What is Vibe Coding?
Is this the start of a new entrepreneurial paradigm? Frank Andrade: How I’d Learn Python in 2025 (ft. ChatGPT, Gemini 2.0 and NotebookLM)
Good Morning,
I’m noticing two interesting ideas that are going to evolve in parallel in the future of work.
One represents AI reasoning models that will utilize agentic workflows to iterate on how we have done science traditionally (for example, science here but could be anything), like Google’s recent AI Co-Scientist, and the other a new kind of kind of vibe worker in the future of work that is enabled by a new set of tools.
Think about this for a moment if you will:
What’s clear is AI changes how we will be iterating along the lines of R&D but also how we build companies, products and “create” fundamentally as human beings. ⚗️
Google’s AI Co-Scientist launched on February 19, 2025, is an innovative system designed to assist scientists in generating novel hypotheses and refining research proposals, ultimately accelerating the pace of scientific discovery.
It’s not clear if either of these two approaches will radically alter the future of work, but they could certainly continue to evolve in the 2020s along very interesting directions.
Andrej Karpathy’s now famous vibe coding tweet follows:
The tweet has 4 million views, why do you suppose that is?
“There's a new kind of coding I call "vibe coding", where you fully give in to the vibes, embrace exponentials, and forget that the code even exists.” - Karpathy, Feb, 2025
What is Vibe Coding?
Vibe coding is a new approach to software development that utilizes AI tools to assist individuals in creating applications and software without requiring extensive programming knowledge.
The term was popularized by Andrej Karpathy, an AI expert, who described it as a method where users interact with AI using natural language to describe their ideas rather than writing traditional code directly.
This allows creators, particularly those lacking technical skills, to build functional applications rapidly by simply explaining their requirements to the AI, which generates the relevant code for them.
AI-Assisted Iteration: How Vibe Coding Works 🔥
The process of vibe coding typically involves several key steps:
Describe Your Idea: Creators start by articulating what they want to build in clear, descriptive language.
AI Code Generation: The AI takes these descriptions and generates the necessary code.
Iterate and Refine: Users can request modifications and improvements by communicating additional requirements or asking for fixes. The AI can adjust the generated code in response to feedback.
“Historically, turning these vibes into something tangible—a detailed plan, a memo, a piece of code—has been a slog. That final 20% of clarity often demands 80% of the effort, as we wrestle our intuitions into structured form. But AI is changing that.” - Azeem Azhar, of Exponential View
Is Vibe Coding real?
As of 2025, it does seem like Vibe coding is indeed a real concept gaining traction, particularly in the realms of non-technical software development.
In a nutshell: Vibe coding democratizes software development by allowing those without coding expertise to turn their ideas into digital products.
When having an idea can be enough…
This is a new kind of No-Code Entrepreneurship
So how does the improving of Generative AI in software engineering change how humanity relates to code?
Traditional Coding
Skills Required: Programming languages, frameworks, debugging
Development Process: Write code line by line, debug, test
Time Investment: Weeks/months to learn, days/weeks to build
Flexibility: Unlimited (can build anything)
Relationship to Code: Direct manipulation of all code
Low-Code Platforms
Skills Required: Basic understanding of programming concepts
Development Process: Drag-and-drop components, configure settings
Time Investment: Hours/days to learn, days to build
Flexibility: Limited to platform capabilities
Relationship to Code: Abstracted, rarely see the code
Vibe Coding
Skills Required: Natural language communication skills
Development Process: Describe what you want, iterate through conversation
Time Investment: Minutes to learn, hours to build
Flexibility: Extensive but with limitations on complexity
Relationship to Code: The code exists but you may never read it
(Credit: Jacob Anderson)
Now I’m not a techno-optimist like Azeem Azhar or many of the X AI bros, but I think over time as AI gets better at coding and iterating, it will lower the bar to some kinds of innovation and entrepreneurship. This as our reasoning models get better at managing AI agents and their evolving workflows. Vibe coding and the vibe worker is essentially an extension of our own BYOAI approach to elevating our game with AI.
I don’t think Vibe coding represents the next step in the evolution of software development where we describe intent rather than write code, but I think No-code is going to evolve with AI and how we are augmented by it will change radically over the next decade. Vibe coding suits GenZ and Alpha cohorts values alignment towards new forms of entrepreneurship and the search for more autonomy, flexibility and work-life balance and trends we are seeing like a four day work week.
Thus vibe coding as a trend will enable and make solopreneurship more viable and make self-employment more flexible, automated and potentially more profitable. Imagine a world where it will be more viable to start a small business, with less risk? Keep in mind Generative AI has been harming the demand for many kinds of freelancers, so we need to adapt. Vibe workers might realize the new minimum viable products (MVPs) they can build suddenly! Suddenly because, what you can do today wasn’t possible previously perhaps even a few months ago.
I think vibe coding has relevance to how Creators build flywheels and diversify their revenue sources. We have to be realistic though, for most and mainstream adoption this is a long way off from today’s tools and agentic frameworks. Claude Code has enormous potential, and with Anthropic expanding globally in the 2025 to 2030 period, they are the company most likely to make vibe coding and vibe entrepreneurship a reality. Many of the high-flying AI startups to reach 100 million ARR in record short time for example, are based on Claude’s technology and relate to the vibe coding trend directly.
If you are interested,
explains even how to vibe code. Naturally, I expect to see a lot more articles and guides on vibe coding and all that it might mean in terms of Agentic AI and new advances in how we use Generative AI and AI tools to iterate, build, create and bring ideas into reality.There’s something minimalistic about vibe coding that’s also endearing somehow:
Benefits of Vibe Coding
Accessibility: Vibe coding democratizes software development, enabling non-programmers to create tools and applications that meet their needs without having to learn complex programming languages.
Speed: This method significantly reduces the time taken to develop an application. What once required weeks or months can now be achieved in hours or even minutes.
Cost-Effectiveness: Reducing the need for hiring developers for simple applications can be financially beneficial for individuals and small businesses.
However Generative AI is also enabling software engineers to be more productive, which has more tangible and immediate value for technology and startups. Even the process of learning to code is changing. As we explore AI tools and how we can integrate AI agents into our workflows vibe coding suddenly becomes a lot more viable. Whatever you think of tools like Cursor and Loveable, AI tools will keep getting better and perhaps engineer a viable path for vibe coding, especially if the trend gains momentum.
Medium alum Frank Andrade’s newsletter is fascinating in this respect. He’s Pycoach, of the Newsletter Artificial Corner.
Behind AI #1: Machine Learning Algorithms Any AI Enthusiast Should Know
Behind AI #6: The 4 Stages of Learning Python for AI & ML
Master The Perfect ChatGPT Prompt Formula
I Tried Multiple AI Coding Assistants. These Are The Best
Video Biography
Frank is based out of L.A. He’s a YouTube and a Newsletter writer and one of the most promising young creators I know involved in AI. (140k Medium, 40k YouTube, 40k Substack). I think you’ll see why I’m bullish about his work.
Introducing Artificial Corner and The PyCoach
He’s also really good at doing tutorials and guides:
As you can imagine, he’s a wizard especially at making Python tutorials.
How I’d Learn Python in 2025 (ft. ChatGPT, Gemini 2.0 and NotebookLM)
In today’s guest post Frank gives us some valuable advice on learning Python in an era of Chatbots and AI tools.
He also shows you how to use NotebookLM to learn code like Python.
He also demonstrates using Github Copilot
We also continue our exploration of Vibe coding
I also list the AI tools most relevant to vibe coding prototyping
By
(The PyCoach)Learning to code will be different in 2025.
Keep reading with a 7-day free trial
Subscribe to AI Supremacy to keep reading this post and get 7 days of free access to the full post archives.