vibe coding — this term coined in February 2025 suddenly got so much traction and there are lot of discussion around this. Even books are getting published on this topic. But what is it? Is it really going to influence the way programmers write coding ? Does the developer do not need to validate the code once created? First let us understand what it is and explore in depth on this topic.
Vibe Coding is where developers primarily write code by interacting with code generating large language models rather than writing code directly.
How does it differ from AI assisted Coding?
I was working on Github Copilot earlier, how is vibe coding different from it. The answer from the research paper is copilot was autocomplete assistant where vibe coding is agentic where AI acts as a collaborator and not a tool. The responsibility shifted from the developer to steer and correct instead of micro manage every line. So which means when I use Github Copilot agent without monitoring the code it has generated, just accepting it with trust is vibe coding.
In wikipedia, they say key part of the definition of vibe coding is that the user accepts code without full understanding. AI researcher Simon Willison said: “ If LLM wrote every line of your code, but you reviewed, tested and understood it all, that’s not vibe coding in my book — that is using an LLM as typing assistant.”
Popular Tools for Vibe Coding
- GitHub Copilot and its Agent Mode within VS Code, Visual Studio and other IDEs (e.g. JetBrains IDEs, Eclipse, Xcode)
- Cursor Chat (previously: Compose) in Cursor
- Cascade by Windsurf, within Windsurf and JetBrains IDEs
- Cline: the “collaborative AI coder” VS Code extension
- Junie within JetBrains IDEs
- Augment Code with support for VS Code, Jetbrains IDEs, and others (including Vim)
Not only IDE we can also use command line agentic tools
- Claude Code by Anthropic
- Codex by OpenAI
There are few design related tools like
- Google Stitch: turn prompts into UI designs for mobile and web
- Figma Make: for creating prototypes from ideas. Start prototyping by sketching a design
- Anima: converts Figma designs to code
There are full stack websites that build and deploy the complete web applications
- Vercel V0 — Good visual prototypes
- Replit — End to end website
- Lovable — For web apps
- Bolt.new — For mobile apps
Combined with the growing popularity of AI-powered assistants and editors, vibe coding’s rise in the enterprise reflects a significant shift in how quickly apps are being conceived and delivered, with many implications for professional developers.
The role of the engineer is still very, very critical to make sure that the boundaries and conditions are set up front for what the vibe coding is going to produce. It doesn’t excuse the engineer from needing to understand what’s going on behind the scenes.
Practical Example
I did built a simple to do list web app using Vibe coding using 3 steps
Step:1 — Prompt
“Create a simple web-based To-Do List app using HTML, CSS, and JavaScript. It should allow users to add, delete, and mark tasks as completed. Make it mobile responsive and clean looking.”
It generated me an HTML that has the components
Step 2: Try the App
- Paste the code into an .html file.
- Open in browser.
- Works ✅
Step 3: You Spot a Bug or Want a Feature
“Can you make the tasks persist even after refreshing the browser using localStorage?”
It has modified some javascript functions to store on local storage and added render tasks function to fetch from local storage.
You’re still guiding the project, reviewing outputs, and requesting changes — but you’re not writing most of the code yourself. The AI becomes a collaborator, not just an autocomplete tool.
Let us think it through —
From all that I have read from the articles and across the web there are multiple opinions. Here are some of the questions we can answer
Is Vibe coding really programming or just prompt engineering
It is actually both. Programming is we write the logic ourself where in vibe coding we write prompts and let AI do the logic. Our job switches from programming to structuring prompts, reviewing results and correcting AI when it fails.
Vibe coding does not eliminate expertise — it repackages it as steering and evaluating AI
How much control are we giving away to AI?
Trust is not static. It is adjusted per experience and depends on the criticality of task.
Some developers simply accept the code it has generated without reviewing it. Some developers reviews each output, tweak prompts and manually fix bugs. But it depends on the programmers on how lenient we should be with the code generated. If the code is critical and should not let a small bug disrupt the working then I would say be careful and test it thoroughly before taking the decision that the code should run in production. It is ok to let the AI generate the boilerplate code which can save sometime for the developers to work on but be very mindful about accepting the code for business critical applications.
Does vibe coding encourage speed over quality?
Yes. With the current world where the thoughts in research papers published last month is turning into production grade applications in just a month. The companies should rethink on how faster the user expects the output to be seen in this fast moving world. Definitely, vibe coding can speed up the development of the applications. But does it has to sacrifice the quality? There is always tradeoff between speed and quality. It should be the business decision to decide what is important.
Vibe coding is great for fast prototyping but not a substitute for good engineering practices.
Can we trust this process for building apps used by real users?
The straight answer is as per our discussion earlier, For MVPs or internal tools, we can use yes. but for production grade applications it is not recommended.
Most of the research papers I read about this vibe coding tried on MVP or prototypes but AI generated code still needs review. This is because the security, scalability and performance is not validated.
Is it creating new barriers?
For a lower level engineers, you dont need to memorize syntax and even a non programmers can build a prototype. But without proper prompting skills and understanding what the AI has generated, without having knowledge you cannot correct the AI generated code. This will lead to poor quality software delivered.
A good coding sense + product knowledge and the skills to instruct it step by step with the clear vision of the output can give better results. These are all new set of skills still needed for vibe coding.
Hope this article has given a good sense of vibe coding!!
