How Claude is Creating a New Generation of Millionaires
A single founder makes the case that Claude Code has erased the cost of building software, using a three-person team's state government contract as proof.
Posted
3 days ago
Duration
Format
Talking Head
hype
Views
77.4K
2.2K likes
Big Idea
The argument in one line.
Claude Code has removed the technical and financial barrier to building software, so the deciding skill now is describing a problem clearly and verifying the AI's work, not writing code.
Who This Is For
Read if. Skip if.
READ IF YOU ARE…
You have a repetitive business task you've been meaning to automate but assumed you'd need to hire a developer for.
You're a non-technical founder or solo operator curious what Claude Code can actually do beyond chatbot Q&A.
You want a concrete first step into AI-assisted building rather than more theory about AI's potential.
SKIP IF…
You already build daily with Claude Code and know the agentic/parallel/memory basics — this is an on-ramp video, not an advanced workflow.
You're looking for technical implementation detail on Vulcan's actual product — it's used only as a proof-of-concept anecdote, not explained.
TL;DR
The full version, fast.
The video's thesis is that Claude Code has collapsed the cost of building software low enough that people with no coding background are launching real, revenue-generating companies. The proof case is Vulcan Technologies, a three-founder team (two of whom couldn't code) that built a government regulatory-tech platform and won a Virginia state contract at roughly 10% of what incumbent consulting firms quoted, later leading the governor to mandate agentic AI use across state agencies. The core mechanism is four traits: Claude does the work directly from plain-English instructions, acts agentically instead of just answering questions, can run several tasks in parallel, and retains memory of a business's context over time. The actionable conclusion is a four-step starter path: get a paid Claude plan, pick one real repetitive task, force Claude to adversarially critique the idea before building (a pattern the host calls 'the roast'), then build in small verified increments rather than trusting the AI's word that something works.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Cold open teasing Vulcan Technologies and promising a framework for building with Claude.
00:33 – 01:49
02 · Why Claude Wins
Anthropic's $65B raise at a ~$965B valuation surpassing OpenAI, and run-rate revenue growth from $1B to $47B in about 18 months, shown via screenshotted press coverage.
01:49 – 02:56
03 · The Vulcan Story
Three founders, two non-coders, built a government regulatory platform by copy-pasting from the Claude chat app, won a Virginia contract at ~10% of competitors' price, and triggered a state mandate for agentic AI.
02:56 – 05:15
04 · The Four Things
The host's framework for what makes Claude Code different: it does the work, is agentic, runs in parallel, and remembers context — illustrated with the YC batch statistic.
05:15 – 09:36
05 · Where To Start
Four-step starter path: paid plan, pick one task, adversarial idea-review ('the roast'), then build and verify incrementally. Closes with a pitch for his Skool community.
Atomic Insights
Lines worth screenshotting.
Anthropic's run-rate revenue grew from about $1 billion at the end of 2024 to roughly $47 billion about eighteen months later.
A $65 billion funding round valued Anthropic at approximately $965 billion, putting its valuation ahead of OpenAI for the first time.
Vulcan Technologies was founded by three people, two of whom could not write code, and one hadn't touched code since a high school JavaScript class.
Vulcan's first prototype was built by copying and pasting code out of the standard Claude chat app, before Claude Code existed.
Vulcan won a Virginia state government contract at roughly 10% of the price incumbent consulting firms quoted.
Virginia's governor signed an executive order requiring state agencies to use the kind of agentic AI regulatory review Vulcan built.
In Y Combinator's most recent startup batch, more than half of the companies build with Claude, versus roughly 90% favoring OpenAI a year earlier.
A first AI-built version of a task typically comes back only 60-70% correct, and that gap is meant to be treated as feedback, not failure.
The adversarial review pattern described (nicknamed 'the roast') runs three sub-agent personalities — a strongest-case advocate, a flaw-hunter, and an objective researcher — to a single go/reshape/kill verdict before any building starts.
Takeaway
Claude Code removed the excuse of not knowing how to code.
WHAT TO LEARN
The barrier to building real software has dropped to describing a problem clearly and verifying the AI's work, and a three-person non-coding team used exactly that to win a state government contract.
The real skill Claude Code rewards is describing a problem in plain language and thinking critically about what comes back, not writing syntax.
Agentic AI that tests, finds bugs, and fixes them on its own closes the gap between having an idea and having a working product far faster than a chatbot that only answers questions.
Before building anything, force the AI to argue against your own idea from multiple angles — a strongest-case view, a flaw-hunting view, and a neutral evidence-based view — because AI models default to agreeing with you.
Never accept an AI's claim that something works; make it run the thing on a real example and show you the output, the same way you'd ask a human to prove finished work.
A first AI-built attempt landing at 60-70% correct is normal and should be treated as feedback to refine, not as a failed result.
Pick one concrete, recurring task to automate first rather than trying to tackle ten problems at once — momentum comes from one finished win.
Being an early adopter of a new capability compounds over time the same way early adopters of basic websites or Facebook ads outran competitors for years.
Glossary
Terms worth knowing.
Agentic AI
An AI system that takes action toward a goal on its own — testing, finding bugs, and fixing them — rather than only answering questions like a standard chatbot.
Run-rate revenue
An annualized revenue projection calculated by taking a company's most recent month of revenue and multiplying it by twelve.
Y Combinator (YC)
A well-known startup accelerator that funds and mentors twice-yearly batches of new companies, several of which have grown into major businesses.
“This is the cheapest employee that you will ever hire.”
punchy reframe of a $20/month subscription→ newsletter pull-quote↗ Tweet quote
07:44
“If you understand language and you understand critical thinking, you can use Claude code really well.”
direct quote from the Vulcan founder, disarms the 'I can't code' objection→ TikTok hook↗ Tweet quote
The Script
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
17px
metaphoranalogystory
00:00Right now, a brand new group of millionaires is being created by AI. There's a company called Vulcan building software that runs inside government agencies. And it looks like the work of a 100 engineers, but it isn't.
00:10Most of their team can't even write a single line code. They just built the whole company describing what they wanted inside of Claude. And this isn't the only million dollar business being run with Claude.
00:19I spent a week with a room full of 7 figure founders doing the same thing. And even I run my own businesses using Claude every single day. So in this video, I'll show you why Claude is breaking the business space and how people with zero coding background are building real companies with Claude code.
00:32So let's get into it. So real quick, in case you're not super deep into the AI space, let me give you guys some context on Claude. Claude is the AI built by a company called Anthropic.
00:40And on the surface, it may just kind of look like another Tratch competitor, but two weeks ago, Anthropic raised $65,000,000,000 in a single funding round. And that round values the company at 965,000,000,000, so just short of a trillion dollars, and they actually just passed OpenAI, the company behind CHATCHBT, for the first time.
00:56Ever. And it's not just investor hype either because the revenue backs it up. Their run rate just hit 47,000,000,000.
01:01At the end of twenty twenty four, that number was $1,000,000,000. So that's 47 times growth in about a year and a half. And run rate basically just means businesses are paying them at a pace of $47,000,000,000 a year.
01:11So the smartest, richest investors on the planet just bet that kind of money on this one company, and companies like Vulcan are a big reason why. Okay. So what actually makes Cloud Code different from every other tool that's out there?
01:22Because there are a lot of tools. So first, think about what it would look like to start a software company over the last decade or so. Let's say you had a great idea for an app or a product.
01:30To actually build it, you needed engineers, and engineers are not cheap. We're talking multiple full time salaries before you even have a first version to show anybody. So unless you had a pile of money sitting around or you wanted to take out a big loan or you could somehow convince investors to fund something that doesn't yet exist, you were just kinda stuck and that idea might just sit in your notes app.
01:49And that's how a lot of good ideas have died, which is what makes Vulcan so crazy because government software is the kind of thing that should have taken a whole floor of engineers. And remember, this is the company from the intro that I mentioned. Three founders, two of them couldn't write code.
02:01One of them, Tanner Jones, hadn't touched code since a JavaScript class that he took in high school. And their first prototype wasn't even built with Cloud Code because Cloud Code didn't exist yet. They built the whole thing in the regular Claude app, basically just copying and pasting code out of the chat until they had something that worked.
02:15And that little copy paste prototype turned into a real product that won a Virginia state contract, reportedly at around 10% of the price the established consulting firms had quoted. Then the governor signed an executive order requiring every single state agency to run the kind of AI regulatory review that Vulcan had built.
02:32The way Vulcan puts it, Virginia required their products by law, and they claim the work is already saving Virginians over a billion dollars a year. And Tanner, the guy who hadn't coded since high school, basically put it like this. If you understand language and you understand critical thinking, you can use Claude code really well.
02:45So basically, the new skill is just being able to describe exactly what you want and thinking critically about what comes back. And that's it. That's the bar now.
02:52Anyone can do it. Okay. So back to the question from earlier, what actually makes Cloud Code different?
02:56I'd boil it down to four things. Number one, it does the work. You describe what you want in plain English, and it builds exactly that.
03:02A full app from one sentence, a website, automating some repetitive task you hate doing every single week. And we're talking, it'll do this in minutes, not months. Number two, it's agentic.
03:10And all that really means is that it takes action on its own towards a goal instead of just answering your questions like a chatbot. So it'll build the thing, and then it'll test the thing, and then it will find the bug, and then it will fix the bug. And with a regular sort of chatbot, you're kind of the one still stuck doing all of the in between.
03:24Number three, it can work in parallel. You can have several different Claude agents running at the same time even while you sleep because they don't sleep. And number four, it remembers things.
03:31And this is the one that I feel most in my own business because everything I do runs through Claude. It knows my business. It knows my team.
03:37It knows my priorities. It knows my past failures. And at this point, honestly, it has a better memory than I do.
03:42Like, it'll literally remind me things that I have to do and things that I might have forgot about. So it's really good. And when you stack all four of those things together, the gap between having an idea and having a real product basically disappears.
03:52So there's this program called Y Combinator, probably the most famous startup accelerator in the world. Twice a year what they do is they take in a batch of new startups, they give them money and mentorship, and they help those startups grow.
04:03And some of those startups have turned into big companies, Airbnb, Stripe, DoorDash, tons more, companies you guys probably know and have used this month. And in YC's newest batch, more than half of the startups are building with Claude. It is the most used AI in the entire batch.
04:17A year earlier, that spot belonged to OpenAI with over, like, 90%. So the founders starting companies right now aren't debating which tool is the best. They've basically picked Claude.
04:26And the same gap is opening up inside normal jobs and small businesses too. The people using Claude for their repetitive work are just getting more done, and they're the ones getting handed the bigger projects and the bigger opportunities. And then the people doing everything manually, they just can't compete with that speed.
04:38It just, you know, makes me think of the cliche saying, which is AI isn't going to take your job. It'll be someone who knows how to use AI. And we've watched this exact movie before.
04:46Think about the early Internet. The people who learned some basic HTML and got their businesses online or understood how to run, you know, Facebook ads. People who did that early grabbed insanely cheap attention while their competitors were still buying newspaper ads.
04:58And that head start compounded for, like, a decade. It's the same window that we're seeing right now. The people winning with Claude aren't smarter than you.
05:04They just started before you. Time is your biggest advantage right now. And being AI native today, like I said, is a real advantage.
05:10But in two years, it's just going to be the baseline. Everyone's going to be AI native. So let's get into how you can actually start right now.
05:17Four steps, and you can do all this today. Step one, get on a paid Claude plan because that's what gets you access to Claude code. The entry level plan is just $20 a month.
05:24Totally fine to start with. You can always upgrade later if you need to. And just think about this like not even as another, you know, subscription like your Netflix or your DoorDash or whatever other subscriptions you have.
05:33This is the cheapest employee that you will ever hire. Step two, pick one real task. Not 10, just pick one.
05:40Something you keep putting off. Something that's repetitive. Something that you might normally outsource to a freelancer and pay someone else to do, organizing receipts, putting together a weekly report, sorting through emails, maybe even doing your taxes, stuff like that.
05:51Or if you've got that notes app of a bunch of ideas that you've wanted to create to, you know, launch a new business, use those ideas. That one task is your first real project. And then step three, you open up Cloud Code, and you describe that task as clearly as you can.
06:04What it is, how you do it today, and what a win looks like. But before you actually let it build anything, make it argue with you. Make it brainstorm with you, and make it play devil's advocate because these AI models kind of lean sycophantic.
06:15They wanna please you. So if you come in excited with your idea, Claude wants to make you more excited. Claude wants to hype you up and start building even if the idea has holes in it.
06:23Planning is so so important. What I do is I make Claude look at everything from every angle first. I literally run a pattern in my own business that I call the roast, where Claude spins up a little council, a bunch of little sub agents, and one of them makes a strong case for my idea, one of them looks at all the holes and tells me my idea sucks, one of them gathers evidence and just purely looks at it from an objective, logic, research point of view.
06:42There's a bunch of little personalities in this council, and they give me a full rundown. And remember how earlier I talked about the memory? All of these little members of the council also know what's currently working in my business, and they know what's going on.
06:53So they're able to use that context when they're running this council, which makes it even better. But the point is that every single point has to be backed by something real, not just vibes. And then at the end of that, it gives me basically one clean verdict.
07:04It says, go. The idea is good. Or it'll say, reshape.
07:06Or it'll say, just kill it. And only if the idea survives or maybe you have to reshape it a little bit, then you move into the actual build, which is step four. And sometimes you've got a big project.
07:15All you have to do is break it down piece by piece, and then you just start from the beginning and take baby steps. You don't ask it to do the whole thing. You start with the smallest version that actually works, you grow up from there.
07:23And when Claude tells you it's done, don't just take its word for it. Make it prove that it's done. Have it run the thing on a real example and show you the output.
07:30Verification is one of the most important elements of building with AI. Basically, just think about it if this was a human giving you work to review, what would you look at? Or what would you do to verify that this work is done and that it's good?
07:42And then you just ask the AI to verify it by doing those things. It can literally control a browser, so anything that you can do on a computer to verify something, the AI can do as well. And every day, they get better.
07:52However the AI is right now, it's the worst it will ever be. Now, your first version is probably gonna come back maybe like 60 or 70% right, and that's completely normal.
08:01You don't treat those misses like failure, you treat that like golden data. You treat that like feedback. Tell it what you liked and what was wrong in plain English, and then you let it fix its own work.
08:09And over time, it gets better and better because you're building skills around it and it's improving its memory. And if it ever does something that you don't understand, you literally just ask it to explain it to you. You treat this thing like a mentor.
08:19You treat this thing like your best friend who is also the smartest person in the world. That's how you actually learn this stuff. And every rep after that gets better because like I said earlier, this thing remembers.
08:27And when that first task works, pick the next one. And that's literally how I ended up running my entire business through Claude, one task at a time. And if you guys want a shortcut for all of this, I put everything in my free school community.
08:36The link for that is down in the description. If you've never opened Claude before, there's a seven day challenge in there that walks you through your first task step by step. The rose skill that I mentioned earlier, that's in there as well for completely free.
08:45And it's literally just a file that you drop into Claude and it runs that whole council for you automatically. And then maybe you probably wanna open it up and customize it a little bit for your own use case, but once again, you can ask Claude to do that. And then when you're ready to go deeper, inside of my free school community, there's also a free course on building what I call an AI operating system, which is basically the full version of everything I've been talking about in this video, and it is how I run all of my businesses through Cloud Code.
09:06Completely nontechnical. You don't have to know how to code. You just have to know how to think clearly.
09:10And look, the window that we just talked about is real, but you don't need to build the next Vulcan this weekend. Just $20 for the subscription and one task. That's the whole barrier to entry right now.
09:19That will put you ahead of 95% of people in the world right now. But, anyways, that is going to do it for this one. So if you guys enjoyed or you learned something new, please give it a like.
09:26It definitely helps me out a ton. I've got tons of videos on my channel about how to actually build with Cloud Code, so just go dive in. And as always, I appreciate you guys making it to the end of the video.
09:33I'll see you on the next one. Thanks, everyone.
The Hook
The bait, then the rug-pull.
A three-person team with almost no coding background just won a state government contract by describing what they wanted to Claude — and the host argues that story is about to repeat everywhere.
Frameworks
Named ideas worth stealing.
02:56list
The Four Things (why Claude Code is different)
It does the work
It's agentic
It works in parallel
It remembers
The host's stated reasons Claude Code specifically (not just chatbots generally) collapses the gap between having an idea and having a working product.
Steal forframing any AI-adoption pitch to non-technical founders
Council converges on one verdict: go / reshape / kill
A named pattern for spinning up several Claude sub-agent personalities to stress-test a business idea before committing to build it, so excitement alone doesn't drive the decision.
Steal forpre-build validation of any new feature or business idea
05:15list
Four-step starter path
Get a paid Claude plan ($20/mo)
Pick one real repetitive task
Describe it and run the roast
Build in small verified increments
The host's concrete on-ramp for a first-time non-technical builder.
Steal foronboarding sequence for any 'first project with AI' guide
CTA Breakdown
How they asked for the click.
VERBAL ASK
08:46newsletter
“if you guys want a shortcut for all of this, I put everything in my free school community”
Soft CTA delivered in the last 90 seconds after the value is fully delivered; points to a free Skool community with a paid upsell tier, plus a like/subscribe ask at the very end.
A 17-minute career roadmap arguing that the next move for anyone who can build with AI is to stop being a builder and start being a consultant — with a four-step playbook to do it without quitting your job.
00:08
00:14
00:22
00:33
00:37
00:46
00:58
01:07
01:16
01:23
01:30
01:35
01:44
01:54
01:55
02:10
02:20
02:27
02:32
02:39
02:45
02:55
03:02
03:08
03:15
03:23
03:29
03:39
03:43
03:52
04:08
04:14
04:21
04:31
04:35
04:47
04:51
04:58
05:06
05:16
05:20
05:27
05:36
05:40
05:46
05:58
06:00
06:12
06:25
06:30
06:39
06:46
06:54
07:08
07:14
07:22
07:28
07:35
07:45
07:52
07:55
08:05
08:10
08:20
08:27
08:38
08:45
08:56
09:02
09:10
09:18
09:28
09:31
Video of the Day1:08:12
05:46
34:20
14:35
16:53
12:15