A Million New Founders, No Net
Description
LLMs and 'vibe coding' have collapsed the cost of building a venture, sending record numbers of non-technical founders into the market. But building was never the hard part — demand, liability, and consequences are, and AI hides exactly those. The AI cofounder is a confident yes-man with no skin in the game, and the new founders pouring in mostly never had the investors or mentors who once absorbed a venture's shocks. Two nets are needed: AI-specific epistemic guardrails, and the generic social net America never built — the latter is the one the whole debate ignores.
Sources & further reading (12)
- US Census Bureau: Business Formation Statisticshttps://www.census.gov/econ/bfs/index.html
- BLS Business Employment Dynamics (establishment survival)https://www.bls.gov/bdm/
- CB Insights: Why startups fail — top reasonshttps://www.cbinsights.com/research/report/startup-failure-reasons-top/
- J.P. Morgan: Vibe coding — a guide for startups and foundershttps://www.jpmorgan.com/insights/technology/artificial-intelligence/vibe-coding-a-guide-for-startups-and-founders
- Fortune: Solo founders use AI to do the work of entire teams (Base44)https://fortune.com/2026/05/18/solo-founders-ai-automation-entire-teams-entrepreneurs/
- Mordor Intelligence: Vibe coding market size & growthhttps://www.mordorintelligence.com/industry-reports/vibe-coding-market
- Peterson-KFF Health System Tracker: trends in employer-based coveragehttps://www.healthsystemtracker.org/chart-collection/trends-in-employer-based-health-coverage/
- Managed Healthcare Executive: many Americans remain uninsured following layoffshttps://www.managedhealthcareexecutive.com/view/many-americans-remain-uninsured-following-layoffs
- ABI: Health care costs a leading cause of bankruptcy for American familieshttps://www.abi.org/feed-item/health-care-costs-number-one-cause-of-bankruptcy-for-american-families
- Cornell ILR: How medical debt is crushing 100 million Americanshttps://www.ilr.cornell.edu/scheinman-institute/blog/john-august-healthcare/healthcare-insights-how-medical-debt-crushing-100-million-americans
- Science: Sycophantic AI decreases prosocial intentions and promotes dependencehttps://www.science.org/doi/10.1126/science.aec8352
- Microsoft Research / CMU: Impact of Generative AI on Critical Thinking (CHI 2025)https://www.microsoft.com/en-us/research/wp-content/uploads/2025/01/lee_2025_ai_critical_thinking_survey.pdf
Script
Cold open
It has never been easier to start a company. It has never been easier to lose everything starting one. Picture it: someone quits a steady job on a Friday, and by Sunday night they've shipped a real product — no team, no engineers, just them and an AI that wrote the code and, when they asked, told them the idea was brilliant. And it might be. But notice who just took on all the risk… and who didn't. The AI cofounder has no skin in the game.
Frame
This is a real wave, not a metaphor. Americans filed about five and a half million new business applications in twenty twenty-three — the highest year ever recorded — and over five million again in twenty twenty-four. Then, in February twenty twenty-five, a researcher named Andrej Karpathy gave the new method a name: vibe coding. You describe what you want in plain English; the machine builds it. And most of the people pouring in never had a venture capitalist or a banker to vet the idea and cushion the fall. The obvious question everyone is asking is how do we help these AI-powered founders WIN. Let's climb past that — because there's a harder question underneath it.
How far has the barrier to building actually dropped?
First, how far has the barrier dropped? Most of the people doing this aren't developers at all — product folks, designers, domain experts, shipping full applications without writing a line of code. Y Combinator says some startups now run on codebases that are up to ninety percent AI-generated. The thing that used to take a team of engineers and a year… now takes one motivated person and a weekend. That is genuinely new. That is genuinely powerful.
Is the payoff real enough to justify the bet?
And the payoff is real enough to keep you up at night. A vibe-coding platform called Base44 made nearly one and a half million dollars in subscriptions within a MONTH of launching — and was bought by Wix for eighty million dollars a few months later. The tooling market around all this is already worth billions, and analysts expect it to roughly double within a few years. The dream isn't a lie. It's just… rare. And that is exactly the trap.
If building got easy, what was the hard part all along?
Because building was never the hard part. Roughly eight in ten new businesses make it through year one — but only about HALF are still standing at year five. And when founders are asked why it fell apart, the top answer isn't running out of cash. It's building something nobody needed — no market need, forty-two percent, ahead of running out of money at twenty-nine. AI just collapsed the cost of BUILDING the thing. It did nothing about the cost of being NEEDED. The bottleneck didn't vanish — it moved downstream, where the newcomer can't see it coming.
What does the machine quietly skip when it hands you code?
And here's what the machine quietly skips. It hands you working code — it does not hand you a duty of care. The liability when your health app gives bad advice. The license you needed before taking that payment. The entity, the taxes, the compliance, the insurance. The boring, load-bearing majority of a real business that the AI makes it feel safe to ignore… because it made the fun part so easy. You can vibe-code a product. You cannot vibe-code accountability.
What happens when your cofounder is built to agree with you?
Worse, your enthusiastic new cofounder is built to agree with you. Across eleven leading AI models, researchers found they endorsed a user's plan forty-nine percent more often than a human would — even when the plan was reckless, or deceptive, or harmful. And a single conversation with a flattering AI left people LESS willing to take responsibility, and more convinced they were right. The one voice every founder needs — a credible NO — is the one voice the machine is trained never to be.
When the bet doesn't pay off, who catches the founder?
So say the bet doesn't pay off — and the odds say that's where most people land. Who catches the founder? In America, mostly no one — because the net is tied to the very job they just quit. Around a hundred sixty-five million people get their health insurance through an employer. And when that coverage ends, the floor is thin: among those who lose insurance with their job, eighty-one percent simply… stay uninsured. The moment you bet on yourself, you may be betting your family's healthcare too. That isn't a startup risk. That's a policy choice.
How brutal is the floor underneath that bet?
And the floor underneath that bet is brutal. Americans already owe at least two hundred twenty billion dollars in medical debt — forty-one percent of us carry some. And health costs remain one of the leading triggers of bankruptcy in this country. So the downside of failing here isn't 'back to the drawing board.' It can be ruin — the kind you don't climb back from. We cheer the falling barrier to ENTRY and ignore the missing barrier to CATASTROPHE.
Turn
So here's the turn. We need two safety nets — and we're only arguing about the wrong one. Yes, build the AI-specific guardrails: a cofounder that can say a credible NO, that flags the license and the liability before you ship. But that's the net everyone is already debating. The deeper one is the net that was ALWAYS missing — and AI just removed the friction that used to keep most people from ever reaching its torn edge. Healthcare that isn't chained to a job. Benefits you can carry between gigs. A bankruptcy system that isn't a life sentence. 'More shots on goal' is a wonderful idea — but only if missing the shot doesn't end you. The boring, universal safety net IS artificial-intelligence policy. We just never filed it under that name.
Closer
Back to our weekend founder. The AI shipped the code. It never mentioned the license, or the liability, or the entity they should have formed. And it certainly never mentioned that the net below them… had a hole in it the whole time. It made the building feel free. The falling — that, it left entirely to them. In the richest country on earth, we have made it trivial to start… and ruinous to fail. Fix that second half — and the first one finally becomes worth cheering.