Blog

Groovy Vibes

Recently, you might have heard the term vibe coding. Which when you think of the semantics, may inspire images of surfers catching a wave or that emoji smiling with the sunglasses. 😎

It’s just a vibe. 

This is an accurate description of what vibe coding entails - using natural language prompts to build a website or an app. And, not to get ahead of myself but it’s incredibly satisfying, so much fun, and also addicting. There’s just something solidly satisfying about starting with an idea and then seeing it become something that you can see and interact with. I’ve played with four platforms and am going to break down the pros and cons of each so that hopefully, you might feel inspired to go and try a bit of vibing yourself.

Which got my solution engineer gears turning. Would there be a way that the barrier to entry could be broken down somehow to make the idea of exploring your thoughts and feelings less intimidating? Which is exactly the app idea I used in testing these vibe platforms. 

The Objective

Build an app that makes journaling more like texting a good friend. Leverage AI to respond thoughtful and supportively. Then, log journals with AI generated summaries for high level perspective on the themes of each entry. 

Of course, the app got a little more elaborate than that. One of my best years instead of keeping a “to-do” list, I kept a “to-done” list, and that was validating to look back on in review. I added this to the concept, as well as a way to track gratitudes and goals.

Vibe Coding 101

Just like with prompting, you want to be specific. 

“Make me a journaling app” is comparatively vague, requires more interpretation, and is not going to get you as good of a result as: 

“Help me develop an app that provides a AI journal, wellness tool for users. I am envisioning a platform where when you come in, AI asks a few questions to check in on how you're feeling and what might be helpful given the current mental state. Something like: "what's going well today?" or "where are you feeling stressed lately" and then allows the user to respond building an entry for that day that can be reviewed and also may impact prompts on future days. This tool should be able to identify trends across areas of stress, life events, and goal progression. One key feature I want to make sure is included is a "done" list that users can elect to fill out at the end of the day. So instead of a "to-do" list we'll call it a "ta-done!" list (like ta-da!) and this can be something built into tracking across accomplishment reporting.” 

And even that isn’t as specific as could probably be started with. But, it gives us a start and with vibe coding, one of the recommendations as expressed in DeepLearning's free Replit course is that you want to build, one component at a time.

One of the great things about vibe coding is that sometimes it adds ideas that you might not have had in mind. Or it might add some styles that you just hadn’t figured out. For example, one of the adjustments I made as I was iterating on a version of my app was to ask for a zen light green type theme and color. And the vibe coding platform I was on built a whole color pallet around that rather beautifully. But before I give too much away, let’s go into the platforms. 

MGX

MGX, short for MetaGPT X, bills itself as the world’s first 24/7 multi-agent AI development team. That’s not just marketing bravado. At its core, MGX is a platform where a cast of specialized AI agents—think team leader, product manager, architect, engineer, and data analyst—collaborate to turn your ideas into working software. The twist? As with all the vibe coding examples we’re going to go through today, you interact with them using natural language, and they follow standardized operating procedures (SOPs) that mirror the best practices of real-world software teams .

Imagine walking into a virtual office where your requirements are parsed, user stories are generated, architectures are diagrammed, and code is written, all by a team that never sleeps and never gets distracted by Slack notifications - which don’t we all. That’s the MGX promise.

Sounds like a dream right? Unlike what goes on in your head at night though, this is a quite tangible tool. MGX simulates the collaborative workflow of a traditional dev team, but with AI agents filling each role. Remember when we talked about agents? That's the core concept. 

The MGX agents communicate, delegate, and execute tasks across the entire software lifecycle - from requirements analysis and system design to implementation and deployment. The platform currently supports building websites, blogs, apps, and more, with support for additional languages like Python, Go, and Java actively in development .

The architecture here is just fascinating. MGX’s agents aren’t isolated bots; they’re orchestrated through SOPs, ensuring traceability and consistency. You can choose between “Engineer” or “Team” execution modes, balancing speed and capability depending on your project’s needs. There’s backend integration (Supabase, GitHub sync), visual editing for UI tweaks, one-click deployment, and even real-time error handling with guided fixes.

In other words - it's pretty cool. 

Building on MGX

MGX has a free tier if  you want to go try it out, which is exactly what I did to get started. I was aggregating ideas and still trying to flush out the direction I wanted my app to go. And MGX made that easy. It searched for a relevant template. I watched amazed as lines of code were written without me having to so much as touch the keyboard. Sipping my coffee, I’m going to look back at that moment as one of several recently where the way the world is shifting is obviously apparent.

The result was pretty basic, which I should mention is in alignment with what you want in an MVP - minimum viable product, or in other words a simple starting point before you start to get all complicated and add features. I made a few adjustments, including changing the aesthetics a bit, aligning the emoji’s the app created for a user to select a mood that was in alignment with how they felt, and then quickly ran out of credits trying to troubleshoot an agent that responded well to the first interaction, but then less well in further conversation. 

But credit to MGX for giving me something concrete out of the idea I had provided and something to build on and continue to refine from a solid starting point. 

Base44

Base44 is an Israeli startup that burst onto the tech scene in early 2025, founded by Maor Shlomo. The company’s mission was as ambitious as it was straightforward: empower anyone—regardless of coding experience—to build and deploy web and mobile apps using natural language prompts. In other words, Base44 is a no-code, AI-powered platform that turns your ideas into working products, fast.

What makes Base44’s story even more compelling is its founder’s journey. Maor Shlomo, working solo, bootstrapped the company to profitability in just six months. He reportedly used AI to write 90% of his code, which feels congruent with vibe code sort of vibes. 

In July 2025, Wix.com Ltd, a WSIWIG web development platform (WSIWIG - what you see is what you get, or a another way of referring to an interface that is click, drag, and drop) acquired Base44 for $80 million. For Wix, Base44’s intuitive, code-free digital creation platform is a strategic leap into “vibe coding” that aligns with their platform of empowering non technical users the ability to build and develop without having to code.

Building on Base44

I put the same prompt in Base44 but got a different version of the app. The MVP Base44 created had styles and theming that I had to prompt MGX to add. For those who don’t have a well flushed out idea in mind, this can be helpful. Base44 added in my Ta-Done list, just like I had asked for and provided a journal tab to look back at past entries. There was also an insights area to turn interactions into analytics. Which, is kind of my love language, so I was pleased. 

The chat agent within my app was also better with it's interactions than what I was testing on MGX, however, where I ran into issues and ran out of credits was trying to get the chat agent to look like a text conversation. I had the bubbles of the response from the chat agent, but couldn’t see the entries that I had written. We went back and forth, trying to understand and fix this problem. Base44 suggested “replacing the MessageBubble component with a simple debug view so we can see exactly what's happening with the messages.” This fixed the issue and the agent assured me that it understood the issue. But when we tried to revert back to chat bubbles, I had the same problem. And needless to say we had to go back on forth this a few times and never did get the chat bubbles to work, at least not within the constraints of the free tier. 

Which I’m sure with more time and credits, is absolutely solvable. A good tip when vibe coding if you hit a sticky spot like this is to explain the issue you are facing in ChatGPT, Gemini, or whatever your AI platform of choice may be, give it a copy of recent logs, and ask if to help you write a good prompt for the fix. It’s a lot better than pounding your frustration into your keyboard and pulling at your hair when the result comes back still not working. Trust me. I have a "friend" who has gone through this a few times. 

Replit

If you’ve ever wished that building software could feel less like deciphering ancient runes and more like sketching an idea on a napkin, you’re not alone. That’s the very friction Replit set out to erase. Founded in 2016 by Amjad Masad, Faris Masad, and Haya Odeh, Replit’s mission is both ambitious and inclusive: "to empower the next billion software creators by making programming accessible to everyone."

Replit’s roots trace back to Amjad Masad’s childhood in Jordan, where his fascination with computers led him to build his first business at age 12. That early experience which included navigating the pain points of setting up databases, collaborating on code, and deploying software is what planted the seed for what would become Replit. Because apparently that is what 12-year olds do these days. The company’s name itself is a nod to the “REPL” (Read-Eval-Print Loop), a concept familiar to anyone who’s ever tinkered with an interactive programming shell.

At its core, Replit is a cloud-based, full-stack development environment. It lets users write, run, and share code directly from their browser. Over the years, Replit has into a robust platform supporting dozens of programming languages, real-time multiplayer editing, and instant deployment .

But the real game-changer arrived in 2024 with the launch of Replit Agent, an AI-powered assistant that can generate, refactor, and even debug code based on natural language prompts. This makes developing something less coding and more conversation. You describe what you want and the Agent scaffolds the project, writes a boilerplate, and even suggests improvements.

By mid-2025, Replit boasted over 500,000 business users and had multiplied its revenue tenfold in just six months, reaching $100 million. Its rapid ascent caught the attention of industry giants, culminating in a partnership with Microsoft to integrate Replit’s technology into enterprise tools.

Building on Replit

So, spoiler alert, this is my favorite platform of the bunch and so this is where I took the ideas that started to form in our other experiments and really got to work. I should also note that Replit required a bit more foundational work than the other platforms. I had to create a funded secret key in OpenAI and populate that into what I was creating for the testing of AI incorporations. This is something I did not have to do in MGX or Base44, and possibly why my MGX AI chat experience was more problematic. I got to a decent MVP in an evening but as mentioned, Replit just makes the satisfaction of creating so addicting that I continued to build out and refine my app from there.

I’ve now got an app that I’m using and testing and have shared with a few friends to do the same. My hope is that for the friends going through difficult things, this is another way I can support them with insights and healthy reflection. 

As far as my experience with this particular vibe platform, in a world where the line between user and creator is blurring, Replit stands as both a catalyst and a companion. This will be a game changer allowing both seasoned engineers and first-time tinkerers a streamlined way to bring their ideas to life.

Lovable

Originally, my intention for this article was to review 3 platforms. But then, I stumbled across an article about the Swedish AI Unicorn “Lovable” and had to try that one out as well. 

Lovable’s roots stem from the stories of two individuals: Anton Osika and Fabian Hedin, both alumni of the KTH Royal Institute of Technology. They crossed paths at Depict, a YC-backed AI startup. There, they saw firsthand how even the most promising ideas could be bottlenecked by the scarcity of coding talent. In mid-2023, Osika built GPT Engineer, an open-source tool that let anyone generate code from simple prompts. The project exploded—54,000 GitHub stars and giving this idea empirical momentum.

Osika and Hedin launched a commercial version, “GPT Engineer App,” which quickly landed on the front pages of Product Hunt and Hacker News. Hundreds of paying users signed up overnight. By December 2024, the app had evolved and rebranded as Lovable, a name that hints at both the platform’s approachability and its ambition to make software creation something everyone can love .

Lovable is, at its core, an AI full-stack engineer. It’s a platform where you describe what you want to and Lovable generates a production-ready web application, complete with front-end, back-end, authentication, databases, and even payment integrations. You can start from scratch, use a template, upload a Figma design, or even clone an existing website’s layout with a screenshot .

What is interesting about their position as a company is the fact that Lovable isn’t about replacing developers, it’s about unlocking creative leverage for the 99% of people who have ideas but not the coding chops to know where to start in creating them. As Osika puts it, “Imagine if anyone with a powerful idea could turn that idea into a product. This is the vision we’re making a reality at Lovable.”

Lovable’s team is small but mighty—just 15 people as of mid-2025, including astrophysicists, sourdough bakers, and a national unicycling champion . Their culture is built on principles like “Driver not Passenger” (take ownership), “No Room for Ego” (collaborate and learn), and “Candor over Comfort” (give and receive direct feedback). 

Building on Lovable

Personally, I’m looking forward to seeing how this platform continues to develop. Because I already had a project that was reasonably well developed in Replit, I didn’t spend that much time on Lovable beyond a few basic prompt tests. From my limited exposure though, this platform is approachable and creates a good initial prototype. 

Notable is that I didn’t have any errors or issues with the first version, which in terms of coding is not often the case. My initial result was very simplistic but that isn’t necessarily a knock given that good vibe coding should be a building process, adding and testing one thing at a time.

Final Thoughts

What I appreciate about the Vibe coding tools is that you don’t have to know everything, or even really anything to get started. For those who want to use these tools intelligently, watch the development decisions that are made with your idea, learn from the agent examples, and use this as an opportunity to get deeper into a stack that is relevant to something you wanted to create. 

There’s no shortage of debate these days about whether everyone should learn to code. Here’s where I land: there’s real value in understanding what you’re looking at, even if you never plan to write a single line yourself. For many people, code feels like an impenetrable wall. It’s easy to lean back and disengage when it all looks overwhelming. But if you take a moment to study what’s in front of you, you’ll notice that code is, at its heart, logical. And sometimes even surprisingly readable, especially if you speak English.

You don’t have to become a developer overnight, but you can become someone who isn’t intimidated by technology. That is something in today’s world is its own kind of superpower.