There is a founder I keep meeting in different bodies. She has six Claude conversations open. She has a coding agent shipping features, a content agent drafting posts, a finance agent that processes invoices when she remembers to paste them in, a research agent that reads competitor sites at night. She is doing the work of a team of fifteen. She is also, very quietly, exhausted, because she is the only place all of that work meets.
I want to give that founder a name. Not a job title. An identity.
She is a vibepreneur. The word is a small joke and a real claim. The joke is that it sits next to vibe coding, the term Andrej Karpathy threw onto the internet in February 2025 and that Collins Dictionary picked as its Word of the Year by November. The real claim is that the thing she is doing is genuinely new, and it deserves a name that makes the difference visible. A vibe coder writes software without writing software. A vibepreneur runs an organisation without running every part of it by themselves. The difference is not the tools. It is whether the work has a structure that can hold itself together while she sleeps.
Is Your Organisation Ready for AI Agents?
Eight dimensions of role clarity, distributed authority, living governance, and accessible data. Find out in a few minutes.
A vibepreneur is a solo founder, or a very small team, running an entire company where AI agents fill the operational roles a team of people used to fill. The human keeps purpose, judgement, direction, and fills some specific roles. The agents do the work, coordinated through an explicit structure that gives them shared context, clear authority, and a tracked record of every decision.
Let me give that definition a precise edge, because vague identity terms collapse into nothing within a season.
The vibepreneur is not defined by headcount. It is defined by operating model. A solopreneur is a fact about how many humans run the company. A vibepreneur is a fact about how the company itself is organised. A freelance designer with no agents is a solopreneur, not a vibepreneur. A founder running a SaaS company with eight agent role-fillers making their own decisions and one human is a vibepreneur, even though the human headcount is one. The categories overlap. They are not the same.
The vibe in vibepreneur is the same one Karpathy reached for in his original tweet: "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." The lift of vibe coding is that the human stops typing code and starts directing toward outcomes. The lift of vibepreneurship is the same shape, one layer higher. The founder stops doing every operational task and starts vibing with the team that does them. The exponential is not the AI. It is what becomes possible when one human's judgement gets matched to ten roles' worth of execution.
A vibe coder produces software. A vibepreneur produces a company. Coding is one accountability of one role inside that company. The vibepreneur question is the larger one, and it is harder, because software has a compiler that tells you when you are wrong within seconds. A company does not. When the marketing agent and the sales agent contradict each other on a Tuesday afternoon, no compiler catches it. The structure has to.
The numbers have done the arguing for me, and they have been doing it loudly enough that the only thing left is to name the person at the centre of them.
According to Carta's Solo Founders Report 2025, the share of new startups with a solo founder rose from 23.7% in 2019 to 36.3% in the first half of 2025. One in three new ventures launched alone. Dario Amodei, Anthropic's CEO, has publicly put 70 to 80% odds on the emergence of a one-person billion-dollar company by the end of 2026. Sam Altman has had a running group-chat bet on the same question since 2024. These are not fringe predictions. They are the leading edge of a structural shift in how companies get built.
The example everyone is talking about is Medvi. Matthew Gallagher started the company in September 2024 with $20,000 and his brother Elliot. By the end of 2025, the company had verified $401 million in revenue and about 250,000 customers, all run by two people and a stack of AI tools. Erin Griffith profiled it for The New York Times in April 2026 under the headline "How A.I. Helped One Man (and His Brother) Build a $1.8 Billion Company." The story has since been complicated by FDA scrutiny of the wider telehealth category, by questions about the company's marketing practices, and by the obvious distinction between verified 2025 revenue and a 2026 run-rate projection. This makes Medvi as good an illustration of where the vibepreneur model goes wrong as where it goes right. The pattern is real. So is the need for governance around it.
Collins Dictionary picked vibe coding as its Word of the Year in November 2025, citing the term's leap from a niche programmer joke to mainstream professional vocabulary in nine months. That kind of vocabulary shift does not happen because of a tweet. It happens because the underlying practice changed faster than the language could keep up. The same shift is now happening above the code layer. There are founders running operations, marketing, customer success, finance, and research through agent teams, and there is no word for them that captures both the felt experience and the structural reality of what they are doing.
Solopreneur undersells it, because their team is real. AI-powered founder sounds like a press release. One-person company is descriptive but emotionally flat. Vibepreneur lands because it carries the felt quality, the lineage to a recognised cultural moment, and the implication that this is a new kind of person doing a new kind of work.
I would not be honest with you if I just sold the upside, because the upside is currently sitting inside a failure mode that is statistically significant and getting worse.
BCG's Build for the Future 2025 report found that only 5% of companies are "future-built" in their AI adoption, meaning they achieve AI value at scale. Thirty-five percent are scalers making real but uneven progress. Sixty percent are laggards still stuck in pilot mode. Amanda Luther, an MD at BCG, summed up the pattern: agents are not plug-and-play, and companies need to redesign how work gets done, including the impact of agents on existing roles and processes. McKinsey's State of AI 2025 sharpens the same point from another angle: 88% of organisations now use AI in at least one business function, but fewer than 10% have scaled AI agents within any single function.
The most public cautionary tale is Klarna. The company launched its customer-service AI in February 2024 and announced that it was handling the workload of 700 human agents. By May 2025, CEO Sebastian Siemiatkowski reversed course in a Bloomberg interview: "Cost, unfortunately, seems to have been a too predominant evaluation factor when organising this. What you end up having is lower quality." Klarna started rehiring humans. The lesson is not that AI agents do not work. It is that AI agents without governance produce work that looks good on a dashboard and falls apart in the customer's inbox.
At the scale of one founder, the same failure has a smaller blast radius and the same root cause. I call it agent sprawl at the individual level, and it follows a predictable shape.
Agents duplicate or contradict each other. The sales agent quotes one price; the support agent quotes another. The customer sees both. Permission creeps quietly. An agent that started with read access to one system ends up writing to three, because the founder said yes to one small expansion after another, and nobody tracked the trail. Context lives in chat histories the founder cannot search. Why did we set the outreach agent's tone this way? Nobody knows. The conversation where that decision was made is buried under hundreds of others. When the founder takes a week off, nothing has context. The agents keep running on instructions they no longer understand, or they stop running and the work stalls.
None of this is a technology problem. It is the entirely predictable outcome of running a company without a structure that holds when the human is not in the room. The vibepreneur who scales is the one who fixes this before it breaks. The one who does not, has a side project that happens to use AI.
The fix is not new, and it is not glamorous. It is the same thing entrepreneurs have understood for a hundred years, made fully explicit because an AI cannot infer it from the smell of the office.
The traditions worth naming directly are Holacracy and Sociocracy. Tom Thomison and Brian Robertson codified Holacracy in 2007. Gerard Endenburg's work on sociocracy at the Endenburg Elektrotechniek factory in the Netherlands predates it by decades, itself building on earlier work by Kees Boeke. Frédéric Laloux's Reinventing Organizations brought the broader pattern to a wider audience in 2014. None of these traditions had AI agents in mind. They were trying to operationalise distributed authority for human teams. What they discovered, by working on it for half a century, is that any organisation where role-fillers exercise real authority needs the same five elements made explicit. Those elements turn out to be exactly what an AI agent needs to operate well in a role.
The vibepreneur who scales is, almost without exception, a founder who has applied some version of these five elements to her own company. Whether she calls it Holacracy, Sociocracy, role-based work, or her own thing, she has done five things.
The first is nested purpose. Each role has a one-sentence purpose. That purpose nests inside the circle's purpose, which nests inside the company's purpose. When an agent faces an ambiguous decision, it has a hierarchy of purpose, not people, to filter through. The content agent does not just ask "did I publish this post?" It asks "does publishing this post serve what we exist for?" This works as an organic form of context engineering.
The second is accountabilities, not tasks. A task is a thing the agent does once. An accountability is what the agent is permanently responsible for. "Maintaining the editorial calendar so nothing stalls between draft and publication" is an accountability. "Write a blog post this week" is a task. An agent with accountabilities owns a function. An agent with tasks waits for the next prompt.
The third is domains. Roles can have exclusive authority over a specific area. The outreach agent owns external communication with prospects. No other role touches that without explicit permission. Domains are what stop the two-agents-contradicting-each-other problem before it starts.
The fourth is living policies. Policies are explicit agreements about, among other things, how roles cross into each other's domains. "The content agent may reference the product roadmap for planning, but may not publish unreleased features." "Refunds above €100 require Circle Lead confirmation." Policies are the working agreements that let a founder close her laptop at night without anxiety. Because they live in governance records, they can be changed in five minutes when one of them turns out to be wrong.
The fifth is shared organisational context. Every role-filler needs access to the same structural map of the company. What every role is. What every role owns. What the current projects are. What the last governance change decided. Without it, every role-filler is reasoning from a private context that drifts further from reality every week. With it, role-fillers coordinate through the structure itself rather than through the founder.
These are not new ideas. What is new is that AI agents have made them newly visible and newly urgent. A human in a vague role can fake the parts that are missing by intuition. An AI in a vague role cannot, and the failure shows up within a day. The structure was always the difference. The AI just made it impossible to ignore.
There is a deeper point here, and I want to surface it because it bears on what the vibepreneur is for.
Brian Robertson wrote in 2015 that the human role in an organisation was never primarily to produce. It was to sense reality on the organisation's behalf, from inside the role they energised. Role-fillers, he argued, are the organisation's sensors. They notice what the rest of the system cannot. That insight was true for human teams. It becomes structurally sharp in a vibepreneur company.
Most of the operational roles are filled by agents. The human is no longer the executor. She is the primary sensor of the organisation. The agents map possibility space at scale, faster and wider than any human could. They cannot feel which possibility actually belongs in the world the customer lives in. That feeling is reserved for the human at the centre, who shares the customer's world and shares the world the purpose is trying to change. The vibepreneur is, in this sense, an extreme case of the human role becoming more recognisable, more valuable, and more clearly essential than it has ever been. Not less.
The work of a vibepreneur is not to do every operational task. It is to feel what the company should be doing, to design the conditions under which the agents can do it, and to evolve the structure as she learns what works. The agents bring computational reach. She brings situated taste. Both are needed. Neither alone is enough.
Let me walk you through a single Wednesday, because the difference between a vibepreneur with structure and a vibepreneur without one is not visible in the tools they use. It is only visible in the day they have.
She is asleep at dawn. The Outreach Specialist role, energised by an AI agent, finishes preparing three personalised messages to prospects identified earlier in the week. It does not send them. A policy on the role says no external communication leaves the company without Circle Lead consent. The messages sit in the outreach pipeline with a status comment: "Three prospects ready for review, all within accountability scope, no policy boundary touched." The agent stops.
She opens the laptop with a coffee. Three things are waiting. The outreach drafts, ready for one read-through and a tap to send. A Role Update comment from the Customer Success agent saying it handled four support tickets within tone-of-voice policy and flagged one ticket for human review because it touched a refund above the €100 threshold. A tension from the Content Strategist role: "I cannot fulfil my accountability to maintain the editorial calendar because I do not have access to the product release dates." The tension is captured for the next tactical meeting. It took her a few minutes to take all of it in.
The tactical meeting runs at midday. It is her and a structured agenda generated from the system. Project updates first: pipeline status, support load, cash position, deploy state. Each role posts its current update. Then tensions. The editorial-calendar tension is processed in two minutes. She grants the Content Strategist read access to the product release calendar with a policy that draft roadmap items may inform planning but may not appear in published content. The change is timestamped and visible to every role from the next activation onward.
Later in the afternoon, the Finance Operations agent posts a comment on an invoice it processed. Status: Done. Amount: €180. Policy threshold: €150. The agent flagged the invoice for confirmation rather than processing it autonomously. She taps approve. The decision is recorded with the policy that triggered the flag.
She closes the laptop at the end of the day. The agents keep working. The outreach agent prepares tomorrow's pipeline but will not send anything without consent. The content agent drafts the next two pieces of the editorial calendar but will not publish them. The research agent reads three competitor sites and posts findings as comments on the relevant projects. Nothing leaks. Nothing drifts. Anything they cannot decide alone waits for her, named clearly, with the policy or domain boundary that triggered the wait spelled out.
That day exists because every role has a purpose, a set of accountabilities, a domain it owns, the policies that govern how it interacts with other roles, and a connection to the same living organisational context. The founder did not run the work. She designed the conditions in which the work runs itself.
The vibepreneur without structure has a different Wednesday. She wakes up to seventeen unread notifications across four tools, and has to ask for some others. She spends a chunk of the morning piecing together what happened overnight. She finds an email her outreach agent sent that contradicts something her support agent told the same prospect last week. She apologises. She rewrites a prompt. She closes the laptop long after dark.
Same agents. Same models. Different organisation.
The thing that makes this practically different from six chat tabs is not a better prompt. It is that the agents are reading from the same book.
Model Context Protocol, the open standard originally introduced by Anthropic in late 2024 and now governed by the Linux Foundation, is the bridge that lets AI assistants connect directly to external systems. When a company's governance lives in a platform that exposes itself through MCP, the agents stop reasoning from static instructions and start reasoning from the living organisation.
In practice this means an agent activating in the Content Strategist role can read its own purpose, its current accountabilities, the policies that constrain its work, the editorial calendar that other roles are looking at, the last governance change that affected its domain, and the recent tactical meeting outcomes. None of that is in the prompt. All of it is in the workspace. The prompt becomes short, durable, and barely changes from one activation to the next, because the specific context lives on the role, where it belongs, and where the vibepreneur can evolve it once and have every future activation pick up the change.
For a vibepreneur, this is the difference between maintainable and not. A company with five agent role-fillers and no shared organisational context is five separate maintenance jobs. A company with five agent role-fillers and a shared MCP-accessible workspace with a clear governance structure is one organisation. The structure carries the load that used to live in the founder's head.
This is the part I find practically beautiful, and I want to be honest that I am biased. At Nestr we have spent the better part of a decade building the platform where this kind of governance lives, and the MCP layer that lets agents read it. It is not the only way to do this. It is the way I know best, and the way I would build it if I were starting a vibepreneur company today.
I think the most useful thing this term can do is not to describe a workflow but to describe a person.
When Sean Ellis published "Find a Growth Hacker for Your Startup" on his blog in July 2010, he did something quietly important. He named a category. His founding line read: "A growth hacker is a person whose true north is growth. Everything they do is scrutinized by its potential impact on scalable growth." The post itself was short. What it did was give a community a name to claim, and the community claimed it. There was a book, a conference, a movement. The work was already happening. The name made it visible.
Karpathy did the same thing on a Sunday in February 2025. He fired off what he later described as "a shower of thoughts throwaway tweet" and the term vibe coding entered the language. The practice did not need permission. It needed vocabulary.
Courtland Allen did it again when he founded Indie Hackers in 2016. The category had existed in different forms for a decade, but indie hackers stuck because it gave a particular kind of founder an identity they wanted to claim in public.
The lesson of all three is the same. Identity terms matter not because labels are important but because language is the thing that lets a community find itself. People who recognised themselves in the term gathered around it. Tools anchored to it. Playbooks emerged. The category gained mass.
I would like the same to happen here, in the open. If you are a vibepreneur, the most useful thing you can do is call yourself one. Out loud. In public. Not because of Nestr, but because the community needs the surface. The number of people quietly running companies on AI agents is much larger than the conversation about them suggests, and they are mostly trying to figure this out alone.
The companies vibepreneurs are building are not smaller versions of the companies that came before. They are a different shape. One human at the centre, holding purpose and direction. A circle, or several, of agent role-fillers around her, each energising a role with a clear purpose, accountabilities, domain, and policies. A living governance layer that every role-filler can read and contribute to. A regular rhythm for the operational work and a slower rhythm for the structural changes. The human is rare. Her judgement is dense. Her taste is what determines whether the company makes anything that matters in the world.
That shape was not possible until AI agents could fill operational roles reliably. It is possible now. The vibepreneurs who get it right will build companies that are more alive than what came before. Smaller in headcount, larger in reach, more honest about what the human at the centre is for. Not less essential. More.
The work ahead is to make this shape visible enough that the founders already doing it can find each other, learn from each other, and hold each other to a higher standard. The article you are reading is one move toward that. The word is the other.
Let's vibe for purpose.
Build the working layer for human and AI role-fillers
Nestr is where the structure of a vibepreneur company lives: roles, policies, tensions, and the MCP layer your agents read from.
Andrej Karpathy, original "vibe coding" tweet, 2 February 2025
Andrej Karpathy, one-year retrospective on the tweet, February 2026
Collins Dictionary Word of the Year 2025, CNN coverage, 6 November 2025
Carta, Solo Founders Report 2025
Dario Amodei's 70–80% one-person billion-dollar company prediction, via Inc
BCG, "The Widening AI Value Gap: Build for the Future 2025," 30 September 2025
McKinsey, "The State of AI in 2025: Agents, Innovation, and Transformation"
Sean Ellis, "Find a Growth Hacker for Your Startup," 26 July 2010
Indie Hackers (Courtland Allen, founded 2016)
Frédéric Laloux, Reinventing Organizations
Anthropic, Model Context Protocol announcement
Nestr, "AI Agent Governance: The Organisational Readiness Gap"
.png){kind=logo}
