The Advantage #10: The $1.5B startup I would build cheaper today, the 1994 window, and the AI team trick

Welcome Note:

Something I keep coming back to: when I started Freshly in 2012, building a meal delivery company required a warehouse, a full-stack engineering team, a logistics operation, and roughly $10 million before you knew whether the idea worked.

I think we could have answered that same question in 2026 for under $50,000.

That is not an incremental improvement. That is a different game entirely.

This week, I want to share why I believe right now is the best time in history to start a company, what history tells us about how long these windows stay open, and one practical framework for getting started without waiting until everything is perfect.

Let's get into it.

What I Am Working On

Over the last few days, I have been building my personal content operation using OpenClaw AI agents instead of a team. I do have one hire helping me spin up my podcast (more on that in the next few months), but there are no full-time content hires. No agency retainers. Just a structured system where AI handles research, drafting, design, and analytics, with a goal of spending about 10 to 15 minutes per week reviewing and approving.

With that said, I have spent almost 10 hours this weekend and probably another 10 hours going back and forth with Will, my AI agent, over the last few days. So we are far from that goal. Early innings. But my mind is blown (way more than with Claude CoWork).

I sat down with my CTO at Petfolk yesterday, and we are figuring out how to spin several up there. We already have one going for Cutting Horse. Security is a real concern. You do not want to run in with your eyes shut. But WOW. OpenClaw is not for beginners, but it is pretty incredible.

I will publish a longer piece on this over the next several weeks as I work through exactly how I am using it. Stay tuned.

Worth Watching

Quick intro:

This week’s Worth Watching: a 12-minute clip where Marc Andreessen makes the case that this is clearly bigger than the internet.

What I loved about it:

Andreessen makes one argument: every major productivity era was driven by a collapse in the cost of something fundamental. Steam collapsed physical power. Electricity collapsed light and heat. The internet collapsed distribution. AI is collapsing the cost of intelligence itself. I wrote about this framework (I called it SHIH, Synthetic Human Intelligence Hours) in Fortune earlier this year. When the cost of intelligence drops 100x, every industry built on expensive human cognition becomes a target.

Here is my 20-second recap if you do not have the full 12 minutes:

AI is infrastructure, not a feature. Just as electricity restructured every industry rather than giving one company an edge, AI will not help incumbents slightly. It will create entirely new categories and make others obsolete.

The cost of starting has collapsed. What required $10M in engineering investment in 2019 requires $50,000 today. The barrier to entry for great ideas has never been lower.

The incumbents are still asleep. Every large company is running AI pilots. Almost none have restructured around AI. Small, fast-moving teams outmaneuver large organizations during transitions. That is how Amazon beat Barnes and Noble.

The window is narrow. The early internet window was roughly 4 years (1994-1998). Andreessen’s argument: we are somewhere in that equivalent window right now.

Pro Move: You do not need to watch it. Put it on like a podcast and listen while you walk or drive.

Lesson From History: The 1994 Window

What Happened:

In 1994, Jeff Bezos was a vice president at the hedge fund D.E. Shaw in New York. He was 30 years old, well-compensated, and on a clear path. Then he read a data point that changed everything: internet traffic was growing at 2,300% per year. He made a list of categories where that kind of distribution advantage would matter most. Books won. He moved to Seattle, incorporated Amazon in a garage, and started taking orders.

That year, fewer than 3,000 websites existed on the entire internet. Amazon was one of them.

By 1995, Yahoo launched. By 1998, Google launched. By 2000, the window was mostly closed. The companies started in that four-year window between 1994 and 1998 (Amazon, Google, eBay, PayPal) became the defining businesses of the next 25 years. The companies that waited mostly failed.

Barnes and Noble had 30,000 employees and $2 billion in revenue when Amazon started. They had every advantage except one: they moved second.

Insight behind it:

There is a brief window at the start of every major platform shift when the infrastructure is in place, the tools have democratized, and the incumbents have not yet restructured around the new reality. Amazon was not the best-resourced company to sell books online. They were the most urgent. Timing plus urgency is what creates category-defining companies. The technology is never the moat. The timing is.

Modern application:

The AI infrastructure window opened in late 2022. The compute exists. The APIs are accessible. The cost of building is a fraction of what it was three years ago. The incumbents are running pilots. And the category-defining companies of the AI era are being built right now, by people sitting in exactly the position Bezos was in during 1994, reading a data point that stops them cold and deciding whether to move.

The question is not whether the opportunity is real. The question is whether you are moving before Barnes and Noble does.

Practical Edge: Build Your First Version in 30 Days Using AI

Why it works:

The biggest lie in entrepreneurship is that you need to be ready before you start. You do not. You need to be curious, you need a clear problem, and you need to move. The 30-day rule forces both.

Here is the reality in 2026: the cost of a minimum viable product has collapsed. You can build software without a technical co-founder. You can run marketing without an agency. You can handle customer support, bookkeeping, and operations with tools that cost $200 a month. The excuse of needing more time, money, or resources is no longer valid for anyone willing to learn the tools.

The data supports it:

First Round Capital’s analysis of 10 years of portfolio data found that the single biggest predictor of startup failure was running out of time, not money. Speed to proof is the most valuable asset a founder has.

How I use it:

When I started Freshly, our first test was a WordPress site and a Google Form. We did not build a product. We built a hypothesis and tested it for $0. The answer came back in two weeks.

I have watched dozens of founders spend 18 months building something nobody wanted. The 30-day rule would have saved most of them.

The challenge:

Pick one idea you have been sitting on. This week, write down the single question that, if answered, would tell you whether the idea is worth pursuing. Not 10 questions. One. Then spend 30 days finding the answer as cheaply and quickly as possible. Use AI for every part of it: research, copywriting, landing page, customer outreach. If you cannot validate the core assumption in 30 days without spending real money, the idea either needs to change or you are not the right person to build it.

The window is open. The tools are free. The only thing left is the decision to move.

Thanks for reading,

Join The Advantage, .