Vibe Coding for PPC: How Marketers Can Build Google Ads Tools Without Coding

Episode Description

In this episode of PPC Town Hall, host Frederick Vallaeys sits down with Nils Rooijmans to dive deep into the world of vibe coding for PPC professionals. They explore how AI tools are changing the way we create applications without traditional coding, and discuss the benefits, practical use cases, and best practices for marketers.

Nils Rooijmans is a seasoned Google Ads automation expert and a pioneer in using AI for PPC management. With extensive experience in developing Google Ads scripts and leveraging AI for marketing automation, Nils shares his insights on how vibe coding can empower marketers to build tools faster and more creatively than ever before.

You will learn:

• What vibe coding is and why it’s becoming popular in PPC
• How vibe coding evolved from generating simple Google Ads scripts to building full applications
• Examples of how marketers can use vibe coding to generate campaign assets from landing pages
• Why vibe coding is ideal for quickly testing new ideas
• The importance of reviewing AI-generated tools to ensure reliability
• Tips for getting the most out of vibe coding while maintaining quality control
• How vibe coding fits into the broader trends in PPC automation

Episode Takeaways

Frederick Vallaeys and Nils Rooijmans discuss what vibe coding actually is, why it’s gaining traction among PPC marketers, and how it’s being used today to prototype tools and generate assets.

Nils takes us through a hands-on session on how to vibe-code a Google Ads tool, suggests use cases, and explains why rapid prototyping matters, where guardrails are necessary, and how marketers need to think more clearly about risk, scale, and human oversight when vibe coding.

Watch your idea take life with vibe coding

One of the biggest takeaways is that vibe coding lets marketers simply describe what they want and let the AI handle the rest. Nils pointed out that this approach removes the technical barriers and makes tool-building feel intuitive. It’s about translating a marketer’s vision directly into a working tool without writing a single line of code.

“Basically, you describe the vibe of your idea and then, through interaction with the AI, you build it together, and the AI is building the code for you. So you’re not writing JavaScript code or Python or any other programming language… through human language, you come up with ideas for software that the AI is building for you,” explained Nils.

Nils also discussed how vibe coding is part of a larger trend in PPC that combines human creativity with AI-driven efficiency. Vibe coding lowers the barrier to building custom tools, helping marketers focus more on strategy and creativity and experiment more freely.

From simple scripts to full PPC applications

Creating Google Ads scripts was the nascent form of vibe coding. But now, vibe coding has transitioned to building complete web applications with a database and set up different systems with various software packages.

AI companies like Google AI Studio v0, Replit, and Lovable have created environments where marketers can build these applications.

“They can actually deploy the software for you. Meaning that they can host the application. So you do not have to set up the servers yourself. You can basically share what you built with your team or with your clients,” said Nils.

This evolution means that marketers can now build and deploy PPC tools without relying on developers. They can create, test, and launch tools directly in their browsers, with the AI handling hosting and deployment automatically.

Nils highlighted that this shift represents a major step forward in PPC automation. Marketers can now quickly experiment with new tools and workflows, adapting and iterating without the usual technical overhead.

Build campaign assets in minutes from a landing page

One of the clearest practical examples in the episode was using vibe coding to generate a full set of Google Ads assets from a single landing page. Instead of treating keyword research, ad copy brainstorming, and extensions as separate steps, vibe coding combines them into a single flow.

Nils demonstrated building a tool that accepts a landing page URL and generates keywords, RSA headlines, RSA descriptions, site links, and callout extensions. The entire process took just a few minutes.

“If I wanted to advertise through Google Ads and I didn’t know anything about this type of business, there would be a lot of work involved in doing keyword research. You would have to think about the USPs and the benefits to come up with the ad copy. You need to create the site links, the callout extensions, negative keywords, etc. It’s a lot of work,” said Nils.

But with vibe coding, you can quickly build a tool that understands the context from the landing page and create ad assets.

“A tool like this took us like five, 10 minutes to develop, and it will increase the productivity of your team immensely because tasks like keyword research, creating ad copy always take a lot of time,” Nils adds.

The key point wasn’t that the AI replaces judgment, but that it removes the repetitive setup work. Marketers still review, select, and decide what gets used. The output is visible, editable, and exportable, which keeps control with the marketer while dramatically speeding up the early stages of campaign creation.

Prototyping quickly is the biggest strength

Vibe coding excels when you need to prototype quickly, but Nils was clear that it’s not always a replacement for production-ready software. He explained that the real value is in how quickly you can turn an idea into something tangible, even if that first version isn’t perfect.

“It’s great for rapid prototyping… the software that is being generated by tools like these is not production-ready, meaning that it does a great job at illustrating what you wanted to achieve, and it brings you 80 to 90%, but it’s not really robust,” explained Nils.

He explained that vibe coding is fantastic for rapidly testing ideas or creating internal tools, but for client-facing or mission-critical applications, you still need to involve developers to ensure reliability and security.

“It’s great for rapid prototyping, but if you want professional software being deployed in a business, then you still need developers to have a look at the code,” explained Nils.

Nils emphasized that this approach makes vibe coding a powerful tool for bridging the gap between concept and implementation.

How you can examine and modify code without being a developer

Even non-technical users can peek under the hood and understand what’s happening. Nils demonstrated this by examining the prompt the AI was using to generate ad copy.

Frederick suggested asking the AI directly about its own code.

“You can ask questions like how did you code this particular piece of the software, or if you’re doing some math, say you’re calculating bids or budgets, maybe on a prediction method, you can ask, what statistical method did you use? And show me the mathematical equation. I don’t need to see the code, but tell me what math went into the code that you wrote,” adds Fred.

This conversational approach to understanding code makes vibe coding accessible even to those who’ve never programmed.

Fred further explains: “It is like a developer, you can talk to it as if it were a human, and it can also start to understand your level of expertise. You can say, ‘Hey, listen. I’m non-technical, so don’t talk to me about React because I don’t know what React is.’”

Take vibe-coded prototypes to production with Cursor

For those who want to take vibe-coded prototypes to the next level, Nils explained how to progress to more sophisticated development tools.

“If I’m happy and I think of something that I really want to deploy, then I like to use Cursor, which is like a development environment for software engineers, to get the code to the next level that is production-ready,” Nils mentioned.

Cursor helps create specialized AI agents that handle different aspects of production readiness.

“Inside Cursor, you can actually create different agents that will help you with different tasks that are required to get the code production ready. So, for instance, you would have an agent that checks for security, another agent that would check for performance and scalability, another agent that would check for API connections, or maybe improvements in the software from a readability and maintainability perspective,” explained Nils.

These agents can work in parallel to improve code quality without requiring manual developer effort for every aspect.

How to build AI agents that handle your recurring PPC tasks

Beyond just building apps, vibe coding can create AI agents that handle ongoing tasks. Nils described using Claude Code for this purpose.

“Inside Claude Code, you can create agents, you can teach them skills by providing instructions like you would to an LLM,” Nils said.

He gave practical examples of agents in action.

“You can have these agents perform different tasks for you on your local machine. For instance, connect with Gmail, go through your inbox, and reply to all the unread emails or create a draft so that you can easily reply to them, or simply archive all the things that are unimportant or spammy,” explained Nils.

The key is training these agents through feedback. Nils even predicted that soon, agents might learn by observing behavior rather than requiring explicit instruction updates.

Vibe coding reveals what you actually need

Nils made it clear that vibe coding isn’t about replacing developers entirely. Instead, it’s a tool for refining and clarifying what you need before handing it off.

One insight Frederick shared was how vibe coding forces you to think through requirements you might not have considered.

He said, “You do that initial prompt and you look at it, and you’re like oh wait… It’s suggesting keywords, but it didn’t look at the other keywords I have in my account, whether there’s overlap in targeting, or whether the budgets are split. And so you very quickly start to understand that what was a simple thing in your head actually has a lot of layers of sophistication.”

Nils agreed, explaining how this process improves collaboration with developers, “If the LLM starts asking questions, oh, you’re right, we should think about this and this as well… And then it will generate a pseudo-code version of what it is going to build. And that pseudo code basically is a very explicit definition of the thing that we wanted to have built.”

Frederick reinforced this point with his own experience. “What was a simple thing in your head actually has a lot of layers of sophistication… You can start to specify what it needs to do. Now you can give it to an engineer, and it is a much better spec,” added Fred.

This makes vibe coding valuable for the early stages of a project, where the main goal is to explore possibilities and nail down what works best.

Keep humans in the loop to manage risk

When using vibe coding in PPC, understanding where risk begins is crucial. As long as humans remain in the loop, vibe coding is a powerful ally. The real caution comes when AI tools are allowed to operate unchecked on live campaigns.

Nils explained that using vibe coding for brainstorming, prototyping, or speeding up manual tasks carries minimal risk. It’s only when these tools start making automated decisions that you need to be more cautious.

“If the software I use is not critical for my business, I can easily vibe code it because the mistakes are not very costly…. On the other hand, if it involves a high risk and could make big mistakes (like, for instance, changing things in my clients’ Google Ads accounts), I want to be sure the software is not easy to hack. I want it to be reliable, and I want it to have certain constraints,” Nils further added.

Nils recommended that non-developers should have their vibe-coded tools reviewed before deploying them in high-stakes situations.

“If you’re not a developer yourself, I would recommend hiring a developer to have a look at these kinds of applications before you start using them,” suggested Nils.

Frederick emphasized the importance of keeping humans in the loop.

“You keep the human in the loop, right? The AI is doing a great job, but it’s our job to basically tell the AI what is good behavior and what is bad behavior. We need to put the guardrails in place so that it doesn’t go out of hand,” Fred added.

Why you should stick with one tool instead of chasing features

With so many vibe coding tools emerging, it’s tempting to constantly switch between them. Nils advised against this.

“I always like to recommend people to stick to one tool and then try to master it and not go after the shiny objects with testing all these different AI tools that are available these days. Every now and then, yes, you should spend some time testing them, but don’t spend too much time playing around with copies of similar tools because it’s just in my experience a waste of time,” Nils advised.

Frederick added that he runs multiple vibe coding sessions simultaneously to manage the wait times.

“Usually, the systems that I use can take two to three minutes per answer. So it can be a little frustrating. So I tend to run a vibe coding window on one side and then another window where I do my usual day-to-day work… I’ve talked to engineers at OpenAI, and they basically have three or four vibe codes running in parallel,” explained Fred.

Small agencies can benefit the most

Nils believes that smaller organizations without technical resources will experience the most dramatic impact from vibe coding.

“Especially the smaller corporations that didn’t have any developers before, they will see a revolutionary change in the way they operate their business… For smaller companies, smaller agencies as well that don’t have any technical skills in the organization, this is going to be massive,” said Nils.

He explained that these organizations can now build custom tools that were previously out of reach.

“These people can literally use software in a way that wasn’t possible before. They didn’t have the money to hire the developers. They didn’t have the know-how to do it themselves, and they didn’t have the ideas of what was possible,” Nils added.

Nils emphasized that vibe coding is something that activates a new way of thinking once you experience it firsthand, and that the AI itself will guide you through the learning process as you go.

He stressed the importance of starting small and building incrementally rather than trying to create something perfect from the start. The real power, according to Nils, comes from gaining that initial feel for what’s possible, which then opens up your mind to new applications and use cases you hadn’t previously considered.

Episode Transcript

Frederick Vallaeys: Hello, and welcome to another episode of PPC Town Hall. My name is Fred Vallaeys. I’m your host, also CEO and co-founder of Optmyzr, a PPC management software. So for this first episode of PPC Town Hall in 2026, we’re going to go back to the word of the year from 2025. And that word is actually two words. It’s vibe coding.

So you’ve probably heard about vibe coding. It’s really gained momentum, and we think it’s going to continue to grow in importance in the next year. So that’s why we wanted to talk about it today. And who better to talk about vibe coding, which is a way of programming, than someone who’s been involved very extensively in programming for Google Ads through script?. And that is Nils Rooijmans.

Nils has been involved in artificial intelligence from before people were talking about this. He’s been one of the leading voices when it comes to Google Ads scripts. And so I’m really looking forward to hearing what his opinion is on vibe coding, where it fits in, how he’s using it, and how we can really use it to do better in PPC management and digital marketing. So with that, let’s get rolling with this episode of PPC Town Hall.

Frederick Vallaeys: Nils, welcome back. Great to see you again.

Nils Rooijmans: Hey Fred, thanks for having me. Always a pleasure.

Frederick Vallaeys: Hey, you’re hanging out in the Bahamas these days, the Caribbean?

Nils Rooijmans: Yes, it’s an island, Curaçao. It’s in the southern parts of the Caribbean, pretty close to Venezuela, which is pretty exciting these days, given the geopolitical things that are happening here. It’s always nice. I love to spend the winter here.

Frederick Vallaeys: So, winter months, November, December, January, sunshine, and they speak Dutch in Curaçao, right?

Nils Rooijmans: Yes, you’re right. It’s part of the Dutch kingdom. So, we have many languages here. The local language is called Papiamento, which is a combination between Portuguese, Spanish, English, and Dutch. But yes, a lot of people speak Dutch. As a Dutch person, you feel right at home.

Frederick Vallaeys: It’s beautiful when you can take the winter months and live somewhere a little bit nicer, warmer. Hey, being there, I really do appreciate you joining us here and talking about vibe coding. But so vibe coding, obviously, this big word. It’s the word of the year in some of the dictionaries for 2025. But that said, it’s very buzzy, and I think a lot of people still don’t really know what vibe coding is. So let’s start there. What is vibe coding?

Nils Rooijmans: Yeah, yeah. So basically, vibe coding is a way to use an AI to explore ideas to build software. So it’s not you writing the code, it is actually the AI writing the code for you, and it is you interacting with the AI to come up with the ideas. Basically, you describe the vibe of your idea, and then, through interaction with the AI, you build it together, and the AI is building the code for you.

So you’re not writing JavaScript code or Python, or any other programming language. You just, I like to speak to the AI, but you can also use the keyboard, of course. Yeah, through human language, natural language, you come up with ideas for software that the AI is building for you.

Frederick Vallaeys: Very cool. And so I think one of the fundamental concepts, and maybe the thing not everyone has picked up on quite yet, is that when you go to a chatbot, whether it’s ChatGPT or Gemini, we all ask it a language, and whether it be Dutch or English or whatever language you speak, and we get an answer in that language. But there’s no reason that these computers and these AIs can’t respond in any other language you want, or that could even be a programming language, right? And that’s where it is able to start building Python. It’s able to build JavaScript.

It’s able to do a lot of the things that are then usable to another computer system. And that’s where vibe coding comes in. But where did that whole concept of vibe coding come from?

Nils Rooijmans: Yeah, it’s actually funny because I remember maybe the last show that we did a couple of years ago, you and I were already using the LLMs like ChatGPT to create Google Ads scripts for us. And that was basically a precursor of something that is now termed vibe coding, because you and I, we were speaking natural language to the LLM. The LLM was generating the JavaScript for us.

And because you and I both can read JavaScript, we knew, oh, this is what we want, but not exactly. So we gave it feedback through text, and then it fixed the scripts for us. That way, we were much more productive. But earlier this year, one of the co-founders of OpenAI, Andrej, I believe, coined the term. It became really popular, and it is basically his way of explaining how developers use LLMs these days to generate code and complex software systems.

So it became a really popular term this year, and many people started using LLMs this way to actually build rather complex software applications that they can easily create themselves without having to go to a developer, which is great, of course.

Frederick Vallaeys: And I think you’re hitting on an important point there, which is complex systems, because like you were alluding to, when we were writing Google Ads scripts with the help of GPT, and I think for me it was GPT-4 when it was the first time that it literally took my request and in a single shot produced a fully working code that I didn’t have to go back and fix. But these were still singular files, right? It’s a single piece of JavaScript code that you copy and paste into Google Ads.

And in that regard, it’s often disconnected from the systems people already use, and it doesn’t have a database, and it doesn’t have authentication, and it doesn’t have a lot of the other things that are basic plumbing for most of the SaaS software that we use today. But we expected, and so it wasn’t able to build those things. But now you’re saying we’re talking about advanced systems. So, does vibe coding enable you to add all of these niceties, like having a login, connecting to databases, etc.?

Nils Rooijmans: Yeah, yeah. Exactly. Yes, it’s good you mentioned it because Google Ads scripts are relatively small pieces of software that live in the Google Ads platform. So they’re relatively simple pieces of software, right? But if you want to build a complete web application, that’s much more involved. Sometimes you need a database, you need to set up different systems that communicate with each other, different software packages.

And luckily for us, the AI companies, they have created environments where we can basically use the AI to create these software applications. So one of the examples, of course, is Google. Google has an environment called AI Studio where you can easily do it. I think you’re a big fan of another one, Vertex, with the v0 dev environment. Replit is another example. There’s many out there.

So you can use different tools to actually vibe code your application. And they will deploy, they can actually deploy the software for you. Meaning that they can host the application. So you do not have to set up the servers yourself. You can basically share what you built with your team or with your clients. And I hope today I can show you an example of how easy it is these days to do.

Frederick Vallaeys: Let’s get queued up here for an example. But so I’m a big fan of Lovable at this point. I was using v0 at the beginning. Now it’s mostly Lovable. But you also again mentioned a very important point, which in fact, vibe coding as it is being used today is often even easier than what we used to do with getting GPT to write an Ads script, because we would still have to take the ad script. We’d have to copy and paste it in the right place. It would have errors.

We’d have to copy and paste those errors, bring them back to the GPT, say fix these errors, then we get new code, and we’d have to copy and paste that. And so it was still this whole slog of back and forth. But as you mentioned, depending on which system you’re using, the result is hosted right there in your browser. They figure out how to deploy it, how to host it.

So the experience, which we’re going to take a look at here in a minute, but it’s literally as simple as having a chat with the system. It produces the code, and rather than showing you the code, it shows you the actual app that’s been built, and you can interact with the app, and if something is broken, it can look at the app. It literally has screenshot capability. It looks over your shoulder, and it sees what it didn’t do correctly, and it goes and fixes it. But yeah, but let’s take a look. I think you have an example of how we can deploy this for PPC, right?

Nils Rooijmans: Sure. Yeah. But before I show the example, maybe just to compare it to how I was involved with software development like 10, 15 years ago, and back then, there was, let’s start like I’m an old guy, sorry, 20 years ago, we had something in software development that was called the waterfall method. So basically, the product manager had an idea about what the product should look like, and he would come in with a team and say okay this is roughly what I have in mind, and that would be a product developer would write requirements, user interface specifications.

Then there would be functional requirements, and technical requirements. These would be written down in documents and handed over to the software architect. He would architect the structure for the software, and that would be handed over to the developers. They would do the coding and then, hopefully, within six, eight, 12, in reality, 30 weeks, you would get an alpha version of the software.

You could test it, and then you would have to write down the feedback that would go through the same process again to the developers, and they would hopefully fix the things that you wanted to have fixed as a product manager. That whole process, that is completely gone, as in it’s really fast, really interactive now with this new software that is available to us. With the vibe code, you can do anything like that in minutes or hours.

So this whole idea of coming up with ideas to build software that will help you grow your business, has come available to people with zero technical expertise, zero development skills. You can simply literally build it in a few minutes.

Frederick Vallaeys: It is this concept of on-demand software. So Sam Altman introduced that at the launch of GPT-5. And so this whole coming out of a world where software has to be built in the way that you’ve explained that also makes it expensive. That means that you have to pay a lot of money to another company, usually to get the software that does the things you need it to do.

But now we’re going into a world where it is much more on the fly. It’s oh, I have a need, I’m hosting a party next weekend, and I don’t like Evite for sending out my invitations, but I know these are the five things I need the system to do. You tell it to a vibe coding tool. It puts it all together, and within an hour, you’ve got something that works well enough. And is it perfect? No. But is it good enough to do what you needed it to do? Is it saving you time? And that’s a point. It’s this whole new way of thinking about tools and capabilities, right? All right. Let’s take a look at Google AI Studio.

So for the people listening, we do have a video of this on YouTube. Nils is going to take us through building an app with Gemini.

Nils Rooijmans: So, Gemini, Google also has an environment. It’s called AI Studio. And inside the AI Studio, there is a builder. And the builder basically allows you to use Gemini to create software in a vibe coding manner. This is the interface. I’m simply going to speak to it, and I’m going to describe my idea, and then see where Gemini will take us. Okay. So, let me quickly test if that is working.

I need you to build a web app that will help me create Google Ads campaigns based on a single landing page. Okay, the input should be one field where I can copy paste the URL of the landing page. And then I need you to present keywords, RSA headlines, RSA descriptions, site links, and callout extensions.

I’d like to be able to select the keywords, the headlines, the descriptions, and the site links, etc. that I want to use. I’d also like to see an output button that would generate a CSV file that I can use as a bulk upload in my Google Ads account, or I can simply copy paste it in the Google Ads Editor.

Okay, so I basically came up with these specifications. I think there’s a lot of things that I probably should be adding to the specification to get something that is really working, but let’s just see where it’s going to take us from here. Yes. So, I just spoke to Gemini. It listened to my idea, and now it is analyzing my train of thought. Basically, it is thinking about what kind of application I’m looking for.

It’s doing some research in the back, and it is creating some files. As you can see on the left here, there’s an HTML file, something that is called JSON. Don’t worry if you don’t know what it means. TS files. This is a little bit of the technical details that normally you would need to worry about, because if you were to build an application like this, you would need to be a developer.

But now the AI is doing the development work for you. So it’s creating the files that will be built, that will create the software for the web application that hopefully will be finished in a few seconds. And in the meantime, it is presenting suggestions based on the initial ideas. So I can, these cards. Okay. So I’m just going to click through them.

Incorporate audience targeting suggestions. Oh, that’s actually a good idea, right? Because if we are creating a new campaign, sometimes we also want to include audience targeting in there. So could be something we could add to it. Images with the Nano Gemini Pro. Great suggestion because I didn’t talk about image extensions, image assets, but these can be of value as well.

So it understood my request, and it already started providing suggestions parallel to working on the application. Okay. So, it finished in, let’s say, two or three minutes. And here we have an ads generator AI that can turn your URL into a campaign. Let’s open up one of my landing pages. So, I’m just going to copy a URL here. This is, I’m quickly going to show you.

So, this is one of my services where people can call, book a call, and I will help them fix any bugs in Google Ads scripts. So, there’s some text here, not a lot. It’s basically a very basic landing page on my website. But if I wanted to advertise this through Google Ads and I didn’t know anything about this type of business, there’s always a lot of work involved in doing research, right? Keyword research.

You would have to think about the USPs, the benefits to come up with the ad copy. You need to create the site links, the callout extensions, negative keywords, etc. It’s a lot of work. So let’s say you have a small agency where you have many different clients that, for instance, you’re doing Google Ads for dentists or for a certain type of lawyers, and most probably you will have tens, hundreds of clients that have similar landing pages.

Ideally, you would have a tool where you could simply insert the landing page and then say generate assets or generate the campaign form. So hopefully this tool is going to do so.

Frederick Vallaeys: Yeah. What we’re looking at here is literally something that was built on the fly, right? You didn’t pre-build this. This is two minutes.

Nils Rooijmans: Exactly. You talk and you go.

Frederick Vallaeys: Okay. Now, what we see on the screen here is a full output with columns for keywords, RSA headlines, and RSA descriptions. Take us through this, Nils.

Nils Rooijmans: So, it analyzed the content on this website, and it suggested keywords. Keywords are Google Ads script support. That’s relevant. That’s good. Google Ads script help, Google Ads script expert, PPC script assistant, fix Google Ads script. That is, these are all keywords highly relevant to the service that I’m offering. It also suggests match types, and I can simply collect as I asked for it. I can select or deselect the keywords. So let’s say I want to target these keywords.

Have a look at the headlines. Google Ads script support, expert script assistance, fix your ad script now, and one-on-one script help. Not bad at all, right? Obviously room for improvement.

Frederick Vallaeys: A couple of things to point out here, too, is that your prompt, what you asked it to build, never specified things like you should be able to select or you should say how long a headline is that it needs to stay within the bounds of what is allowed in Google Ads. But all these things the LLM figured out because they had an understanding of oh Nils is trying to do Google Ads, and so it understands that Google Ads has headlines and there’s a 30 character limit, and it understands the character limits for descriptions, and it naturally decides that that is a logical thing to show and that’s what it’s done. So that’s the beauty of these systems is you don’t have to go into a ton of detail. It figures out what you likely would have wanted.

Nils Rooijmans: Exactly. Yeah. And you bring up a good point because that’s also one of the reasons I like to use AI Studio because it is Gemini, a Google service, right? So it has a lot of knowledge about the Google ecosystem with all the different tools including Google Ads. So I find that if I develop an app inside AI Studio it has an, it’s less effort to integrate it with my other Google tools including Google Ads.

So that knowledge that you just described about the character limits for headlines and RSA descriptions, I think the other LLMs would also know about that. But yeah, it got that. But here’s for instance, it created six descriptions, but all six of them have more than 90 characters. So probably I can, if we go back, provide feedback that can only suggest descriptions that actually meet the limitations of the 90 characters. It will fix it in the next run.

Frederick Vallaeys: What’s very cool here is that despite showing headlines that are too long it actually has marked itself already as that’s an error because it knows it should only be 90. So there’s already, because you can imagine a system that just writes 100 characters doesn’t even bother to tell you and then you try to upload it. That’s when you find out it’s not going to work.

And I see these things evolving quite quickly. So when I was vibe coding, even a couple of months ago, the results would come back. There would be many errors. They would take six iterations and it would sometimes struggle to fix those errors. But nowadays when I use Lovable, sometimes I ask it to fix one thing and then it fixes that one thing and then it comes back and says, “And by the way, while I was looking at the code, I also noticed this other thing that seems wrong, and I fixed that as well on the fly.” So these systems are evolving super, super quickly when it comes to doing the right thing.

Nils Rooijmans: Exactly. Yeah. Yeah. That’s also what tools cannot do today. It’s amazing what they can do next week. And also the difference between the tools like I mentioned, the distinction between AI Studio and v0 or Lovable, there the providers are always playing catch up as in if the one tool is providing a feature that creates high value for the users, the other ones will simply copy it and the development is really quick these days.

I always like to recommend people to stick to one tool and then try to master it and not go after the shiny objects with testing all these different AI tools that are available these days. Every now and then, yes, you should spend some time testing them, but don’t spend too much time playing around with copies of similar tools because it’s just in my experience a waste of time.

So, just to finish this demonstration, maybe, yes, to finish this demonstration, we’re going to copy the output to the clipboard. See if that works. And hopefully if I create a new Google Sheet here with a little bit of luck, it will generate output that we can, there you go.

I’m having a look at the heading here. I’m not 100% sure this is going to work as a bulk upload, but people who have worked with Google Ads Editor or bulk uploads already recognize that this is almost what we want for an upload to Google Ads. So in theory, this could already work and otherwise we would need to instruct it to have a look at the specifications of the templates that Google Ads Editor or bulk uploads use to upload this information to Google Ads.

But getting back to my preference for AI Studio is that you still need to copy paste this into Editor or a bulk upload. Why not have this app talk to the Google Ads API directly? So that is actually one of the things, it’s a bit more complicated and we can get to this within the hour, but you could have an app running that has access to the Google Ads API and then you can literally say okay I’m happy with this, instead of downloading a CSV just add it to my account. So you could create an app like this literally within an hour that would make the changes inside your Google Ads account.

And then if you’re happy with the result, I want to show you how easy it is to share it with your colleagues or even maybe your clients or your boss. You can simply deploy the app online. So this is running inside my AI Studio right now. So nobody can see it but me. And if I want to share this with other people in my team, I simply click the deploy the app button. I selected a cloud project. So I prepared the vibe coding demo. It’s basically a Google Cloud project where that’s going to host the app. I’m going to say deploy it.

Okay, this is going to take a few minutes, but then it’ll be running inside the Google Cloud and I will get a URL and I can simply direct people to the URL and they can use this tool. So again, oh there you go. It’s already there. There you go. So it’s not a really pretty URL, but we can change that as well. You can upload the app to Vertex or some other environment.

Frederick Vallaeys: But the point that you’re making, very easy to get it onto a URL that’s publicly shareable. And then if you wanted to actually start selling this to a customer, maybe you can put a nicer domain there. Okay, so we’re doing a test here using this new tool for the optmyzr.com domain, and let’s see if it understands what we do, which if it doesn’t that’s maybe a reflection on our website, and maybe then we need to tell our web team to fix a few things, but let’s see what comes out of here.

Right, so ideally this would already generate something that is highly relevant, but if it’s not then we can go back and provide some feedback to the development part. As you can see, I’m going to zoom in a little bit to make it easier to read. I think this is already really good, right? PPC management software, yes. Google Ads optimization. Maybe that’s a little bit too broad. PPC automation tool.

So this, a tool like this, it took us like five, 10 minutes to develop, and it will increase the productivity of your team immensely because stuff like this, keyword research, creating ad copy, it always takes a lot of time, right?

Nils Rooijmans: It does. Yeah.

Frederick Vallaeys: So hey, a lot of work is being done here by the computer in a very little amount of time. This must be super expensive. How much does it cost?

Nils Rooijmans: Ah, that’s a good question. Yeah, AI Studio is free to a certain extent. I’m using the pro model. I forgot, it’s $20 or $50 a month. If you want to host it in the cloud, the application that you develop, you also have to pay, I don’t know, 10, 20, 30 bucks. It’s not really expensive. You have other environments like Lovable and v0 or Vertex. If you deploy there, it’s a bit more expensive, but I think the deployment part is a bit easier, and the look and feel, in my opinion, feels more professional and nice than AI Studio.

So you have to play around with it a little bit to see your preference, but it’s never really that expensive, and especially not if you compare it to the cost of developing this software with the developers that you need to hire.

Frederick Vallaeys: Exactly. And the pricing models, I think, are still evolving to a large degree. So again, I’m mostly using Lovable, so I know that system the best. You can start for free. You get five credits per day for free usage. And those five credits will let you build an app like the one you saw here today from an initial generation.

But now the next thing that probably happens is you’re like, “Oh, this is cool, but I wish that it suggested different ad groups, and maybe it was theming some of the keywords together, and maybe got all these bells and whistles,” right? And that’s when you start eating into these credit systems. And that’s when you end up upgrading to the $20 per month plan that gives you a couple of paid credits. And then as soon as you hit that limit, it just asks you to upgrade, and you keep upgrading.

And interestingly enough, Lovable was the fastest company to $100 million in revenue in history, in the history of the world. So clearly a lot of people, including myself, are spending a decent amount of money on this. But again, the alternative is I hire a developer and I go through this months-long process, and we have a great development team at Optmyzr, and we’re continuing to build, but they also become more effective developers thanks to these types of capabilities.

But let’s talk about that for a second, Nils. So you have a programming background, there’s this whole other class of tools which are say like Cursor that help engineers do this type of stuff. Like where’s that break then, and why would people use one type of system versus another?

Nils Rooijmans: Yeah, great question. So yeah, so the interface like this, like I just showed you, and also the v0 and the Lovable, I think they’re great for rapid prototyping. So I can imagine you in your organization are using tools like this to illustrate to your developers what kind of functionality you want to add to Optmyzr, and you can illustrate it easily because you can literally build it within a few minutes or an hour.

But the software is typically, the software that is being generated by tools like these is not production ready, meaning that it does a great job at illustrating what you wanted to achieve, and it brings you 80 to 90 percent, but it’s not really robust. It’s really not reliable. It’s not really to scale to the level that you need for all your clients. It’s not secure because it’s easy to hack things in.

So in my experience, it’s great for rapid prototyping, but if you want professional software being deployed in a business, then you still need developers to have a look at the code. And that is also one of the good things about these environments because you can simply go in and have a look at the code. So I’m here. So if you understand code, you can simply go in there and for instance this Gemini service. Maybe if I zoom in here, I can see the prompt.

So another thing is that I basically didn’t explain anything about writing ad copy, right? So the quality of the site links and the ad copy is mediocre. But if I improve the prompt that is being used by this software, then I get much better headlines, much better descriptions. So that is an example of the changes you can make if you understand a little bit of the code, or you can do it through the suggestions as well.

Long story short, the quality of the code is good, but it’s not at the level of professional software that is being developed by real software engineers yet. So I wouldn’t recommend using vibe coding tools like this for very complex applications that need a lot of backend logic, database connections. It’s good for basic, simple web apps like the demonstrations that I just gave. And I think you also created a Chrome extension, which is also a great example of how you can use tools like this, but for complex software systems it’s good for prototyping.

Frederick Vallaeys: Exactly. That’s the whole notion of on-demand software, right? It’s if this is something that you’re building for your own productivity, like it’s almost a no-brainer because you’re the only one using it, security is not that big of a concern. But the moment that you do deploy it and want to scale it, you do have to have secure login. You have to make sure the data doesn’t get stolen. You have to make sure that the servers stay up.

And again, it then depends on what system you use. But Lovable allows you to scale up the size of server infrastructure that you put behind it. And obviously, it costs more money if you go to a bigger deployment. But these things can be done. Now, Nils, you did bring up an interesting point, right? This app that you’ve just vibed is using Gemini AI to look at a site and then generate ad headlines for that.

You’ve gone into the code to try to find what is the prompt it is using to do that AI magic. But I suspect you can also just ask the code assistant the question. So can we try that? Can you tell me the prompt you’re using to do the ad generation?

Nils Rooijmans: That’s right. Because I think what that illustrates, if this works, is that you still don’t have to be technical, right? It’s nice you can go and look at the code, but even if you don’t want to look at the code, you can ask it like how did you code this particular piece of the software, or if you’re doing some math, say you’re calculating bids or budgets maybe on a prediction method, you can ask it, well, what statistical method did you use? And show me the mathematical equation. I don’t need to see the code, but tell me what math went into the code that you wrote.

Exactly. Yeah. So, let’s give that a try. Show me the prompt you use to generate the headlines and RSA descriptions and three ways to improve that prompt.

Frederick Vallaeys: Okay. So, play by play here. But Nils was dictating that into the system. So, it’s much faster than typing. So now it is thinking about that using Gemini 2.0 Pro, and it gives you little updates there along the way. It says what it’s doing so that you get a sense for it not being stuck, and then usually the systems that I use, it can take two, three minutes per answer.

So it can be a little frustrating. So I tend to run a vibe coding window on one side and then another window where I do my usual day-to-day work or some online shopping. But I’ll always do two things at the same time. And I’ve talked to engineers at OpenAI, and they basically have three or four vibe codes running in parallel. They can each complete when they complete, and then they can go deeper on the ones.

Nils Rooijmans: Exactly. Yeah. Yeah. I have to say, and that also brings back to the question about the different tools, right? Because I’m using a tool like AI Studio, v0, for these rapid prototyping things, and then if I’m happy and I think okay this is something that I really want to deploy then I like to use Cursor, which is like a development environment for software engineers to basically get the code to a next level that is production ready.

And inside Cursor you can actually create different agents that will help you with different tasks that are required to get the code production ready. So, for instance, you would have an agent that checks for security, another agent that would check for performance and scalability, another agent that would check for API connections, or maybe improvements in the software from readability and maintainability perspective. So inside Cursor you can create multiple agents that work in parallel to upgrade the code to the level that is required to get it production ready.

Now, as you can see here on the left, it has provided the answer to our question about the prompt. So it shows the JavaScript where it has the prompt. I’m a Google Ads expert, I need to analyze the following landing page URL, blah blah blah. So this is a long prompt with constraints, and then it also suggests three ways to improve this prompt. If we read through this based on our knowledge of what a good ad copy looks like, we might agree with the suggestions and say, “Okay, incorporate these into improving the prompt, and we don’t have to go into the code to do it ourselves because this tool will do it for us.”

Frederick Vallaeys: Cool. So, yeah, it is like a developer, you can talk to it as if it was a human, and they can also start to understand what your level of expertise is, right? You can say, “Hey, listen. I’m non-technical, so don’t talk to me about React because I don’t know what React is.” And sometimes it’s also going to say, “Hey, you just told me you want to connect to the Google Ads API, and you told me to go into the Google Developer Console and like something about OAuth.” Like, I get confused. Okay.

But so you tell me, can you explain this more? Can you point me in the right direction? Here’s where I am. Like, can you help me get to the right place? And it will walk you through all of these things. So it’s really cool because it can write code, but it can also guide you to where you need to be to take that next step that’s necessary to get you where you want to be, right? And Nils, you also talked about deploying, right? So maybe you go into Cursor as a next step. So the code that you just produced in vibe coding, is this now throwaway work, or how do you get that to your engineers to do something else with?

Nils Rooijmans: So I could send it over to my engineers, and they would be like okay we’re going to completely refactor this, as they call it, which would be a lot of work for them, or they could use Cursor and agents to do the work. Many of the developers, most of them, already use something like Claude Code or Cursor to help them improve their productivity by using the AI to help them improve the quality of the code in ways that are important to them as software developers.

For us, these technical requirements to make it production ready, us as non-developers, we’re not really that interested in that. We’re interested in the functionality that the tool has. But the software developers know that there is more to it. And they use tools like I mentioned, like Cursor, Claude Code, Cursor, GitHub Copilot. These are the kinds of tools that the techies use to improve the quality.

Now, I’m not saying you shouldn’t use these tools or start playing around with them, because you can use these tools to build agents, which is the next level, right? You can create AI agents that will help you with completely different tasks. So you could, for instance, inside Claude Code, you can create agents that will help you have a look at your emails and already generate standard replies to all the emails you have today, or I don’t know, transcript and generate blog posts from your YouTube videos.

So you can create different agents for different tasks that run on your machine and do the work for you. So there’s another example of using AI in an interactive way, sort of like vibe coding the AI agent to teach it what needs to be done, having a look at what it is doing and how it is performing, and then through interacting with the agent improve the quality of its work, and that’s really cool.

Frederick Vallaeys: So when it comes to these agents, are you saying you would use the Cursor or AI Studio from Google to build an agent, or do you have a different sort of tool that helps you build these agents?

Nils Rooijmans: So the thing is with the agents I like to use Claude Code, and inside Claude Code you can create agents. You can give them, you can teach them skills by providing instructions like you would give instructions to an LLM with a system prompt. A system prompt would basically be something like you put the LLM on stage, right? So you act as an experienced PPC expert with a lot of experience in optimizing ad copy. Your job today is to improve the ad headlines for a blah blah blah.

So these kinds of instructions you can use as well to create agents inside Claude Code, and then you can have these agents perform different tasks for you on your local machine. For instance, connect with Gmail, go through your inbox and reply to all the unread emails, or create a draft so that you can easily reply to them, or simply archive all the things that are unimportant or spammy, and based on its behavior you can train that agent to improve its performance.

So you have an agent that is going through your email, and then you see okay you decided that this was a bit spammy, but it’s not because of this and this reason. So, I’m going to update these instructions. So, it’s still me updating the instructions. So I can see this changing, well maybe already next year, where I don’t have to explain why it’s not spammy, or I don’t have to explain why I was changing the reply that the agent has generated for me.

I simply make minor changes to the reply and then send it, and then the agent will be observing my behavior and learning from it. I think that is very near the horizon.

Frederick Vallaeys: And I think that maybe raises another good question. So at what point do you just pay someone for a tool that does this versus build it yourself? And I think in this scenario, because you constantly have to retrain it and make it better and it’s clearly responding to email, you’re going to be doing that for the next couple years of your career, right? So maybe that is where you pay someone to do the upkeep for you, keep making it better.

Whereas something like, hey, I’ve got this one-off project for a client and it’s going to take me eight hours of work, but I could vibe code something that helps me do it in two hours, and after that I might never do that sort of project for a couple of months, right? But by that point, I’ll just vibe code a new version of it that is more in line with the current state of the industry. So that’s how I would think about when I buy versus build. What do you think about buy versus build in that case?

Nils Rooijmans: Yeah, great question. The software that I like to use that is not critical for my business, I can easily vibe code it because if it can make mistakes that are not very costly, if I’m convinced that it cannot make a mistake that is very costly, that I’m really happy vibe coding it. If, on the other hand, it is something that I feel there is a huge risk involved if it does make big mistakes, like for instance changing things inside Google Ads accounts for my clients, then I want to be sure that the software is not easy to hack.

I want it to be reliable, and I want it to have certain constraints. And because of that, and because I have a background in reading code, I like to have a look at the code and make sure that it is doing exactly the kinds of things that I would allow it to do. Now, if you’re not a developer yourself, I would recommend hiring a developer to have a look at these kinds of applications before you start using them.

Frederick Vallaeys: Yeah. And I also think, like you mentioned in your books, you keep the human in the loop, right? The AI is doing a great job, but it’s our job to basically tell the AI what is good behavior and what is bad behavior. We need to put the guardrails in place so that it doesn’t go out of hand.

Nils Rooijmans: Exactly. Guardrails are such a big thing in 2026 with all this automation.

Frederick Vallaeys: Now maybe in what you said, maybe we’re starting to discourage people from doing some vibe coding, but I think one interesting element that’s come out for me is often as a marketer or a business person or whatever role you have in your head, it’s, oh, there is this simple thing that I need to do and just make the code, right? Make the software. And then you do that initial prompt and you look at it and you’re like, oh wait, like it’s suggesting keywords but it didn’t look at what other keywords I have in my account and whether there’s overlap in targeting and whether the budgets are split.

And so you very quickly start to understand that what was a simple thing in your head actually has a lot of layers of sophistication. And that’s a big benefit because if you can go through that with this vibe coding tool, and even if it’s not perfectly connected to Google Ads and it’s not perfectly pulling in all the data you have, but you can start to specify what it needs to do.

Now you can give it to an engineer and it is so much better spec because part of what you were describing is that 36 months of the developer going and doing the thing and then it comes back and you’re like wait, you didn’t consider the factors, other keywords and other campaigns and ad groups, and it’s, you didn’t tell us, right? So maybe that’s why they missed it. But this forces you to really think through what the tool needs to do, and so the developer is still going to take 36 months, but what comes out of it is going to be a hell of a lot better in that first version than if you hadn’t gone through that process initially.

Nils Rooijmans: Yeah, that is a great point. Exactly. It’s also during my training. That’s exactly how I like to use LLMs for generating Google Ads scripts. First, to basically convince the LLM that it needs to ask a lot of questions about the requirements because we’re not aware of the requirements most of the time, right? We don’t know exactly what we want.

So if the LLM starts asking questions, oh, you’re right. We should think about this and this as well. We should make choices like that. And then it will generate a pseudo-code version of what it is going to build. And that pseudo-code basically is a very explicit definition of the thing that we wanted to have built. And if we can get there through this natural communication, with the interaction with the AI in a way that doesn’t require us to know anything about coding, that’s just a huge, huge, huge win.

Frederick Vallaeys: And so in the example that you gave today, we did a single prompt generation, but what you’re saying now is like that was just an example, right? Like maybe the better scenario is that you go and talk to GPT and you tell it this is kind of what I’m trying to do. Ask me questions like what am I missing? What should I be asking the vibe coding tool to do for me?

And so now you spend half an hour going back and forth, and you put it in voice mode, right? Because nobody wants to type anymore. So you put it in voice mode, you have that conversation as if it was a human product manager. And then what it spits out might be, and I’ve talked to OpenAI people, but like they have 10,000-word prompts, wow, as the initial prompt. Yeah.

And that’s pretty sophisticated, but that includes things like what is our preference for the language to use and stylistic preferences, and what do we need to consider in terms of security and UI design. And so all of that can go into that initial prompt if you wanted to. You don’t, I also don’t want to make it sound scary like that’s what you need to do to do it right. You can go either path, but it is your choice. You can make it as sophisticated as you want.

Nils Rooijmans: Yeah. Yeah. Start small. Start small and then just take incremental steps to improve the quality and also the complexity. And it’s relatively easy to get started, and once you get started, you get a feel for what is possible, and it will activate a mindset that is required to take you to the next step. And then the AI, the LLM, will actually guide you in the next step. And if you enjoy the experience, which I guarantee you will because it’s a superpower, you will gradually increase your understanding of what is possible and also the complexity of these apps you’re going to build.

Frederick Vallaeys: And Nils, I can’t tell you the number of nights that I’ve been up past midnight vibe coding, something I haven’t done in a long time because we got a big engineering team at Optmyzr. But it’s fun. It’s addictive. It’s productive. Yeah.

And I’ve actually put some vibe coded tools into production. I’ve got one fun little one I did. It’s called trivithis.com. It lets you generate a little quiz about some topic you want, and then you can play it right there. I’ve just vibe coded an advent calendar, a countdown calendar for the holidays that has prizes from a lot of PPC experts and put that into production.

And again, these are still relatively low risk. This is not changing budgets and multi-million dollar ad campaigns. But still, it’s got security, it’s got a privacy policy, it’s got all of those things in place, and you can launch it. Again, really cool stuff. And I agree with what you said. Go out there today, use any of the vibe coding tools, put something in, see what it does because it’s going to be wild and it’s going to be fun. Nils, a couple of pointers maybe, like any mistakes you have faced that we can help people avoid.

Nils Rooijmans: Good question. Yeah. So one of the biggest mistakes is thinking that once you have something that is working, then that it’s production ready because there will always be edge cases, edge scenarios that you haven’t thought of yet, that if you put it into production, other users will encounter these errors. And it could be a big mistake that can cause some harm. So vibe coding can get you to 90 percent, but don’t expect it to be 100 percent unless it’s a relatively easy, simple task. That is one.

Another one would be, yeah, that’s actually a bit contradictory, because I’d like to say if you spend too much time thinking about the code and looking at the code, that is not vibe coding anymore, then you’re becoming a software developer. So if you’re in the vibe coding mode, don’t think about the code or the JavaScript or the Python itself that much, even if you understand what’s going to happen, because it will slow you down in that creative process of improving the idea. You’re really in the vibe of improving upon the idea and building the app. So don’t worry about the code too much.

Frederick Vallaeys: Yeah. And know what your purpose is, right? And I find myself at some point you just have to take two days to connect stuff to the backend systems, to make the security correct, to put in the terms and conditions, work with lawyers to put in privacy policies. That stuff is not fun. But you go into that mode, okay, this is what I’m going to get through. And then you go back to the mode of, okay, now I’m going to look at the app. Now I’m going to make it more usable. Now I’m going to make it more fun. And that’s when you’re vibing, right? And jumping back and forth is important.

But then the other thing that I find myself sometimes doing is going down rabbit holes. And so you have to be, I have to keep a list on a piece of paper. Here’s the big thing. I just thought of this other thing that’s like an edge case, like you said. I need to check on that edge case. Because then you look at the page and like the button is blue, but I wish it was red. Okay, let’s go change it.

And then you go down these rabbit holes, and eventually you’re like, “Oh, but the big things that actually matter the most I haven’t even gotten to.” So just keep that list so that you don’t go too far off track.

Nils Rooijmans: Yeah, good point.

Frederick Vallaeys: Good. Nils, you’re doing a lot of teaching, right? Are you teaching any vibe coding in addition to Google Ads scripts?

Nils Rooijmans: Not yet, but I might in the future. Yeah. So currently I’m just explaining how to use the AIs to create Google Ads scripts because the main advantage of using Google Ads scripts is that they still operate inside the Google Ads platform, relatively cheap, easy, and they can manipulate entities inside the Google Ads platform. So a lot of the software that we create using vibe coding, unless you have an API connection to Google Ads, which is relatively difficult to set up, it cannot change anything inside Google Ads.

So Google Ads scripts are my preferred tools to monitor the performance of Google Ads and make minor changes inside the Google Ads platform. Vibe coding is to create apps that live outside of the Google Ads platform that can help me. For instance, I don’t know, yesterday I created a web interface to help me generate compare-to lists.

So for instance, if I have a shopping campaign and I decide to increase my max CPC bid with a portfolio target CPC, then the way shopping campaigns work is it’ll show your ads to a whole different set of search terms. So before I upped the bid, there was a set A of search terms, and after there was set B. I want to see the difference between the two inside the Google Ads interface. It’s relatively hard to do, right?

Of course, I could create a script to send me an email, but sometimes for me it’s just easier. Okay, I have these two Google Sheets, just compare them and show the difference, right? So I created this app that allows me to simply insert two links to different Google Sheets, and it will create a nice table with the differences between them. So I can see what new search terms my products are being shown for after I’ve upped the bid.

Frederick Vallaeys: Yeah. And that kind of makes me think of another mistake that I’ve made, which is sometimes I do something the old way. I have a big project. So we did a CRM migration, and I decided to use the tools from the CRM platforms and bulk sheets to move data back and forth. And eventually I just hit some limit because the bulk sheet wasn’t able to take every entity, but these entities existed in the API. And I was like, if I started vibe coding this from the beginning, like all the simple stuff, like the bulk sheets, it would have done that in five minutes, and it would have set me up to immediately get into the more complicated edge cases, which you can do with the API.

And I ended up spending two days not doing vibe coding to do this project, and I was like, I do need vibe coding because I’m stuck. And then vibe coding got me over the finish line. But I think again, it’s that experience of having done it, having known what vibe coding can do, so that you start to solve your problems, like you’re saying. A very simple problem you had, and you were like it’s going to take 10 minutes, but if you hadn’t tested it before, if you didn’t know what vibe coding was, you would never have gone down that path, and you would have probably never had a solution for this thing.

And so that’s why I’m so excited, because all of these little frustrations that we have, or all of these great ideas that we all have, all of a sudden, like you said, it’s a superpower. We can go and build these things. We can solve them today, not tomorrow. We don’t have to ask someone. We do it today.

And I’ve worked with friends. I’ve got a vibe coding lounge here at the Optmyzr offices in Mountain View. And I have friends coming in here who’ve literally been sitting on ideas for years, and they’ve just never had the drive to talk to their developers or talk to their team because they knew what was coming. They knew they were going to spend half a year describing it and waiting and having frustration along the way. And they come into the vibe code studio here, and they walk out the same afternoon, and they get a working prototype.

Nils Rooijmans: Awesome.

Frederick Vallaeys: And some of these things actually, like one guy that I was working with, presented his vibe code to the CEO of Adobe, and the CEO was super happy. At Adobe, like they have engineers, right? But vibe coding was making this person stand out above everyone else. So it’s really, really cool what people are doing.

Nils Rooijmans: Nice. I’d like to join one day. I’m buying a ticket to San Francisco now.

Frederick Vallaeys: All right. Yeah. Hit us up in the comments, anyone who wants to come hang out at the Vibe Studio and build some cool stuff together. Yeah, let’s do that. We got the space.

Nils Rooijmans: Yeah. Yeah. Also, maybe one last point, because you mentioned the person presenting a prototype at Adobe. I think especially the smaller corporations that didn’t have any developers before, they will see a revolutionary change in the way they operate their business. I think the bigger corporations, they already have some, they have developers, they have product developers. So the mindset of what automation can bring to the business is already sort of implemented in the organization, and AI and vibe coding will be the next evolutionary step in that process.

But for smaller companies, smaller agencies as well, that don’t have any technical skills in the organization, this is going to be massive. This is going to really be revolutionary because these people can literally use software in a way that wasn’t possible before. They didn’t have the money to hire the developers. They didn’t have the know-how to do it themselves, and they didn’t have the ideas of what was possible. And I think that that’s going to be a dramatic change in the next few years.

Frederick Vallaeys: Yeah. And I’m also seeing a lot of recombinant innovation where I build one thing or I see one thing and I’m like wait, if we change this one little thing, all of a sudden it becomes a marketing piece of software as opposed to whatever it was doing before. And again, I’m thinking about this advent calendar, right? So it’s a countdown calendar, and I’m like why don’t we turn that into a teaching course? Every day you get a small nugget because people are busy.

A lot of people don’t want to spend four hours in a class learning something, but we’re all on social media, right? We can consume two minutes of something that helps us along the path. So it’s a recombinant innovation. You see these things that are like, what if for my business it was able to do these things? And that’s a cool thing too in Lovable. When I build something, I can publish it with a remix button where you can then literally go and grab it and say remix this.

It puts a copy into your Lovable, and now you can prompt it, and you can start branching. Basically, make your own version of it. So as a community, we’ve always been so good in PPC, and you’ve been great at sharing stuff. I’m really excited about how even vibe coding is going to become a community effort.

Nils Rooijmans: Yeah, I didn’t know that. That’s cool. I should check it out. Yeah.

Frederick Vallaeys: Well, good. Nils, I know you’ve got some chores and outdoor activities to get to. Thank you so much for sharing all of your wisdom around vibe coding. This has been awesome. Hope to have you back later in 2026 to share more of what you’ve been working on. And people find Nils on LinkedIn, see what he’s up to. He’s got a great newsletter. I always enjoy reading that. So subscribe to that as well. Nils, remind people where they can find you.

Nils Rooijmans: Yeah. So if you simply Google my name with scripts, you will find my website and a lot of free Google Ads scripts. There’s also a newsletter on my website. So nilsrooijmans.com/newsletter , where I share daily tips to optimize and boost your Google Ads performance.

Frederick Vallaeys: Great. Okay. Awesome. Thanks everyone for watching this episode. When you, if you want to know when Nils is coming back on or about other guests, please subscribe, hit that like button, interact with us in the comments if you have any questions or any ideas, or you get stuck on vibe coding. We’re happy to, of course, help you with that. With that, thank you for watching PPC Town Hall. Nils, thanks for joining us, and we’ll see you for the next one.