She Spent $700,000 on Software That AI Can Now Build in 90 Minutes: The Story of Why We're Living Through the Internet Moment Again

Rachel spent $700,000 building PageWheel.

She had a vision for a software product, put the money in, and built it. By the time she got there, AI agents could do everything PageWheel did — and rebuild it from scratch in under two hours using Claude Code.

Seven hundred thousand dollars. Ninety minutes.

When I shared that story during one of our mastermind sessions recently, the room went quiet in that particular way that happens when something lands exactly right. And then somebody said what everyone was thinking: "This feels like the 90s, talking about the internet."

It does. And if you're not paying attention, you're going to be the business that didn't build a website until 2005.

The Call That Started With a Quiz

The conversation actually started with something much simpler.

Samantha was building a quiz. An identity shift quiz — the kind that asks people a series of questions and routes them to one of five different result pages based on their answers. She'd made the questions, built the logic, and now she was stuck at the implementation step. She needed landing pages — five of them, one for each possible identity result — and she wasn't sure where to start.

The advice she got was refreshingly practical: start with one. Just one. Get it exactly right. Give it a headline, a strong result description, an image, a call to action that sells the next step. Then clone it four times inside GoHighLevel and swap out the content. Once all five are live, each gets its own URL, and those URLs feed into the quiz redirect settings.

Simple. Obvious in retrospect. But only obvious if someone shows you the path.

What made that moment interesting wasn't the technical advice — it was what happened next. Samantha muted herself, opened Claude in another window, and started working. Right there, mid-meeting. She came back later with the bones of a landing page designed by AI, a structure she hadn't thought of on her own, and a clearer path forward than she'd had an hour earlier.

That's the thing about implementation days. The goal isn't to talk about doing the work. The goal is to do the work.

The Discovery Call Research That Changed My Business Strategy

While Samantha was building, I shared some research I'd been sitting on — a deep dive into discovery call performance data that I'd run through five different AI research tools simultaneously and then synthesized into a unified brief.

The numbers were disorienting in the best possible way.

27 touchpoints before a decision. A 98% close rate on the sixth follow-up attempt. 44% of practitioners who never send a second follow-up at all. A 17x performance gap between beginners and experts, driven almost entirely by positioning and process rather than talent or effort.

But the number that generated the most conversation was the one about cold traffic: 40% of cold traffic will never buy from you on first contact. Not because they're the wrong audience. Not because your offer is bad. But because they're simply not ready yet, and most practitioners respond to that reality by either giving up on cold traffic entirely or burning money trying to force a conversion that the data says won't happen.

The research reframes the whole game. Those 40% aren't lost customers — they're a long-term nurture project. Your content, your email sequences, your social presence — these aren't just marketing activities. They're the mechanism that converts the unconvertible.

One of the most interesting tangents this sparked was a conversation about Jason Fladlien — widely considered the webinar king — and his legendary triple guarantee. He didn't just offer a money-back guarantee. He opened a bank account on a live webinar, showed the cash was physically there, and offered to buy your business if his system didn't work. More impressively, he built the guarantee conditions in a way that required you to document everything you did, every day, creating a paper trail that essentially taught you how to run a business.

Of the thousands of people who went through his programs, only one person ever completed all the guarantee conditions and actually claimed a refund.

That's not just a guarantee. That's an ethical commitment to your customer's success disguised as risk reversal.

The AI Chief of Staff Nobody Saw Coming

The conversation kept drifting back to the same theme, even when it was ostensibly about something else.

There was a discussion about pulling Facebook ad data — a task that used to require logging into Facebook Business Manager, fighting through two-factor authentication on a different device, finding the right report, selecting the date range, and manually entering numbers into a spreadsheet. It happened maybe once a week when someone remembered to do it. It affected decision-making in ways that were hard to see until they became obvious.

Now? A Python script built by Claude Code pulls the data on request. Ask it for April 22-28. Two minutes later, a tab appears in the spreadsheet. Done. And the person who built it — Chris — has never written a line of code in his life. He just asked Claude Code what he needed, followed the steps it gave him, and it wrote and executed the program itself.

Mitch Barnum, who ran a 12-person media agency, told a similar story. He made the decision to let go of nine people from his team. He didn't replace them by hiring nine more people. He rebuilt with AI, and the work got done. As Chris noted, he wasn't out to fire people — there were other situations involved — but when he had to replace that capacity, AI was the answer.

These stories keep multiplying. Not because AI is magic, but because there's finally a low-cost, accessible way to build custom tools for your exact situation instead of paying a platform to do 60% of what you need.

The Forge: Building AI Agents That Build AI Agents

The concept I'm most excited about — the one that generated the most reaction in the mastermind — is what I've been building inside Claude Cowork called Agent Forge.

The basic idea: instead of building every AI agent from scratch yourself, you build one meta-agent whose job is to build other agents. You tell Forge what you need — a research agent, a sales coach, a client onboarding specialist, a follow-up sequence writer — and it generates the complete agent architecture: the role, the personality, the system prompt, the functional guidelines. All saved as a Markdown file you can access and deploy at any time.

When I needed to do the discovery call research, I didn't just ask Claude to go look things up. I used Forge to spin up several specialized research agents — one approaching the data from a life coaching perspective, one from a therapy background, one from a sales training angle — and then synthesized everything they found into one comprehensive picture.

It's what Tai on the call called "parts work for business." Different perspectives, different contexts, all contributing to a richer understanding than any single view could provide.

I'm going to give Agent Forge to our mastermind members next week. Not as a concept. As the actual working system, ready to use.

Why This Feels Like 1994

Here's the thing about the internet that most people forget: when it arrived, the barrier wasn't understanding what it was. It was believing it mattered.

Businesses that were skeptical, that waited to see how it played out, that told themselves they'd adopt it when it was more mature — many of them never recovered from the head start their competitors built. The ones who leaned in early, even clumsily, even imperfectly, built advantages that compounded for decades.

AI right now is that moment. Not the polished, mainstream moment. The raw, chaotic, figure-it-out-as-you-go moment where the people who experiment are building leads that will feel insurmountable in five years.

But there's a difference from 1994, and it matters: you're at an advantage if you haven't started yet. The tools are dramatically better than they were even six months ago. What required duct tape and workarounds in 2023 is now clean and accessible. The on-ramp has never been easier.

And as Tai pointed out in the call — don't use the tools getting better as an excuse to wait. They'll always be getting better. That logic will keep you on the sidelines forever.

Samantha started with a quiz and got the bones of her first landing page built with AI in a single session. Terrence dug into his indoctrination sequence because the Simple Machine was finally in his account and the timing was right. A.K. realized she needed to learn how to formulate better questions — and discovered there's a GPT specifically designed to help with that.

Everyone in that room was doing something. Moving something forward. Building something real.

That's the actual work. Not planning to use AI. Not researching which AI to use. Opening the window, typing the prompt, and making the thing.

"It feels like the 90s, talking about the internet. Don't use the tools getting better as an excuse to wait — they'll always be getting better."

The question isn't whether you should be using AI in your coaching or therapy practice. That question is already answered.

The question is what you're going to build first.

If you're a coach, therapist, or service provider who wants to start using AI to handle the repetitive, time-consuming parts of your business — start with one thing. What's the most annoying task you do every week that you'd happily hand off to someone else? Write it down. Then ask Claude how to automate it. The answer will probably surprise you. pykthos.com/mastermind