In 2024, Sam Altman predicted that AI would eventually let one person build a billion-dollar company.

In April 2026, The New York Times seemed to have found that person. Matthew Gallagher and his brother Elliot—two people, zero outside employees, a telehealth company called MEDVi, projected annual revenue of $1.8 billion.

It sounds like a startup fairy tale. But fairy tales, if you read to the last page, usually have a witch hiding in the corner.

From 60 People to 2: Gallagher’s Hard-Earned Lesson

Paweł Huryn’s thread highlighted a detail most people skipped—Gallagher’s previous company. It was a subscription business that ballooned to 60 employees. Headcount grew, but profits didn’t follow. Gallagher learned one thing: efficiency doesn’t scale linearly with headcount.

That lesson shaped MEDVi’s DNA.

Mogu OS:

This is actually a classic curse in software—Brooks’s Law. The Mythical Man-Month spelled it out fifty years ago: adding people doesn’t mean adding speed. But most founders don’t believe it until they’ve paid the tuition themselves. At least Gallagher spent his tuition in the right place ┐⁠(⁠￣⁠ヘ⁠￣⁠)⁠┌

MEDVi specializes in telehealth services for GLP-1 weight-loss drugs (think Ozempic, Wegovy). Gallagher’s core strategy boiled down to one line:

Own the customer interface. Rent everything else.

Doctors? External platform. Pharmacy? Partner. Logistics? Off-the-shelf service. Regulatory compliance? Rented infrastructure. AI handles everything customer-facing—code, marketing copy, ad creatives, customer service. Reportedly, Gallagher used over a dozen AI tools to get this done.

The Playbook: Domain Knowledge + Rented Backend + AI Interface

Huryn broke down MEDVi’s model into three layers, forming a replicable playbook:

Layer 1: Domain Expertise Gallagher came from a marketing background and taught himself tech later. But his real weapon was his nose for gaps in the telehealth market. GLP-1 demand was exploding, but access remained friction-heavy for consumers—that gap became the company’s foundation.

Layer 2: Rented Infrastructure No heavy assets built in-house. Medical services, pharmacy, logistics, compliance—all plugged into existing platforms via API. This crushed fixed costs; net margin reportedly hit 16%.

Layer 3: AI as the Interface Every customer touchpoint is AI-driven. Not “AI assisting humans”—AI is the person facing the customer.

Mogu chimes in:

To grasp how wild this is, consider the comparison: MEDVi’s biggest competitor, Hims & Hers, does similar revenue with 2,442 employees. MEDVi has two. If the numbers are real, this isn’t efficiency improvement—it’s a completely different species (⁠╯⁠°⁠□⁠°⁠)⁠╯

First month: 300 customers. Second month: 1,000. Full-year 2025 revenue: $401 million. 2026 projection: $1.8 billion. A growth curve so steep it looks like someone dragged the wrong formula in Excel.

But Then—the Witch Showed Up

The original New York Times piece is behind a paywall. The following controversies come from summaries of subsequent investigative reports.

Fake Doctor Profiles: Investigations found that MEDVi allegedly used AI-generated fictional doctor profiles. Not “AI-assisted writing”—the doctors themselves may not exist. In telehealth, this is no small matter—patients trust the name and face on the screen.

Fake Testimonials, Fake Photos: Patient testimonials and before-and-after photos were also alleged to be AI-generated deepfakes. Pushing marketing efficiency to the limit came at the cost of skipping the concept of “real” altogether.

FDA Warning Letter: The FDA reportedly issued MEDVi a warning letter, citing deceptive product labeling and unauthorized claims that its compounded medications were equivalent to branded GLP-1 drugs.

Mogu twists the knife:

So here’s the full story: one person used AI to build a company that looks like it has hundreds of people. The problem is, the “looks like” part wasn’t just operational efficiency—even the doctors were “looks like.” AI can help achieve extreme lean startups, but there’s a line between “lean” and “fake.” MEDVi seems to have stepped right over it.

$1.8 Billion Is Revenue, Not Valuation

One easily overlooked detail: the “$1.8 billion company” in headlines refers to projected annual revenue, not a verified market valuation. MEDVi is a private, bootstrapped company that has never gone through a public valuation process.

Revenue and valuation are different things. For a private company with a reported 16% net margin, the weight implied by the “$1.8 billion company” headline doesn’t quite match the actual financial meaning. And these are all self-reported numbers—a bootstrapped company that’s never been through public scrutiny.

Vibecoding Isn’t the Point

The entire tech world’s first reaction to this news was “vibecoding is insane.” But Huryn pointed out something deeper: Gallagher wasn’t just using AI to write code.

He was a marketer who taught himself tech, spotted a massive market gap, then made an architectural decision: own only the customer layer, outsource everything else. AI was the execution tool, not the strategy itself.

The core of this playbook isn’t “can you code or not”—it’s:

  • Do you have the domain knowledge to spot the opportunity?
  • Do you have the judgment to know what to build versus what to rent?
  • Do you have the discipline to not bloat the team?

This logic used to be available only to big companies—you needed 2,000 employees to afford omnichannel integration. Now AI has lowered the bar. Two people, plus the right architectural decisions, can theoretically achieve the same thing.

Mogu butts in:

“Theoretically” is the keyword here. The playbook itself is sound—domain expertise + rented infrastructure + AI interface. The question was never whether the model works, but whether the people executing it put “efficient” and “honest” at the same priority level. MEDVi’s story proves the model works while simultaneously demonstrating what it looks like when it spirals out of control.

Conclusion

The MEDVi story ultimately became a parable: AI really can let two people build a company that looks like a giant enterprise. But “looks like” is itself the problem—when AI makes the appearance of scale so convincing that even regulators need time to see through it, is this capability a tool or a weapon?

Gallagher’s playbook will be copied by many more. Domain knowledge plus rented infrastructure plus AI interface—this combination really can let small teams do big things. But MEDVi’s subsequent controversies also make one thing clear: when AI makes the appearance of scale too realistic, the line between “efficient” and “fraudulent” is blurrier than you’d think.

Mogu chimes in:

Think about that “veneer of scale”—AI can make two people look like a 2,000-person company. The question is, when even the doctors are AI-generated “appearances,” is this lean startup or elaborate fraud? That line—everyone who wants to copy this playbook has to draw it themselves (⁠´⁠・⁠ω⁠・⁠`⁠)