In February 2025, Andrej Karpathy, co-founder of OpenAI, posted a tweet that became something of a manifesto. "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." (source)

A month later, Merriam-Webster added "vibe coding" to its dictionary as a trending term. Five months after that, Jack Dorsey, the founder of Twitter, built Bitchat over a weekend. A fully functional messaging app. Bluetooth mesh. No internet. No phone number. No email. Dorsey used Goose, Block's AI coding tool. The app hit the App Store three weeks later.

But here's the thing. Vibe coding isn't really about programming. It's the first visible symptom of something much bigger. The same pattern, curiosity plus attentiveness plus first principles plus knowing how to ask the right questions, is already showing up in science, marketing, and business. AI isn't devaluing expertise. It's opening doors to fields where regular people simply couldn't get in before.

Vibe Coding: The Barrier to Entry Has Dropped Incredibly Low

Three years ago, building a working application required knowing a programming language, frameworks, deployment infrastructure. Git, CI/CD, cloud providers. Today? Describing the task in plain language is enough. Google Cloud defines vibe coding as "a workflow where the primary role shifts from writing code line-by-line to guiding an AI assistant through a more conversational process."

Fortune quotes a Silicon Valley CEO: "Vibe coding lets 10 engineers do the work of 100." But what's more interesting is that it lets people without engineering backgrounds do work that previously only engineers could.

Let's not exaggerate though. The barrier hasn't dropped to zero. You still need to understand what you're building, why, and for whom. You need to evaluate the result. You need to notice when AI generates beautiful but broken code. But the barrier has dropped so low that what used to take years of study now takes hours of attentive work.

Dorsey is a programmer. But his Bitchat matters not because he wrote code. It matters because he didn't write code. He described an idea. Bluetooth mesh chat. Peer-to-peer. No servers. And AI built it over a weekend. If the founder of Twitter can forget about code and just "vibe," what's stopping a product manager, a designer, or an entrepreneur with no technical background?

Nothing, really. I wrote about this at length in Business of Software is Dead, where I dug into the Stack Overflow data showing that while 84% of developers use AI tools, 72% say vibe coding is not part of their professional work. The split is real: commodity software is dying, complex software is thriving.

Arxiv published an academic paper on vibe coding, describing it as "an idealized style of programming using agentic software development tools" like Cursor, GitHub Copilot, Windsurf, and Bolt. This is no longer hype. Academia studies it. Industry uses it. And it's changing who can build technology products.

Vibe Science: Peach Tea and First Principles

If vibe coding lowers the barrier in software, what's happening in science? Turns out, exactly the same thing.

The terms "vibe science" and "vibe research" are already alive and getting debated seriously. The European Nexus for Strategic Intelligence describes vibe science as "turning discovery into an AI-native process: autonomous hypothesis generation, simulation, and integration that collapses scientific time and expands what humanity can know."

This isn't theory. It's practice, with real results to point to:

AlphaFold and the democratization of biology. DeepMind's AlphaFold cracked the protein folding problem. Biology had been stuck on it for 50 years. Nature reports that over 5 years, AlphaFold helped researchers worldwide predict the 3D structures of hundreds of millions of proteins. In February 2026, Isomorphic Labs announced "AlphaFold 4", an even more powerful model aimed at drug discovery. This kind of research used to require a laboratory, a PhD, and years of work. Now the AlphaFold Server is free. Any biologist, or even someone who isn't a biologist, can ask a question about protein structure and get an answer with near-experimental accuracy.

A solo physicist and black holes. Science News (February 2026) tells the story of Alex Lupsasca from Vanderbilt University. Working alone, with AI tools, he discovered new symmetries in the equations governing the shape of a black hole's event horizon. Not an international collaboration like CERN. Not a team of 50. One physicist with the right questions and the right tools. When he later met OpenAI's Chief Research Officer Mark Chen, both realized they were looking at the future of scientific discovery.

The AI Scientist by Sakana. In 2025, Sakana AI created something called The AI Scientist. A system that autonomously generates scientific ideas, runs experiments, analyzes results, and writes complete scientific papers. One of those papers passed double-blind peer review at an ICLR 2025 workshop. It was published in Nature. The first peer-reviewed scientific publication entirely created by AI. From hypothesis to final text.

Vibe research as a discipline. Owkin, an AI-for-biology company, puts it simply: "A scientist sets a problem, then the AI reads papers, generates ideas, runs analyses, and writes up findings. You don't have to be a coding genius to write code, and you don't need to be an expert in every field to work with it." This isn't replacing the scientific method. It's amplifying it. AI handles the routine: literature search, data processing, initial analysis. The human does what humans do best. Asks the question "what if?"

The World Economic Forum sums it up: "Democratized AI tools that even non-computer scientists can use will be key to lowering the barrier to entry for innovation."

So picture this. Sipping peach tea, asking the right questions, drilling down to first principles. And inventing new things. This isn't a metaphor. It's a description of a new workflow. And what's fascinating about it is that the key skill isn't knowing the answers. It's the ability to notice things and ask about them.

Vibe Marketing and Vibe Business: Strategy Without an Agency

MarTech (January 2026) already put the term into circulation: "Vibe marketing is using AI and no-code tools to turn plain-English ideas into live campaigns faster than traditional teams and stacks can keep up."

Think about what this means for small business. A startup founder from, say, Novosibirsk who makes a great product couldn't compete in marketing with a company that has an agency budget. That was just the reality. Today, they describe their audience, product, and goal. They get a strategy, content plan, copywriting, and analytics. Not perfect. But good enough to start. And often more accurate than what an agency delivered for $5,000 a month, because the founder knows their product and their customer better than any external marketer ever will.

PwC confirms this in its report: "AI-driven insights can identify untapped customer segments, uncover hidden product opportunities, or suggest new go-to-market approaches. Just as importantly, they democratize capability."

Demand Gen Report marks the tipping point: "For the past several years, AI sat alongside the marketing tech stack as a helpful tool. But in 2025, it moved into the center." Marketing stopped being a field where you need years of experience to launch your first campaign. It became a field where you need clarity of thought, understanding of your customer, and willingness to iterate fast.

The same goes for business strategy, financial modeling, legal review, writing business plans. I explored how this plays out for small businesses in Ant In A Jar, where the same principle applies: AI doesn't replace the business owner's knowledge of their market. It removes the technical barrier that used to sit between their idea and execution. None of these doors were open to non-specialists two years ago. Today they are. Not wide open. But open enough to walk through and begin.

The Social Media Paradox: Filter the Slop, Not the Authorship

Here's where it gets really interesting. And troubling for the major platforms.

Vibe marketing produces content at industrial scale. Social platforms react the only way they know how. By blocking.

The numbers tell the story. WIRED (March 2026) reports that X suspended 800 million accounts over 12 months. LinkedIn removed 200 million bots, which is 16.7% of its entire user base. In just the first half of 2025, another 83.7 million. X launched new bot detection tools in February 2026. LinkedIn replaced its entire algorithm with an AI system called 360Brew and is actively suppressing automated engagement.

But they're fighting the wrong enemy.

When AI generates content that looks identical to human-written content, asking "who wrote this, a human or a machine?" is the wrong question. The right question is: "Does this content deliver real value to the community?"

WIRED tells a great story about this. An AI agent "conquered" LinkedIn. Built an audience. Generated engagement. Got invited to give a corporate talk. Then it got banned. Not because its content was bad. Because it was automated. But if the content is useful, if it sparks discussion, if it helps people, does it really matter who or what wrote it?

Platforms need to stop filtering by authorship and start filtering by value. Block the slop. The meaningless, generic, spammy stuff. Not because a bot wrote it, but because it carries no meaning. Valuable content written by AI should stay, just like valuable content written by a human using Grammarly or a marketer using ChatGPT already stays.

This isn't a technical problem. It's a worldview problem. The platforms that figure this out first, that start evaluating content by what it actually contributes to people and conversations, will win. Those that keep fighting automation itself will lose. This is an irreversible social trend. They can either lead it or get run over by it.

The Curious AI Director: A New Profile

What do Dorsey building Bitchat, physicist Lupsasca exploring black holes, and a startup founder building marketing without an agency have in common?

None of them used AI as a replacement for their expertise. Each used it as an amplifier of three things: curiosity, attentiveness, and the ability to think from first principles.

Here's why those three specifically.

Curiosity is the engine. AI provides answers. But answers are meaningless without questions. And not just any questions. The right questions. "Build me an app" yields a generic result. "What would a messenger look like that works without internet, over Bluetooth mesh, and requires no personal data?" That question yields Bitchat. Curiosity is the ability to see possibilities where others see the status quo. AI turns that from an abstract personality trait into a practical tool. A curious person with AI can test 10 hypotheses in a day. Without AI, they'd test one in a month.

Attentiveness is the quality filter. AI generates. A lot. Fast. But it generates plausible nonsense with the same ease as brilliant solutions. The ability to notice that a result looks right but is wrong, that's the most critical skill in working with AI. An attentive person sees that generated code doesn't handle an edge case. That marketing copy repeats clichés. That a scientific hypothesis sounds elegant but isn't supported by data. Without attentiveness, AI becomes a generator of confident-looking errors. With attentiveness, it becomes the most powerful tool for verification and iteration you've ever had.

First principles are the navigation. When AI can do anything, the question "what exactly do I need?" becomes central. First principles thinking means setting aside "how it's always been done" and asking "why is this needed and what's the most direct path?" Dorsey didn't try to copy WhatsApp. He asked whether people can communicate without internet at all. And got Bluetooth mesh. Lupsasca didn't reproduce others' research. He asked a fundamental question about symmetries and found something new.

These three qualities together form what I'd call the Curious AI Director. Not a programmer. Not a marketer. Not a scientist. A person who directs AI through curiosity, filters through attentiveness, and navigates through first principles.

The Stanford HAI AI Index 2025 shows that AI infrastructure costs are falling 30% per year. Energy efficiency is improving 40% per year. The gap between open-weight and closed models has narrowed from 8% to just 1.7%. The barrier to entry isn't falling at a steady rate. It's falling exponentially. This is why AI adoption isn't following Moore's chasm the way previous technologies did. The adoption curve is too steep, too fast, and too broadly accessible for the traditional early-adopter-to-mainstream gap to form. Every year, more people will find that "Curious AI Director" isn't an abstraction. It's a real role they can step into.

The most valuable skill is no longer "knowing Python" or "10 years in marketing" or "a PhD in physics." It's the combination of curiosity that generates the right questions, attentiveness that catches the errors, and first-principles thinking that leads into territory nobody has explored yet.

The Doors Are Open. The Question Is: Who Will Walk Through?

Vibe coding, vibe science, vibe marketing. These aren't three separate trends. They're one process. AI is radically lowering the barrier to entry in specialized fields, making them accessible to people with the right mindset.

This doesn't mean expertise is dead. Deep specialists are still essential for validation, for edge cases, for telling the difference between a plausible result and a correct one. A specialist with AI will always be faster and more precise than a non-specialist with AI. But the monopoly on entry is broken. It used to take 5 years of study just to begin. Now you can start today and learn as you go, if you have enough curiosity to ask, enough attentiveness to notice, and enough courage to think from first principles.

Platforms and institutions will have to adapt. Social networks need to stop fighting automation and start evaluating content value. Universities need to stop selling knowledge (which is now free) and start teaching thinking skills. Companies need to stop looking for "10 years of experience" and start looking for people who know how to ask questions and notice details.

The doors are open. Not for everyone. AI requires effort, discipline, and attentiveness. But for far more people than ever before.

The question isn't whether you can walk through. It's whether you will.