Vibe coding has lowered the barrier to software development. What next?
Friday, November 20, 2026
AI tools may make it easier to turn an idea into an app. Knowing what happens behind the screen, how to protect users and how to keep the product working is a different skill set. COURTESY

Social media is full of videos promising that almost anyone can now build an app with artificial intelligence, publish it on an app store, and start making money from downloads. The process often looks simple: describe an idea to an AI tool, let it generate the code, and watch your product take shape.

There is some truth to that promise. But there is a part the videos tend to leave out: getting an app to work is different from making sure it is secure, reliable, and safe for the people who use it.

ALSO READ: Rwanda Coding Academy shortlisted for $1m Global Schools Prize

The practice, popularly known as vibe coding, involves using ordinary language to instruct AI tools to generate software code. It has lowered the barrier to experimenting with app development, particularly for people without formal programming experience.

However, the less a developer understands the code being generated, the harder it can be to spot problems that could put users and their information at risk.

Trust is a big part of choosing software. Before using an unfamiliar app, many people want to know who built it, what information it collects, and whether their details are secure. Those questions become harder to answer when the person who built the app may not understand how it handles sensitive information.

ALSO READ: So, what is the big deal about coding? Everything

I spoke with a DevOps professional, a specialist who helps teams develop, deploy and maintain software, about the risks emerging from AI-assisted development. Among the examples we reviewed were published applications with security weaknesses that could expose secret keys, activity records, cached data and, in some cases, payment information.

One case involved a developer who had placed an OpenAI application programming interface (API) key in publicly accessible code. An API is a connection that allows different software systems to communicate. The key essentially gives an application permission to use a service, sometimes at a cost.

Because the developer had not set a spending limit, automated bots found the exposed key and began consuming credits. The developer only realised something was wrong when a bill running into thousands of dollars arrived.

ALSO READ: "Rwanda’s coders have a future” despite rise of AI – experts

Fixing the problem required more than deleting the key from the code. Once a secret key has been exposed, someone may already have copied it. The engineer had to replace the compromised key, set a spending limit, and move calls to OpenAI and Stripe into Supabase Edge Functions. These are small server-side functions that handle tasks away from the user's browser, keeping sensitive credentials out of publicly accessible code.

The lesson: AI can help people write software, but it does not automatically give them the knowledge to protect it. Understanding basic security practices, operational security, is an important part of the learning curve for aspiring developers.

ALSO READ: University of Kigali opens registration for AI and cybersecurity certification programmes on November 16, 2026

That is one area structured training can address. GIZA Institute's February 2027 intake offers two six-month Certified Professional Short Courses: Full-Stack Software Development and AI and Machine Learning. Both programmes begin February 1 at the institute in Kicukiro-Niboye, with in-person practical sessions.

The Full-Stack Software Development course is aimed at students, graduates, and working technology professionals who want to develop the ability to build complete software applications. Full-stack development covers both the frontend, which is what users see and interact with, and the backend, which handles the application's logic, data and services.

The curriculum includes programming foundations, React for building user interfaces, backend engineering, APIs, databases, authentication, security, testing, and deployment. Learners also work on a capstone project, allowing them to demonstrate what they have learned.

The AI and Machine Learning course takes a different route, starting with Python programming and data processing before moving into machine learning, deep learning, and the deployment of AI applications.

Machine learning is a branch of AI in which systems learn patterns from data to make predictions or decisions. The programme also introduces MLOps, the practices used to build, deploy, monitor and improve machine-learning systems.

Learners will use tools including Python, Jupyter, Google Colab, GitHub and Streamlit. The course includes a capstone project based on an industry or community problem, alongside training in portfolio development and entrepreneurship.

Applicants need an Advanced Level secondary school certificate or equivalent, basic computer literacy, English communication skills, and a personal laptop. An 8GB RAM laptop is recommended, but previous programming experience is not required.

Training is available on weekday mornings, weekday evenings, and weekends to accommodate different schedules. Applications are being accepted on a rolling basis and will close once all available places have been filled.

This demand for practical skills is reflected in the findings of the University of Rwanda’s 2025 Labour Market Study, which surveyed 5,923 participants. While 71 percent of employers reported high demand for specialised skills in artificial intelligence, cybersecurity and data management, 90.9 percent of alumni expressed a preference for practical, hands-on learning.