The Promise of Low-Code
When Microsoft launched the Power Platform, I saw it as an Excel 2.0 for apps. The pitch was irresistible: business users could build applications, automate workflows, and connect systems — no programming required. The timing was perfect. Demand for software was exploding, while IT teams were stretched thin. Low-code seemed like the answer.
The growth numbers supported the hype:
- Power Platform reached 33 million monthly active users in 2023, according to Microsoft.
- Gartner predicts that 70% of new enterprise applications will be built with low-code/no-code platforms by 2025, up from less than 25% in 2020.
- Gartner also projected that citizen developers would outnumber professional developers by 4 to 1 in large organizations.
It felt like history was repeating itself: just as Excel democratized modeling, low-code would democratize app development.
Hitting the Ceiling
But after years of building real-world solutions in Power Apps, Power Automate, and Dataverse, I found myself hitting the ceiling:
- Performance: Apps that work for 10 users often collapse with 200, or even fewer.
- Licensing costs: unpredictable, complex, and hard to justify (AI Builder credit costs are a mystery, for example).
- UI/UX: clunky compared to frameworks like React or Flutter. No advanced effects or native responsiveness on Power Apps canvas.
- Cloud unpredictability: flows failing without explanation, delegation errors impossible to debug, and a frustrating expression editor.
And I realized it wasn’t just my impression. Other citizen developers with years of VBA experience admit to feeling more limited in Power Apps than in Excel. Others say being a citizen developer is “extremely stressful and unsustainable” because development is always a second job on top of business responsibilities.
Then came the high-profile failures: in 2021, according to Wired, 38 million sensitive records were accidentally exposed because of default Power Apps settings that allowed public access. It was a brutal reminder that democratizing development without guardrails creates serious risks.
The lesson: you can’t really become a part-time developer — at least not without adopting professional development practices.
The Alternative: Will Vibe Coding Fill the Gap?
Recently, while launching FinModeler, the limits became too obvious, and I began exploring an alternative tech stack:
- React for the frontend
- Azure SQL as the data backend
- Azure Functions as the business logic layer
This, of course, with the help of my AI friends: GitHub Copilot and Codex.
The experience so far has been extremely positive. With the same data, programming logic, and business utility of a Power Platform solution, the time it takes to go from zero to proof of concept is about the same as with low-code.
The difference is two-fold:
- Freedom. With vibe coding, there are no platform limits. I can design the architecture and implement the solution exactly as I envision it.
- Compounding experience. Since there are no hard limits, every hour invested pays off more over time. Skills accumulate rather than being boxed in by platform constraints.
And what do we lose by going vibe coding? Nothing.
You do need to understand what you are doing. It makes sense to learn the basics of programming languages, data, design, and good software practices. By that I mean adopting simple but powerful disciplines: using Git to track changes so you never lose your work, writing clean and modular functions instead of long tangled scripts, validating inputs to avoid corrupted data, documenting why you made a design choice so your future self (or your team) understands it, and testing the logic before it reaches production.
These are not abstract rules — they’re the difference between a fragile prototype that breaks under pressure and a solution that scales with your business.
The Tectonic Shift: AI and Vibe Coding
Something bigger is happening now: we are entering the era of AI-assisted development.
Excel will never die. It will remain the business user’s tool for decision-making. But low-code might — or it might transform beyond recognition.
Vibe coding, powered by AI, may become the next “bicycle for the mind” — combining business knowledge with software power, without the middleman. And Microsoft is already reacting. The Power Platform today is infused with AI at every level: in Power Apps, you can simply describe the app you want in natural language and Copilot scaffolds the screens and data model; in Power Automate, entire flows are generated from a single prompt; even Power BI now allows you to build DAX calculations or dashboards by asking questions in plain English. On top of that, new capabilities such as Power Apps Code Apps allow developers to blend traditional coding (e.g., React) with the Power Platform ecosystem, supported by GitHub Copilot and access to more than 1,500 connectors.
These are not incremental improvements — they blur the line between “low-code” and “pro-code,” or perhaps create something beyond both. That leaves the very notion of low-code in a strange place. If building an app no longer requires dragging controls or learning Power Fx, but instead describing your intent to an AI assistant or embedding full React components with Code Apps, can we still call it low-code? Or has the category already outgrown its name?
The Power Platform is evolving into something closer to AI-first vibe coding, where prompts, copilots, and code coexist seamlessly. Low-code may not die, but it is being redefined into something quite different from what it started as.
The essence remains the same: data, logic, design, and solving business problems.
