Programming has always been a slightly difficult skill to acquire, but now, with AI, anyone can be a vibe programmer. Here is how you can be one
Vibe coding is the next frontier
Vibe coding is a new buzzword and may be your ticket to a better job that involves actively participating in the creation of software rather than being a user. Vibe coding refers to the skill of having Large Language Models (LLM) and specialised programming software suites that let you program what you need with just prompts. Even Andrej Karpathy, who co-founded Open AI, tweeted about it recently.
ADVERTISEMENT
If this sounds like a fun task with the LLMs and software where they are right now, it is still a task that a proper programmer would rather not do. However, if you stick to it, there are jobs in the future that will need this skill or even be the main job profile itself. While most programmers feel threatened by this tech, I must point out that people who are skilled at programming would still have a function further up the food chain as supervisors. Imagine a company where vibe coders are the majority of coders, and you, as a programmer, are the supervisor. Of course, this means that even as a professional coder you would need to know how vibe coding works.
How vibe coding works
To be a vibe coder, rather than understanding the code, you need to know what the program needs to do, and you have to coach the LLM (Large Language Models) or AI-based IDE (Integrated Development Environment) to do your bidding. Getting a perfect result every time right now is very difficult as we are at the cutting edge. This means you may need to work a little harder than people in the future who will probably be able to get better results as technology gets better. This also means you will need to know a bit of coding and learn to think like a programmer, even if you don’t actually program.
The coding part that is needed is the ability to debug; this means while you may not know the syntaxes and the codes by heart, you can still logically read through the code and figure out what it is doing wrong. This requires a fair amount of practice and some basic programming knowledge. You may want to do some of those quick learn X programming languages in X-hour-type courses on YouTube. If you can already think logically, these courses should be super easy to get into.
Tools of the trade
Get familiar with the programs you will need to use for this. The best LLM options for vibe coding are Claude AI, DeepSeek R1, and ChatGPT. Alternatively, you can also set up your own local LLM using LMStudio or Ollama. However, hosting your own LLM takes significant resources, such as investing in a graphics card and setting everything up.
For a coding environment, you can use Cursor AI, Windsurf, Continue or Github Co-pilot. There are also direct app builders like Bolt, Replit and Lovable. They essentially create the entire application for you from scratch using a prompt. These are ultimately where you want to get with vibe coding so it is worth exploring these options while on your coding journey.
How to start
As is with anything in life, practice makes perfect. However, you will want to start small and move up. First decide what it is you want to code is it web apps, mobile apps, software etc, this will decide the programs you want to concentrate on. For example, if you want to program a mobile app there is no point in learning and using Python. So, research the field you want to program in and start there.
Find programming challenges and start small coding challenges, moving to more complex ones as you get better at extracting the code from the AI IDE or LLM. There are plenty coding challenges that you can find Google. Once you are confident in solving almost any challenge, search for the things you want to specialise in and find projects related to that on Github that you can contribute to.
Alternatively, you can just make your own project if you have an implementable idea. Vibe coding is all about experimenting and getting an imperfect AI to cough out code that works. As AI improves, this code will get better and you will know exactly how to get this done.
Links
Claude AI: Claude.ai
Deepseek R1: groq.com
ChatGPT: chatgpt.com
LM Studio: lmstudio.ai
Ollama: ollama.com
Cursor AI:
www.cursor.com
Windsurf: codeium.com/windsurf
Continue: www.continue.dev
Github Copilot: github.com/features/copilot
Bolt: bolt.new
Lovable: lovable.dev replit.com
Subscribe today by clicking the link and stay updated with the latest news!" Click here!
