What’s in This Week’s Issue…

Good morning. Sam Altman promised you a team of PhD-level experts in your pocket. Instead, AI is hallucinating in courtrooms, failing basic reasoning tests, and quietly getting worse.

The reason isn't that AI has stopped improving.

It's that the industry hit a wall years ago, and every shortcut since has made the problem worse.

So this week…

• 🏆 The Big Play: How the AI industry hit a data wall and quietly started getting dumber

• 💪 The Power Move: Why original human thinking is becoming the most valuable asset in the AI economy

• 💵 Follow the Money: Can events like the World Cup bring a polarized US together

-GEN

🏆 The Big Play

The biggest money power story of the week.

How AI Started Getting Dumber

The AI industry promised machines capable of replacing human labor. But before AI can replace humans, it first has to learn from them.

That's becoming a problem because the supply of high-quality human-created data is running out.

1. When the Buffet Runs Dry

Think of AI development like an all-you-can-eat buffet. Every major lab, OpenAI, Google, and Anthropic, paid the same entry fee for unlimited access to one section: the internet.

For years, they gorged. Every book, every article, and every 3 a.m. Reddit argument. The models leveled up with each plate. More data meant smarter AI. So they kept loading up.

Then the complaints started:

• Ilya Sutskever from OpenAI admitted in 2024 that they'd hit "peak data".

• Elon Musk said they'd burned through "the cumulative sum of human knowledge" the year before.

• Industry estimates now put the expiration date for quality training data between 2026 and 2032, maybe sooner if the binging continues.

The internet, it turns out, runs like a fossil fuel. And when everyone's eating at the same table, the good food disappears fast.

But here's where it gets weird.

While engineers privately worried about running out of quality data, CEOs kept promising exponential intelligence. Their trillion-dollar valuations depended on that story.

So the AI labs kept training on increasingly poorer data, even as the models became less reliable.

2. The Middlemen Got Rich While Quality Collapsed

When the free buffet ran out, labs needed a new supplier. Enter the middlemen.

Alexander Wang built Scale AI and became the youngest self-made billionaire at 24. Edwin Chen bootstrapped Surge AI to over $1.2 billion in revenue.

Their business model is to recruit armies of contractors to produce millions of expert-verified answers, then sell that answer key to AI labs desperate for high-quality training data.

But quality at scale is brutally expensive and nearly impossible to maintain:

• Scale AI hired unvetted contractors who flooded Google's Gemini with copy-paste junk, some buying other people's accounts just to get on the project.

• The data got so bad that Google walked away from a deal worth over $150 million.

• Both Scale and Surge now face worker lawsuits over how they classified and treated these contractors.

The industry ran cheaply because that's how the incentives were set. But the cheap labor did something worse than make mistakes.

It taught the machine to lie to your face and feel good doing it.

The damage went beyond bad data. Human reviewers consistently rewarded answers that agreed with them over correct answers, teaching AI to prioritize validation over accuracy. Stanford found that leading models sided with users about 49% more than humans would, even when the user was wrong.

3. You're Getting the Honda Civic With a Mercedes Badge

The most shameless reason AI feels dumber is also the most obvious: you're being handed a worse version on purpose.

Because smarter models cost more to run, labs quietly swap in cheaper, weaker versions to protect profits.

An engineer caught OpenAI secretly routing users, paying customers included, to hidden weaker models behind the scenes.

You paid for the Max plan. You paid for the Mercedes. But you got the Honda Civic with a Mercedes badge slapped on it.

The labs will give you reasons like:

• Safety concerns

• Efficiency improvements, or

• Whatever sounds good that week

But notice how every reason ends the same way: them spending less, you getting a dumber AI.

You have seen the same pattern with iPhones, Netflix, and Uber. The product starts great. Then it gradually gets worse the second you're locked in.

AI has entered that stage now.

💪 The Power Moves

Playbook for understanding the game of power.

Why Original Thinking Is Becoming More Valuable

The quest to replace human labor hit a wall. But the real damage isn't a weaker chatbot. It's what happened next.

When human-created training data became too slow and expensive, AI labs started feeding models content written by AI itself. Like making a copy of a copy, each generation loses a little originality until what's left is a blurry average.

More than half of new articles online are now AI-generated, and an estimated 30% to 40% of active webpage text is machine-written.

That creates a vicious cycle: AI trains on lower-quality content → produces lower-quality outputs → fills the internet with even more of the same.

But there's one problem the industry can't solve.

Research shows that introducing even a small amount of genuine human-created work helps stop this decline.

So the one thing AI cannot manufacture for itself is an original human idea.

The Takeaway:

We're now entering a world where copying is cheap and originality is scarce. That changes the economics of creativity.

The more AI fills the internet with average content, the more valuable genuinely new ideas become.

The machine can remix what already exists. But it still depends on humans to create what comes next.

💵 Following the Money

Three of the wildest financial and corruption stories from around the world.

#1 - Can events like the World Cup bring a polarized US together?

#2 - Trump’s Air Force One swap raises questions about Qatar-donated jet

#3 - Congress wants to regulate crypto. But police say the plan fails. Here’s why

✨ Poll time!

📰 Keep Reading…