The video discusses the emerging trend of Vibe coding, a method where both technical and non-technical users employ Large Language Models (LLMs) to generate code and build applications with the assistance of AI models. This approach allows users to provide natural language instructions and iterate on the outcomes to generate web apps or websites, bypassing traditional, manual coding processes. The hosts explore how this trend leverages components like Stripe and Clerk to build more efficiently, and highlight the range of users and companies benefiting from this approach, including developers, entrepreneurs, and small businesses.

Vibe coding is considered a game-changer because it provides opportunities for non-coders to create personalized applications previously out of reach and allows developers to focus on enhancing functionality rather than starting from scratch. The video also describes how the underlying models have evolved, benefiting from a wealth of internet data and the maturation of web-based development environments, which makes the creation of applications simpler and more accessible. Businesses are taking advantage of this tool to quickly generate economically viable projects, enhancing productivity.

Main takeaways from the video:

Please remember to turn on the CC button to view the subtitles.

Key Vocabularies and Common Phrases:

1. iterating [ˈɪtəˌreɪtɪŋ] - (verb) - To repeatedly perform a process with the aim of approaching a desired goal, target, or result. - Synonyms: (repeating, refining, cycling)

Interestingly, some really technical people have been tweeting about how instead of like coding something from scratch, they're just like going to an LLM and saying, this is what I want to make and then iterating on it.

2. authenticating [ɔːˈθɛntɪˌkeɪtɪŋ] - (verb) - The process of verifying the identity of a user or device, often as a security measure. - Synonyms: (verifying, confirming, validating)

In reality you would just ask the coding agent to actually add Stripe as a component or add Clerk as a component for authenticate.

3. divergence [daɪˈvɜːrdʒəns] - (noun) - The process or state of diverging or differing in opinion, character, or style. - Synonyms: (separation, deviation, variance)

Yeah, definitely seeing a divergence of different Persona using different tools and reaching for different tools.

4. latency [ˈleɪtənsi] - (noun) - A delay before a transfer of data begins following an instruction for its transfer. - Synonyms: (delay, lag, downtime)

And the most frustrating thing ever because it gets stuck in some bug or loop that it is not seemingly not able to fix. Or I'm having to tell it like, think step by step and figure it out.

5. ramp [ræmp] - (noun) - A steady increase or rise in activity or condition. - Synonyms: (rise, incline, escalation)

They are saying things like getting to 20 million in ARR in two months or 10 million in ARR in two months, which is an insanely fast ramp.

6. modality [moʊˈdæləti] - (noun) - A particular mode in which something exists or is experienced or expressed. - Synonyms: (method, form, manner)

It's kind of similar honestly to what we've seen AI do for content generation across all of modalities, like in some way you can think of like, I don't know if it's code or website or web app, it's like another modality like AI image generation didn't replace like photography and it didn't replace like artists

7. confluence [ˈkɒnflʊəns] - (noun) - An act or process of merging or flowing together. - Synonyms: (junction, converging, merging)

The confluence of that made generating code much easier.

8. prototyping [ˈproʊtəˌtaɪpɪŋ] - (verb) - The process of creating an initial version of a product before full-scale production. - Synonyms: (modeling, designing, sampling)

Built for engineers who are like prototyping something to show to their team during standup.

9. containment [kənˈteɪnmənt] - (noun) - The action of keeping something harmful under control or within limits. - Synonyms: (control, restraining, limiting)

So as a result it's very good at building those apps. Another reason I actually think is web based developed environment. Now it's very mature. So before you may need a front end which is like HTML, JavaScript and backend like Rust or maybe back in this C NET or something. But nowadays the full stack apps are all JavaScript. Most of the apps we see that are being generated are all JavaScript and TypeScript based. So it's in itself a very contained runtime.

10. emerge [ɪˈmɝːdʒ] - (verb) - To come into view or become apparent. - Synonyms: (appear, surface, arise)

We're seeing more products emerge here every day.

Vibecoding is Here - How AI is Changing How We Build Online

I think this is a perfect time to be building with or building for these tools because the components out there are just so mature. Like now for payments. You don't want the coding agent to build Stripe from scratch, because I know there's a lot of people building Stripe clones, Airbnb clones, but in reality you would just ask the coding agent to actually add Stripe as a component or add Clerk as a component for authenticate.

Justine Yoko, welcome to the show. What is Vibe coding and why has it at least seemingly taken the Internet by storm? So my understanding of Vive coding is that you give a coding agent a set of instructions and you just let it go. And the only thing you need to do is to provide natural language instructions. Like, I like this, I don't like this. You know, this Vibes. Well, with me, this doesn't. So that's how Vibe coding is, I think.

Yeah. And I think we've seen it used by both technical and non technical people. Interestingly, like, some really technical people have been tweeting about how instead of like coding something from scratch, they're just like going to an LLM and saying, this is what I want to make and then iterating on it. And then maybe they're kind of exporting the code to wherever they like to edit it and making their own edits and then deploying it. And then there's also a lot of completely non technical people who are like purely just going on Vibes who have never been able to make anything before, who are like entirely reliant on an LLM to write code for them, which is very cool.

Yeah. And to your point, it's not just the non technical are doing this and then the technical are kind of using this paternalistic term. The term actually comes from Andrej Karpathy. Right. Who is like extremely technical, is behind the LLMs that are actually helping people Vibe code. They're entire Reddit communities behind this. And on the back of that, there are these companies that are being built that help people Vibe code.

So who's actually participating in this trend? What are the companies that are being used and how would you also kind of like frame them in terms of like, are there different categories within this trend? Yeah, so we've, we've seen different companies serve different types of users. So I think there are some more sort of like IDE based companies like Cursor that are targeting developers and just making it a lot easier for them to code by giving an agent a prompt and then it writes code or edits code for you. And then we've also seen a new emergence both for non technical and also being used by technical users of these like text to web app or text to website companies that are in the browser and like you literally go, it's a web app.

You type it in, you say, hey, I want an app to track if my dog has been fed. Or I want a website for my small local business where people can find information and contact me. And then you sort of get this interface where usually they all look quite pretty similar. On the left side you prompt it and then on the right side it kind of shows you what it's generating, what the interface looks like. And then you can say, no, I want this button to do this or no, I want the design to do that.

So we've seen a ton of companies emerge in this category. I, I think the biggest ones we've seen thus far are probably like the Replit, Agent, lovable, bolt and V0 from Vercel. But we're seeing more products emerge here every day.

Yeah, definitely seeing a divergence of different Persona using different tools and reaching for different tools. So for example, as an engineer, maybe my first instinct is like IDE is where I live, so I want to do everything my IDE or vim back in the days, now it's Cursor. But for someone who may just want to build a marketing landing page, they may reach for, you know, bold, lovable and V0 because they generate just very visually appealing assets just for different use cases and what, you know, people are used to from the very get go.

Yeah, and give me a sense of scale here. Right, Because I think maybe Cursor is becoming more of a household name. A lot of people refer to it as one of the fastest growing companies over the last two years. But what about some of these others? Like are people really using these tools at scale scale or are we really in the early innings? Yeah, it's a great question.

Not every company has released their metrics. I think Bolt and Lovable both have either tweeted about it or talked about it on podcasts and are are saying things like getting to 20 million in ARR in two months or 10 million in ARR in two months, which is an insanely fast ramp. And so I think to us that indicates like there's a ton of latent demand for people who want to use these tools or want to code or make something for the first time and are finding a ton of value and being able to do it in a more accessible way. Obviously there's also A ton of demand from developers to code more easily because Cursor is growing incredibly fast as well.

I think it's actually just very qualitatively. Like metrics aside, I think it's solving a very real problem. Like my cousin who wants to say, build his personal app for tracking like his how often his flowers are watered. Like before consumers like this, they couldn't have made an app for themselves. They have to, you know, find an app that works for them. But now, like I think the beauty is that they can create some of this themselves and you know, they can run it on their local machine forever and never deploy it for anyone else. It's like there's so much personalization needs this is solving.

Yeah, software for one. I feel like I'm hearing that more and more where it's like, yeah, I have a friend who wrote a book and he's like, I've always wanted my own personal page to show my book sales and I've tried it, I've tried to learn to code. Just never happened. So tell me more about that because it does seem like, okay, it seems like you could maybe create a static webpage, maybe something like a tracker, some of these more early stage websites, maybe not full blown apps. Am I getting that sense or am I wrong about where these tools are and where they end perhaps in what you can build? Yeah, I think the difference with like the, the vibe coding products versus products we'd seen before to help non technical people generate websites like a squarespace or a wix is that they're actually writing code so they can actually like do more kind of interactive and dynamic things.

Like one example is like even me with Lovable made a web app where you input a book that you liked growing up. It does a call to the OpenAI API gets a response about like three books you might like. Now based on the book you liked growing up, then you can save the book in a database that I like prompted within Lovable and then you can even log in with your Gmail account which like a Google Authentication API if you want to like save books and then rate them. And so that sort of thing I think is beyond like far beyond just like a basic static website. It's like truly a dynamic web app.

That was a little bit harder than just like pure text prompts. It required me going and I mean the LLM would explain it to me. I would say like, hey, I want to add authentication. And it would be like, okay, here's the five steps. But you do have to click out to this external link. And you do have to be able to like copy in these codes and keys into one places or another. But someone like me who's non technical was, was able to do it. And so I think our belief is it's just going to keep getting easier.

Yeah, the components, I think this is a perfect time to be building with or building for these tools because the components out there are just so mature. Like now for payments, you don't want the coding agent to build payment build Stripe from scratch because I know there's a lot of people building Stripe clones, Airbnb clones, but in reality you would just ask the coding agent to actually add Stripe as a component or add Clerk as a component for authenticate.

Yeah, I mean, to that point these tools are pretty magical, right? You can build way more than what especially non technical person could have built in the past. Can we just quickly revisit that idea of the why now? Why are these tools so good? Are we a little surprised even that these LLMs are so good with code? Because that to me would have not been intuitive coming in and maybe just tell us a little bit more about the building blocks that led us to having these tools in the first place.

Yeah, so I guess on the very lowest level of layer, which is the foundational models, all of these tools are powered by very good coding models out there. So why are the models good? One, obviously it's like transformer architecture and everything. But two, I think it's the plethora of data. The data distribution is there on the Internet. Most of the apps nowadays are JavaScript apps. Like if you go on Stack Overflow before AI age, most of the questions people ask are like, why can't this node JS app work? And then you get answers for that. And now there's a lot more frameworks that's very well defined, like Next js, React.

Right. So you really travel up the abstraction encoding. So as a result, when you train a coding model with the data, there's just way more examples of web frameworks, web data. So as a result it's very good at building those apps. Another reason I actually think is web based developed environment. Now it's very mature. So before you may need a front end which is like HTML, JavaScript and backend like Rust or maybe back in this C NET or something. But nowadays the full stack apps are all JavaScript. Most of the apps we see that are being generated are all JavaScript and TypeScript based. So it's in itself a very contained runtime. The agents can automatically verify, like, did I do a good job generating this or is this actually correct? Can they run the browser? So all of that, I think the confluence of that made generating code much easier.

Totally. Yeah. And I think we're seeing on the foundation model layer just to add the companies are really focusing on code as a benchmark. Like we see when OpenAI or Anthropic or anyone puts out a new model, they're generally measuring performance on code related tasks or benchmarks. And the great thing for people like me then who can't code myself, it's like amazing. The LLMs are now learning to code much faster than I could and I can provide some direction or guidance as it's trying to debug itself, but it's able to do things that would take me years to learn for sure.

And maybe we can answer the question as well in terms of how these work. Right. Some people might think that this is just an API call to a foundational model. Is it that or is there more built on top? And then adding onto that, how do these different companies add value right past the integration with the foundational model? And how do they differentiate across the competition? These are actually very sophisticated systems.

When we lift the covers and look into how they could be implemented. So the consumer interface is always like Justine mentioned, there's generation of the code, there's a preview, there's a prompt window. But then when you kind of lift the COVID what it had to do is that it has to have a execution environment for the agent to actually work in. So it needs to pin the foundational model like, hey, my user have this prompt like give me these examples of a code. And then you have to, you know, the agent will have to tell the foundational model, like generate the code only to this, these set of standards. Then once that happens, it runs this code in a execution environment, which is mostly a browser.

Nowadays it's either something called web container, which is a very cool technology that leverages your own laptop as a compute and then spin it up locally, or it's running on a server somewhere so it produces that preview for the user. Then we discussed on the components side, a lot of apps can't be built without a database. So you just have to persist the data somewhere that's either in memory in a browser or it's more persistent in the actual database, like Supabase of the world.

Great. The real question is how well do these work? Where would you say again, the tools perform well, maybe perform perfectly all the way to the end of like, things are really breaking down. Maybe we're not quite there yet. Yeah, it's so funny because to someone like me, it often feels totally magic that you can prompt and like get a working web app. And then sometimes it's like the most frustrating thing ever because it gets stuck in some bug or loop that it is not seemingly not able to fix. Or I'm having to tell it like, think step by step and figure it out.

I think Yoko can speak more specifically to, like, where they, where they go wrong. But I think one of the problems is just like, as these apps get bigger and bigger and more complex, they lose some of the context of like, what they've already generated and the environment in which they're operating. And like, you might ask for one design change. This has happened to me many times with these tools. You ask for one design change and then it generates that new element, but then somehow something in the new element it generated breaks like 10 previous things it did before. And so suddenly like the authentication doesn't work because you like added an image to the top of the page and then you have to tell it like, hey, now this other thing is not working.

Can you go back and fix it? And yeah, and the other thing that sometimes doesn't work about them that is really interesting is the LLMs are always convinced that they can fix something. They're very optimistic. They're extremely optimistic, which is sometimes great, but it's sometimes so frustra when you're on the 40th try of like, we need to have a different approach. This button is still not doing what you said it was going to do. And it's still just as cheery as it was Time two. And it's like, we're going to do this. And you're like, I need to turn it over to a professional at this point.

So I think it depends on sort of the complexity of what you want to build. They can do some things, especially simple things where a code base isn't getting too big really well and really reliably, but they're definitely not perfect. Yeah, this is like a very timely podcast. Just because Cloud just released their latest coding model and then all over Twitter people are like, oh, here's a one shot prompt on creating this app. Obviously there's a lot of advancement in the coding capabilities, but at the same time, I think on the infra layer softwares are always very stable, stateful.

So the question is, where do you keep this data and how do you reconciliate this data? Where is Your auth? Where's your database? Where's your caching? Where's the core logic? I think for that part, it's still a very complex problem that it hasn't done a good job in reducing the problem space just yet. But I'm actually very hopeful in that it's a verification problem at the very bottom in that before we always ask engineers, we commit code and you're like, this is a Shaq link to this list of changes. Now I find myself just keep prompting and then when it rewrites the whole app for the next prompt, I don't even care about what's a previous version as long as it works.

So I actually think the next step for a lot of these tools is adding, you know, how to verify it still works. Is it adding tests? Is it adding something else? There's a lot to be, you know, desired today, but I have high hopes that it's going to get there pretty soon.

Yeah, it seems like a solvable problem for sure. Let's talk about what people are building. I mean, you've mentioned a few examples here, but I'd love to have you both since you've seen so many projects. Kind of talk about like the depth and breadth of all the projects that people are creating.

Yeah, I think it varies. There's an extremely long tail of things that people are building, which is fascinating because you can truly build like an app for one person yourself, which you could never do before. I love that app for one. Yes. It's such a good catch. I know one of these companies should use that.

Yeah. And so on the consumer side, that's everything from like Yoko's cousin's app to track if a plant was fed to like, I need that, by the way. Yeah, so aper too. I need my dog. So slightly different to like, we've seen like a dad generating like a bedtime story for his kid that like always, you know, has his kid's name and interests and things like that. And then it'll weed aloud a different story every night.

So consumer use case is just like anything you can imagine wanting to build, you can create. We're also starting to see more usage on like, the kind of moving up into the enterprise part of the spectrum. So like small business agencies, there's all of these people who were consultants or could be hired to make a website or a web app for like, your local business or any sort of kind of professional service provider. And those folks used to use like a Squarespace or wix or something like that. And many of them are now experimenting with like, replit or Bolt or Lovable, because they can build.

They can make projects so much quicker than they could before, which is very cool. Which allows them to take on more work, take on smaller projects that might not have made economic sense before. And then I'll let Yoko talk about what, like, true engineers, more in the enterprise are experimenting with. Yeah. So I think true engineers actually reach for a slightly different tool.

Like true engineers, they would probably start from cursor, but then for a simple, like, landing page stuff, I think Lovable and bolt and like V0 of the world, they're all just very visually appealing. And the UI is very different because they can render it. Whereas in Cursor, like you kind of render it. As a developer, you're responsible for that. But you just think. I think it's.

There's something we heard that's so interesting in that a lot of the marketing agencies, or just agencies in general before, you know, Bolt, Lovable and, you know, V0 of the world, they wouldn't want to take on like a thousand, $2,000 projects. Nowadays. They can, because they can churn out these projects so fast that, like, it is still a very good quality. So what we're seeing is definitely the sum of the long tails is bigger than the head.

So that's why people are actually migrating from Squarespace wix of the world to this, you know, like software for one business. Yeah, I think that's great. How do you see these products evolving? Right. Because today they do seem quite similar. I think you even called out maybe one differentiator, which is, okay, maybe certain products will be more focused on design.

You also talked about how maybe certain features aren't quite there yet. Where does this go in terms of, like, how you see maybe the market segmenting or what kinds of new features you see these companies adding? Yeah. I think today what we've seen is most of these products are trying to be like everything to everyone. So it's just like a text prompt box. And anyone from an everyday consumer to an engineer, they're like, come one, come all, you can all build something here. And they have very similar interfaces. I think we'd expect to see sort of more segmentation in the features they build and the customer base they focus on based on a variety of things.

Like you could imagine segmenting by, like, technical sophistication of the end user. So, like, if for an everyday consumer like me, I probably, if I'm just making a simple website or something cool to show my friends, I Might even want to make it like on my app like with the new REPL.apple, the AI agent that does it super easily and the experience of prompting and the level of code it shows me and the level of sophistication will be very different or probably should be very different than a product that's like built for engineers who are like prototyping something to show to their team during standup.

And then I think sort of beyond the technical and non technical user segmentation, there's also probably segmentation in terms of like the level of control you expose on various parts of like the website or the web app. Like we've talked a lot about design, how basically most of these tools today you like prompt that you say like hey I want a website or web app with this vibe. And then you can say like slightly different color or move this button and for a real designer it's like so frustrating via text.

Yeah. So we're expecting to see more people add Figma esque features where you can have like true fine grained control over design. Yeah, I think it's so interesting because a lot of people like to talk about like we should empower everyone to learn to code like you know, like now maybe my mom can use the tool to learn to build her own website with code. But I actually don't agree with the view in that I don't think, I mean as an engineer myself who have, you know, worked in the industry for a long time, I don't think most of people should learn to code with how good AI is nowadays. I don't think code is right abstraction for majority of the world and there's just a lot of latent demand there on what's the right knobs for the actual audience out there to build software from.

Is it natural language? Is it drag and drop? Is it like something like Figma like so the question becomes like what should we ship for a majority of the world to be able to build software for one. So it's also interesting because when we talk to you a lot of more consumer perspectives who didn't come from a coding background, they actually have no desire to learn to code. Their goal is to ship what they want to ship, coding or not. And then before it's more static, no code tools like Squarespace of the world nowadays you have more fine grained control on what you want the software to do. So I actually think the next step in a lot of where this is going is like finding what are the things that someone can, you know, do without code. But not as rigid as the previous generation. I think that's such a good point in terms of the level of abstraction because even if you take natural language like English, there are different levels that people interface with that, right?

You can, some people can speak in English but not read or write in English. Right. Some people may not recognize speech, specific letters that people write, but again, they understand a specific message that's being delivered or created within that language. Let's talk about pricing because I actually, I think these different products are so underpriced given the products that people are able to create in terms of, you know, they literally could not create something in the past and now they can create something of immense value in their life.

So tell me about how companies are currently pricing and then how you think about the comparison, right? Like should we be comparing it to the wixes and the squarespaces? Is this a completely new market and where a new pricing model needs to emerge? Tell me about how you think about that. Yeah, the pricing is a really interesting one.

I think most of the companies in this space are doing some variation of usage based pricing. So they'll probably, you'll get some amount of tokens for free and then they'll press you to convert to a monthly subscription where you get a certain amount of tokens. They have different kind of tiers based on the number of tokens you need. You could sometimes buy more if you need them. But the, the interesting thing about that model is like most people, even like technical people are not thinking about the cost of generating a website or web app in terms of how many tokens they need.

Like most of us have no idea. We're like, I'm like, I don't know, is it going to cost 105,000, 50,000? Like, I don't know. And, and some of it depends on like the performance of the LLM. Like if I have to go back and forth with the app a ton that's going to spend a ton of tokens, whereas if it gets it right the first time, it's not going to spend a ton of tokens. And so I think like we would love to see more clarity around the pricing models, like making it easier for people to understand upfront.

Like how much is it going to cost both to build this original website or web app, but then also to maintain it on an ongoing basis. I think the dream for like many products in this space is some sort of value based pricing which better aligns like kind of your incentives with the user of like, hey, if this, if Someone is creating like an E commerce store where they're doing like tens of millions in annual revenue and they've created it through your product. Like they're probably willing to pay you several hundred thousand dollars a year to like keep it up, keep it looking good, like maintain it, that sort of thing.

Whereas someone who's just making their personal webpage is definitely not going to be paying you hundreds of thousands of dollars a year to do that. Yeah, yeah. I think there are two ways we traditionally look at pricing. Like this is not a hard and fast rule, but just general trend we have seen. Is the product being used by a machine or a script or is the product being used by a human? There's only so many requests a human can make even though we keep prompting the app and there's a lot of tokens and then usually if it's used by a machine, it's a consumption based pricing.

Right. So we see this a lot in the infrastructure players like, you know, like either databases or like API products. And then for things that humans use, say maybe canva, figma of the world, maybe that's like a per c per month kind of pricing. I totally agree with Justin. I don't think tokens is just the right measurement in the long run. And then I actually do think that the value these tools kind of bring to individual users and enterprises are dramatically different.

And then for the individual users maybe it's more like a productivity gain, like is it how much money am I willing to pay per month to get to where I couldn't have before? And then for enterprises actually think it's more packaging question instead of outside of pricing, like what other team collaboration features I can factor in so that I can pay more per person, per se.

Interesting. Speaking of incumbents, do these products survive? Right. If you take some, some of the newer companies that you mentioned that are doing extremely well, getting to millions in error very quickly. Is it just a matter of time before large companies, let's say like a Google or you know, maybe a better example would be something like a figma, right? Which is already has the connective tissue to designers, for example, is it just a matter of time before they start integrating these features and basically blow these new startups out of the water? Or how do you think about if there's really defensibility there? We're in the business of investing in a lot of up and coming companies and then we also think, I mean figma is a portfolio company and they're doing amazing things in the AI front. So I will actually definitely count Figma as one of the trendsetters in the space, especially for design and engineering.

There are many ways to tackle the problem. But I think compared to, you know, Google of the world, which if you have used Google Docs or you know, Excel before, there's always an argument on why didn't they build AI native Excel like a year ago or why didn't they, you know, have a good way to incorporate AI into the writing experience Just yet you ship your org chart. And that's the fact I actually know that, you know, at Google you have, you know, different orgs shipping for AI and not AI. I don't think that's a way to ship AI native products. So I obviously like, we're very optimistic about like up and coming startups kind of taking on the incumbent for this very reason.

Yeah, and I think those are great points. And also the good thing about this space is you're creating new markets and it's truly becoming massive. It's kind of similar honestly to what we've seen AI do for content generation across all of modalities, like in some way you can think of like, I don't know if it's code or website or web app, it's like another modality like AI image generation didn't replace like photography and it didn't replace like artists. It just allowed people like me, who were never creating like images or art before to create it. And so I think we'll see something similar in this space where there will be products like Figma that have immense user base, really great products who will be able to approach this problem from like, hey, we know what our users need and like we can get gain an advantage from the existing platform we've built by adding features that allow people to go like from design to web app or something like that super easily.

But then there'll be a ton of a huge market of users who, like me, who are not in Figma every day and don't start in Figma and who are going to start in another sort of product. And so I think that's what's super exciting is like it's not like we have this really small constrained piece that everyone is fighting over. It's like the pie is getting bigger and bigger every day, especially as these models get like better and better because more people can do things and more use cases are being discovered, which is part of what's like so exciting about this space.

Totally. I mean, I think that is really the most important point. Right. Which is just the Number of people that can participate here is so unprecedented. Like before Vibe coding or really just AI code in general, what was it like? Less than 1% of the world knew how to code and therefore could create things with code. And so now, even to your point, Yoko, maybe those people don't need to learn to code, but now they have this resource that they can build with.

So maybe ending there. Is there any high level takeaways or things that you want to share in terms of what you're excited about in these next steps? Looking ahead, whether it's next, the specific tools we have now, or new tools that you see on the horizon that should be built, what do you think in terms of what's coming? Even on the developer side, I'm excited to see a new set of tools here.

Instead of the traditional tools like Git, you should have something else out there that's more AI native. There's a new way of creating software even for people who can code. And then it's hard to put my fingers on where exactly it will evolve, but I know it will be very dramatically different. Maybe two years ago it was hard to see what would it look like for something like Cursor to come out.

Now we just use it every day and we're so used to it. I actually think it will happen to most of the developer productivity tools out there. Yeah. I think from a consumer perspective, two things I'm super excited about. One is the problem of like, how do you integrate other services that you need to make, like a truly working web app? Like I would love to not have to go somewhere else to create a database and then go back to a Bolt or Lovable or a replit or a V0. I would love to like not have to go to Stripe separately or Authentication separately.

Like as much as all of that stuff can be packaged in just like one click, it does it for you. I think that will massively explode the consumer usage and small business usage as well. Like honestly not even consumer, just like non technical people. And then the second thing I'm really excited about is like, what are the different ways that we can enable prompting here? Like thus far prompting and creation has been very text driven, which is great because like if you can speak English, you can make something, which is incredible.

But I think especially for more like aesthetically oriented people or just everyday consumers thinking about it, it's like, wouldn't it be cool if I could like upload a screenshot of my Instagram page or if I could upload my like Spotify Listening history or like, my five favorite movies and be like, this is my vibe. Like truly like vibe coding. Exactly. Or like, you know, here's screenshots of like six websites that I love. Can you make something similar? I think that sort of thing will be super interesting and exciting and especially young people, I think are gonna love, like the whole Gen C and even gen Alpha.

Younger populations love that sort of thing. Wait till they realize code can animate. Yeah. Crazy things that they'll start creating with that. I think we're gonna have a return to the like 90s vintage aesthetic for a bit where you have like the giant mouse that like leaves like a trail across the screen or like those sort of like fun things that people don't really do anymore. Yeah, I don't know. That's one of my personal predictions. I love that.

That's almost like mood board to, you know, apps instead of prompt your apps. Yeah. And I think also bi directional. Like I was telling you both, my husband started vibe coding and he was telling me the things he was doing while he was prompting. And I was like, oh, I would have never thought to ask the AI for design examples. Like, show me five examples of the way that I could visualize this. And I was like, I did not think that that was possible, but of course it is. Right.

And so there's also an element of like knowing how to prompt, but also I think a bidirectional nature where maybe the AI becomes more vocal about like, hey, have you considered this? Yeah. Instead of it just being like, hey, build this and then the AI goes and does it.

Yeah. I guess more to the point of software for what I just really love, like this, you know, like this tagline you kind of came up with recently. This may be worth a whole different topic, a whole different podcast episode is that we are seeing a lot more of the developers building MCPS model Complex Protocol Server. Can I just quickly share what that is for first? Yeah.

So the way MCP works is that you could build something that extend the capability of a coding agent. So on cursor, I can build, say, a MCP server to teach agent how to send email. Now I can highlight something and then send emails there. I think it will be the next generation API. I don't know if it's MCP or a different protocol, but these platforms are very quickly becoming very extensible, whether that's from components like Clerk Supabase or from MCP actions that agents can call as tools. Very excited for the next phase of that. As more and more people build the plugins here.

Love that. Great. Well, I am so excited to jump from this recording and go vibe code.

TECHNOLOGY, INNOVATION, ENTREPRENEURSHIP, VIBE CODING, PROGRAMMING TOOLS, AI DEVELOPMENT, A16Z