How A.I. Helped One Man (and His Brother) Build a $1.8 Billion Company

The New York Times ^ | 3rd April 2026 | Erin Griffith

[Cronos]

Matthew Gallagher took just two months, $20,000 and more than a dozen artificial intelligence tools to get his start-up off the ground.

From his house in Los Angeles, Mr. Gallagher, 41, used A.I. to write the code for the software that powers his company, produce the website copy, generate the images and videos for ads and handle customer service. He created A.I. systems to analyze his business’s performance. And he outsourced the other stuff he couldn’t do himself.

His start-up, Medvi, a telehealth provider of GLP-1 weight-loss drugs, got 300 customers in its first month. In its second month, it gained 1,000 more. In 2025, Medvi’s first full year in business, the company generated $401 million in sales.

Mr. Gallagher then hired his only employee, his younger brother, Elliot. This year, they are on track to do $1.8 billion in sales.

A $1.8 billion company with just two employees? In the age of A.I., it’s increasingly possible.

Mr. Gallagher works on Medvi from his house basically anytime he’s not showering, sleeping or spending time with his two children, he said during a two-hour conversation. He even made an A.I. clone of his voice to help manage his personal life, using it to call and schedule appointments so he would have more time to work.

...He used many A.I. tools to build Medvi’s website, including ChatGPT, Claude and Grok. He created custom tools, including A.I. agents, or bots that perform tasks on their own, to get his software systems to communicate with one another. He tested A.I. voice tools from ElevenLabs and others for communicating with customers. And he used the image and video generators Midjourney and Runway to create media for his website and ads.

---

TOPICS: Computers/Internet
KEYWORDS: ai; industry

[Cronos]

[MarlonRando]

good for him. This is like the guys who learned BASIc in the heyday and made software

[Theophilus]

He works hard for the money. So hard for it honey.

[Resolute Conservative]

So, the “learn to code” meme is false. Sounds like a good idea to put 4.4 million (from a simple search) or more people out of a job and render a couple valid hi-tech degrees worthless. Glad I am not going to be around to watch the bread lines and our submission to machines running our lives.

Note: been a real software engineer for 30 years and AI makes up less than 10% of my work effort (I use it for research and spot questions only, it never writes code for me) and compared to the new breed of developers my work product is delivered faster and accurate while they sit around trying to figure out how to give their job away.

[Cronos]

I just wonder about the productionizing of the code.

Claude Code or CodeMie puts some good small code, but then if we are talking of larger projects or programs, then I don’t see this working.

and the maintenance of AI generated code is going to be nutz

[Brian Griffin]

We provide medical care online—simple, direct, and led by licensed providers. No waiting rooms. No unnecessary steps. Just care that works.

Healthcare,

redefined for real life.

We provide medical care online—simple, direct, and led by licensed providers. No waiting rooms. No unnecessary steps. Just care that works.

Weight Loss

Everything you need—included:
Prescription to fast, effective GLP-1
1:1 physician guidance
24/7 support
Weight loss guarantee
Fast & discreet shipping

MEDVi Meals is the perfect partner for your GLP-1 journey—delivering chef-prepared, medically aligned nutrition designed to help you lose fat while preserving muscle.

Peptides & Longevity
Men’s Health
Women’s Health

https://home.medvi.org/

[Brian Griffin]

ADVANCED Peptide SUPPORT

Recovery support (BPC-157, TB-500)
Performance (CJC-1295, Ipamorelin)
Recovery cycles (Sermorelin, MK-677)
Strength (Tesamorelin)

[delapaz]

believe it or not this is already working.

and you end up not caring about the maintainability, you just spend more tokens fixing whatever is wrong. You do have to start with good design, good specs, but its working.

[Brian Griffin]

https://www.peptidesinstitute.org/peptides/bpc-157

[Brian Griffin]

“TB-500 is a synthetic version of the naturally occurring peptide LKKTETQ.”

https://muscleandbrawn.com/peptides/tb-500/

[Brian Griffin]

“you just spend more tokens”

*****

“Tokens are small units of data — words, sub-words, characters, or pixel patches — that AI models process during training and inference.”

“Context matters too. The word “lie” could mean resting flat or speaking falsely. During training, a model learns these distinctions and assigns different token IDs to the same word depending on surrounding context.

“Tokenization is not limited to text. Visual models break images and video into patches of pixels or voxels, each mapped to a discrete token. Audio models sometimes convert sound clips into spectrograms — visual representations of frequency over time — which are then tokenized like images. Other audio systems skip that step and extract semantic tokens that capture the meaning of spoken language rather than raw acoustic data.

“Efficient tokenizers reduce the total number of tokens a model must process, which directly lowers computing costs.”

“Training an AI model begins with tokenizing the full training dataset. For large language models, that dataset can contain billions or trillions of tokens. A well-established pretraining scaling law holds that larger token counts during training lead to higher-quality models.

“The core training loop works through prediction. The model sees a sequence of tokens and tries to guess the next one. When it guesses wrong, internal parameters update to improve accuracy on the next attempt. This cycle repeats across the entire dataset until the model reaches a target accuracy threshold — a state called model convergence.”

More at:
https://www.sentisight.ai/tokens-explained-new-currency-of-generative-ai/

[Brian Griffin]

“MK-677 is a growth hormone secretagogue, meaning it increases your body’s production of human growth hormone (HGH) and insulin-growth factor 1 (IGF-1). It first showed up as an investigative drug when looking at child growth hormone disorders and frailty in older adults. This drug is still in investigate status, meaning it has not been approved for human consumption.

“Despite not being approved for human consumption, this compound is still being found online and popping up in dietary supplements. Its use comes with serious risks, especially when taken without medical supervision. The most common side effects include insulin resistance, increased appetite, weight gain, and swelling due to water retention. Long-term use may increase the risk of developing type 2 diabetes and contribute to the growth of cancerous tumors due to elevated IGF-1 levels. Users have also reported side effects such as joint pain, numbness, and fatigue. Because MK-677 is not approved by the FDA for over-the-counter use and has not been thoroughly studied in healthy individuals, its long term safety profile remains uncertain. One of the main reasons it has not moved beyond investigatory status is due to observed increase risk of cardiovascular damage during one of the research studies.”

https://www.getsmartaboutdrugs.gov/news-statistics/2025/07/08/beyond-hype-potential-health-risks-mk-677

[Bon of Babble]

My nephew was diagnosed with stage 4 colon cancer and given a prognosis of “terminal” with months to live, if that.

He went to New York City and was able to get a brand new peptide treatment - he had nothing to lose and everything to gain if it worked.

The doctors told him it would slow or halt the cancer but not cure it. It’s been three years so far and he’s returned to a relatively normal life. He did three rounds and was told that’s all his body could take.

[ckilmer]

I’m on a very, very low-to-bankrupt budget.

At 73, I live inside a hornet’s nest of tinnitus.

My vision in one eye is fuzzy. I wear glasses when I look at the computer.

I need to get more exercise because my gait is getting wobbly.

The county is covered with server farms.

I’m hoping the county will hire me. They have a program for guys like me.

But I’ve never written a stitch of code in my life.

About three weeks ago, I built an email software that self-improves. using Claude and a Claude code writer whose name I’ve forgotten. I worked the two together like R2D2 and 3PO (Star Trek reference). (Sometimes I would double-check claud’s work with Grok.) I put the email software up for sale online. But to make any sales, I need a whole ecosystem of funnels to bring in customers. I did not see a path to money anytime soon on this project.

So I saw some articles two weeks ago about what the guy in this news story was doing and set out to see if I could do the same thing.

I researched. I’ve talked a lot to Grok. Probably, Claude would be better. But I started with grok. And I’m finding that Grok will maintain the context window. over days and now over a week. So we are learning together about how to do this job.

I don’t have much money. The first thing is to find a way to do everything for free.

Oracle provides about 24 GB of free server space. Github provides all the free tools

I do pay grok $20@mo. So Grok helped me set that up. Grok turned the conversation into code, which I plugged in. It worked. Often, there was a lot of back-and-forth, trial-and-error. Often, my questions caused open loops. I’ve learned how to get grok out of them. Grok would sometimes try the same thing over and over again. I’ve learned how to tell Grok to spot failure patterns and to describe them so well that Grok can use them to interrogate forums and the software literature to find a solution. That solves persistent failures pretty fast.

So last week I started building a company that makes stuff and sells it using agents inside the Oracle Cloud. I have about 143 agentic tools right now and a comprehensive self-improvement architecture, both at the corporate and business levels. Grok keeps converting our conversation into code, then checks it for errors. Then we check to see if it runs.

I don’t know if I’ll be successful.

But this work gets me out of bed in the morning. Before, I struggled to maintain my full attention for more than a couple of hours. I work on this all day and night.

In short. It’s fun. And addictive. (I may go bankrupt. But I’ll have fun all the way down the slide until the drop.)

[atomic_dog]

Note to self: Remember to download the “Claw Code” from Github while it’s still there.

[dfwgator]

You mean the Claude Code source, that got leaked?

[Resolute Conservative]

Good luck to you.

[atomic_dog]

Yep. GitHub

I understand this was rewritten in Python so Anthropic cannot get a DMCA takedown order on it. I also heard they are rewriting it in Rust. What would be funny is if they are using an AI model to do the rewrite.

[dfwgator]

Seems to me Python would be slow as a dog, Rust would make much more sense.

[The Duke]

"Note: been a real software engineer for 30 years and AI makes up less than 10% of my work effort (I use it for research and spot questions only, it never writes code for me) and compared to the new breed of developers my work product is delivered faster and accurate while they sit around trying to figure out how to give their job away."

I've been a software and aerospace engineer for 45 years, and I use AI to create software tools that I use all the time. I also use it to speed up DevOps operations by a large factor. Many of the things that AI has empowered me to do I simply could not justify taking time to do in years past ... and don't even get me started on clearing out tech debt with AI!