In the days of yore, becoming a coder involved years of study in college or university, or perhaps an intense bootcamp lasting months at a time that drilled participants in the basics of programming. Then artificial intelligence (AI) entered the picture.
Countless viral “hacks” to get code from chatbots began circulating shortly after ChatGPT’s debut back at the end of 2022, and the rest is history.
Since then, there has been a boom in AI-powered coding tools, with some from big names like Microsoft, Google, and even ChatGPT-maker OpenAI, which are promising to make programming more accessible than ever.
This has given rise to what is being described as “vibe coding”, which is loosely defined as a software development practice that generates working code through prompts written by users who might not have extensive technical knowledge – similar to how AI chatbots are being used to plan trips and give explanations.
Daren Tan, the founder of local IT community Developer Kaki, says that this is a phenomenon that has developers relying more on AI tools and autocomplete suggestions instead of systematically thinking things through on their own.
Tan says that vibe coding has developers relying more on AI tools and autocomplete suggestions instead of systematically thinking things through on their own. — DAREN TAN
“It’s like coding by feel. You start typing something, the AI suggests the next bit, and you go with whatever feels right in the moment. Tools like GitHub Copilot, ChatGPT, Cursor and Claude are incredibly powerful for this,” he says.
When it comes to approaches to vibe coding, Mafas Raheem, a senior lecturer at the Asia Pacific University of Technology and Innovation’s (APU) School of Computing, breaks it down into two categories: pure vibe coding and AI-assisted responsible development.
Pure vibe coding involves a developer working with an AI tool to generate code, engaging in a back-and-forth driven by their ideas, instincts, or simply the overall “vibe”, hence the name. This flow makes it especially useful for prototyping or quickly iterating on a project.
Meanwhile, AI-assisted responsible development involves more rigorous testing and quality assurance to ensure reliability, security, and maintainability.
“At its heart, this new paradigm powers a dynamic, conversational workflow. You describe what you want in natural language, the AI generates the code, you test the result, and the loop continues, where each iteration sharpens the product until it meets your needs.
“But the impact extends far beyond single functions or scripts. With a high-level prompt, AI tools can now generate entire applications, including front-end, back-end, and infrastructure that are ready for iterative refinement, expert review, and deployment at the click of a button,” Mafas says.
For some, this promise has become a reality. Full-stack developer Naman Iqbal, for instance, has experimented with several of such vibe coding tools after first coming across them in the early days.
“In the initial stages of my intro to ChatGPT in 2023, people were using it to fix their (programming) logic or convert one syntax’s code to another,” he says, adding that back then, it was very much in its infancy.
Pure vibe coding involves a developer working with an AI tool to generate code, engaging in a back-and-forth driven by their ideas, instincts, or simply the overall “vibe”, hence the name. — Pixabay
Now, he says it’s gotten to the point where it can fix bugs in code and include the proper logic for an intended function.
Naman’s not alone in this. Shee Tze Jin, an AI and machine learning specialist with Taylor’s University, shares that a strong community of vibe coders has already begun to take root locally.
One such community is Build with AI (BWAI), which Shee is involved in. While other communities tend to skew towards a more technical crowd, BWAI is exclusively for those without a background in coding, but who have ideas that they want to bring to life.
Run via a free private chat group where members can seek advice and share their challenges, he explains that each person is screened first to understand their goals and ensure they’re a good fit for the community.
“We host weekly online Office Hours (a collaboration and knowledge sharing session) every Tuesday night, where members can learn about new tools, ask questions, and get direct support from experts.
“Recently, we also introduced Co-Working Fridays for AI Builders at the Asia School of Business, offering free coworking space for members to collaborate, experiment, and work on their AI-driven ideas together,” he says.
Open access
From Mafas’ point of view, the greatest strength that vibe coding has is in lowering the barrier of entry to the field as a whole. This means that students get the chance to turn their ideas into working prototypes and gives them a real sense of achievement.
“This early exposure can ignite passion, encourage creative problem-solving, and help students see the big-picture applications of software, rather than getting bogged down by syntax and setup.
“For non-traditional learners, hobbyists, or those just starting, this accessibility is transformative,” he says.
Mafas says that vibe coding tools can act as a supportive companion for a student’s software development journey when used responsibly, giving them the chance to experiment and perform rapid iterations on projects. — APU
This is echoed by Shee’s involvement in BWAI’s community outreach efforts, noting that people with ideas but lacking the technical skills previously needed to jump through a multitude of hoops just to get their ideas off the ground.
Budding entrepreneurs often had to bring in a team of developers to make up for their lack of know-how, a process that could be uncertain, expensive and time-consuming.
“For beginners who have little to no coding experience, I usually recommend starting with Lovable. It’s very beginner-friendly and, as long as you have everything properly set up (eg, GitHub and Supabase), it can quickly build something functional for you.
“However, once users start to feel the limitations of Lovable and want more control, I suggest switching to Cursor (the most popular AI Integrated Development Environment or IDE).
“Using Cursor alongside a proper IDE not only provides greater flexibility for editing and customising code at a granular level, but it’s also a more cost-effective approach in the long run,” he says.
He further encourages those with ideas still on the sidelines to get involved with the community, especially since becoming a member and participating in activities is usually free of charge.
“Today, the barrier to trying AI is lower than ever – you can often build your own idea for the cost of just three to five meals. The cost of failure is equally low, which means you can afford to experiment, fail, and try again until you succeed.
“If you’re curious and have the time, there are countless tutorials on YouTube that can help you self-learn.
“And if you’d rather learn alongside like-minded people, join one of the many AI vibe coding communities such as BWAI, AI Tinkerer KL, or Builder Club,” he says.
Experts believe vibe coding and its tools still have a place, as long as they’re used responsibly and with the right precautions in place. — Pixabay
But despite the warm reception so far, it hasn’t come without its downsides.
Co-pilot or on autopilot?
From Tan’s perspective, a significant limitation is that “you’re essentially outsourcing your thinking process. The AI might give you working code but you might not understand why it works or how to debug it when things go wrong.
“I’m not against AI tools. They are incredibly useful and can boost productivity significantly. But there’s a difference between using AI as a smart assistant versus using it as a crutch.
“The concern is when developers start relying on these tools so heavily that they lose the ability to think through problems, solutions, design and quality independently.
“It’s like having GPS for every journey – convenient, but you never learn to read a map or understand the terrain,” he says.
Shee says that with how much faster vibe coding is in development, fundamentals like thorough testing become all the more important. — IZZRAFIQ ALIAS/The Star
Looking at things from the academic side of things, Mafas believes that overreliance on such tools can introduce some real risks.
This comes in the form of a failure to cultivate key skills, which include coding fundamentals, problem-solving, and the discipline needed for professional software engineering.
This is something Tan, who is also the CEO of IT firm Alphv Technologies and a frequent hackathon judge, agrees with.
He has observed a mixed impact on participants, with AI tools both allowing them to produce an output that looks polished at first glance, but often without a deeper understanding of their submitted code and the reasons behind certain features.
“On the positive side, AI tools help participants move faster and tackle more complex features they might not have attempted before. I’ve seen teams build impressive-looking demos in record time.
“But here’s the thing… when I dig deeper during judging, I often find that participants can’t explain their own code or struggle to modify it on the spot. The output looks polished, but the learning is superficial,” he says.
Mafas further points out the technical and security downfalls that may be overlooked due to a lack of technical know-how.
“AI-generated code can contain subtle but serious security flaws, such as injection vulnerabilities, insecure data handling, or broken access controls that may go unnoticed by beginners.
“Spotting and fixing these issues requires a solid grounding in secure coding principles, something that vibe coding alone cannot teach,” he says, adding that generated code may also be messy and hard to understand, further causing difficulty in maintaining projects down the line due to a lack of proper documentation.
But even with those shortcomings, experts believe vibe coding and its tools still have a place, as long as they’re used responsibly and with the right precautions in place.
For instance, Shee says that with how much faster vibe coding is in development, fundamentals like thorough testing become all the more important.
“With vibe coding producing code at a much faster pace, building unit test cases becomes even more critical to ensure everything works as intended.
“Since AI can sometimes ‘hallucinate’ or generate incorrect logic, a solid testing methodology is the only way to consistently catch errors,” he says, adding that maintaining good backups and version control is just as critical in case something goes wrong, whether from human error or AI-generated code.
Noting the downside of vibecoding, Shee says that similar concerns were noted with the rise of Google Search, Wikipedia, or even digital phone books.
“People once argued that relying too heavily on Google would prevent students from truly learning from books.
“The same argument applies here. Yes, there are valid concerns that some coders may struggle to write high-quality code without assistance. However, the benefits that vibe coding brings to society far outweigh these risks.
“Vibe coding has effectively turned the once tedious and complex process of coding into a commodity – making app development accessible to anyone with an idea,” he says.
Mafas too says that vibe coding tools can act as a supportive companion for a student’s software development journey when used responsibly, giving them the chance to experiment and perform rapid iterations on projects.
“Under APU’s coding policy, students are encouraged to use AI-assisted tools as learning aids, provided these tools are used to deepen understanding, accelerate knowledge acquisition, and foster higher-order problem-solving skills,” he says.
Likewise, Tan positions them as fulfilling a more supportive role, helping to cover blind spots and enhance a programmer’s final output, rather than being the deciding factor in what that output should be like.
“My advice is simple: use AI tools, but don’t let them think for you. Try solving problems on your own first, then use AI to help optimise or catch things you missed.
“Understand what the AI-generated code does before using it. And please, focus on solving real problems rather than cramming in every cool technology you can find.
“The best programmers I know aren’t the ones who can use the most tools. They’re the ones who can think through complex problems systematically and build solutions that actually matter to users.
“Genuine technical skill shows up in how someone thinks through problems and adapts, not just in their final code. We need to distinguish because ultimately, we’re trying to identify and nurture problem-solvers, not just code generators,” he says.
