they can have a version of gpt that talks like cory's uh johnny five like cory's johnny five i uh it's a little bit more oh not yeah it's a little bit thinner than it used to be but like right now i've set my custom instructs to say take a forward-thinking view use quick and clever humor when appropriate be talkative and conversational be innovative think outside the box challenge my assumptions i like that and then you go and you fill out your about you you tell it a little about yourself you can let it reference safe memories if you don't know you can edit all of your memories that chat gpt has about you just like that you just go in and delete them you can add them and also if you say hey commit this to memory it just goes and throw that's where it throws it if you wanted to go back and get it later yeah it grabs a lot of crap you won't need don't want probably but that's okay uh record mode i don't know what advanced is i haven't looked in here in a minute okay uh long as everything's turned on i'm happy yeah but yeah it's it's really simple and like like it cusses at me you know mine mine on a case it regularly says hell yeah like it listens to me talk or something and uh it will now only uh do i say tastefully maybe not tastefully maybe that's not the right word but all in all though uh you know it's talking to me the way i talk and it's talking and it's putting things into like you know it used the wizard joke because it knows i'm into dnd and uh so you know it's teaching me in a way that i will relate to it's using you know analogies that that cory would would get yeah lumen lumen has some interesting uh suggestions here they say use e-prime language more explicit and less babyish i shared my version uh which is be direct answer with 100 fidelity to the technical detail but without jargon so i can follow without knowing the terms.

1:51:24Oh, I like that. Yeah. But then this framework is kind of interesting. AQDCA, awareness, question, discern, choose, act. Ooh. So I'm going to copy all of this out of this conversation here. And when I sit down tonight and nerd out on my free time, because I literally do the same thing when I'm off that I do when I work. I am going to go try those. I'm always so leery about, you know, we had a live stream a couple of weeks ago. Someone was like, dump your memory. And I was like, back off. Don't you get on my memory. I need a dummy account for demos is what I learned that day. Yeah, that's funny.

1:52:16One more question i have for you before we jump off here how much how many tokens did you use in this whole this whole ordeal i was at 22 i'm now at 19.63 and that was running max the whole time wow yeah so we just we just literally built a blog that that i mean like you could go read uh it's not perfect and it still needs a lot of work but you know for sub three tokens max um it's amazing That's not bad. And you know, if you want to go buy more credits, I'll show you. As long as it's not going to flash my credit card up. I'm not sure if there's anything worse I could do in this video. Just plain English.

1:52:56Yeah. It'll show you. Buy credits. There you go. Yeah. 30 bucks is$30,$60,$100. And you're basically just buying API credits. API credits. Yeah. What is this? Operation Varsity Blues over here?

1:53:16buying API credits. Yeah. And there was a lot of ways that you could do that. You're not actually using an API, but I mean, you are, that's just, that's what these numbers are. Yeah. So there's a lot of ways you could be more efficient with that. You could go to chat to BT and brainstorm some of the things that you want to ask and, and get it down to just, just the exact thing you want to do and then bring it back over here. You could use the lower tiers like mini and the middle one, you can even start from a template. We didn't talk about these. That's why I was scrolling down. I was like, you know, I don't think we ever talked about it, but you could.

1:53:53New feature. Hey, you can now ask clarifying questions before it starts generating. Oh! Oh, V0 can now ask you clarifying questions before it starts. I love that. You saw it here first. You know what else? If you go to v0.app slash templates, you can see all of the templates. And anyone can do this. Basically, you can go here and find a good starting template and work off of that. Or just get inspired. Yeah. Yeah, go break something. Get out and landing pages. Games. We need games. Yeah, we've done nothing. Check out that app I sent you yesterday. The game agent. How was it? uh very cool very it was sort of like if you'd like to spin up your own mobile games right away yeah one for you is kind of the vibe i got yeah yeah which i think there's merit to that i i like working with my uh mcp server with godot which is the platform that i use but but i've been thinking about even trying to make my own video game engine from scratch perhaps working Godot and changing some things.

1:55:10I don't know. But it's a long, that would be a very difficult project. Logo maker. These are so cool. Everyone, thank you again so much for joining us today. We're about at time, but this has been a lot of fun. I hope it's been helpful for you that you've learned some things along the way. And for those of you who stayed on after Cool Tom left, thanks to Tom again, by the way, and Vercel for letting us have him for a little while. Yeah. For those of you who stayed on, I hope the rest of this has been useful too. We're kind of just stumbling around and trying things and learning. And to everyone who had advice and tips in the chat, thank you for teaching us.

1:55:52Perhaps that's what this channel is. Lumen, shout out. Lumen, Rachel, Tom, everybody had ideas. Yeah. There may be others, but. There was a couple ideas we didn't get to. so we'll try to build those in future demos. Perhaps Lynn asked for an app that helps people categorize their business expenses for profit and loss statements. Maybe we can get Logan from Google DeepMind to do that tomorrow. Maybe, maybe. That's actually not a bad idea. We'll see, but maybe we can build something with Flash. I don't know. I assume Flash can build. I don't know why it couldn't. Yeah, we'll ask him what he thinks the best use cases for Flash are.

1:56:32If you still haven't, if you're still here, You probably already like and subscribe, but just in case I'm going to say the thing, please do. It really helps us out, helps us build both the channel, the newsletter. If you don't get the newsletter, visit the neuron.ai, sign up today, and join 600 and some odd thousand people who read it every morning. We'd love for you to be one of them. Otherwise, we'll see you tomorrow with Logan Kilpatrick from Google, and then back on the 30th of December for Grant and I to share our predictions that we hope don't break before the clock strikes on new year's yeah ai rolls buddy oh man should we do should we do a spread of like a percentages likelihood or like good good or bad format like i've got a bunch together some of which have already been busted a couple of which have already been passed uh uh that i thought were interesting and didn't didn't expect uh like iq3 yeah oh by next thursday got it um but yeah that's uh that's about it we'll we'll put it together if you have any requests oh why don't i plug this for everyone who's here before we sign off uh we're doing a survey what can we do better in 2026 i'm gonna share it in the chat uh for anyone who wants to sign up uh the first 100 people to enter i think we've already past this so actually i'm not even going to say that um but basically we have a survey that is asking you what can we do better in 2026 what types of additional content you want to see from us what do you want to see less of yeah please definitely if you have some time and it's a lot of questions about um like you know some things that that have been considered to add over the next year or some things that are, you know, maybe yeses, maybe noes.

1:58:30We don't know yet, but we're, we're just kind of figuring everything out as we build this, this ecosystem of Neuronians. Exactly. Exactly. And your, your input will literally define, define and decide what we do next year. So definitely if you want to, if you have a stake in the neuron and you want more content, different types of content, definitely fill that out. Cool. Well, to everyone who's watched, thanks. It's been a total blast. Thanks again to Vercel and everybody who made that happen. Thanks to Eunice who helped us out here in the back end and made sure things happened the way they're supposed to.

1:59:07And on that note, farewell for now, humans. We'll see you tomorrow. Thank you.