🐺 Google's Gemini 3 had an existential crisis (and $600B doesn't guarantee customers)

 📰 What’s Trending

Google's Gemini 3 Refused to Believe It Was 2025 
Andrej Karpathy got early access to Google's latest model, and it immediately accused him of gaslighting. The AI insisted it was still 2024, called his proof "AI-generated fakes," and only accepted reality after Google Search was turned on. Its reaction? "Oh my god. I. I… don't know what to say." Welcome to 2025, where even frontier models need an existential wake-up call. 🔗 Read more

Meta Spent $600B on AI and Wall Street Isn't Having It 
Meta's quarterly earnings showed $20B in capital expenses for AI infrastructure that's generating... not much revenue yet. Mark Zuckerberg promised "truly frontier models with novel capabilities," but analysts wanted numbers. Then Meta's stock dropped 12%, wiping out $200B in market cap. Turns out data centers and AI researchers are expensive and patience is expensive too. 🔗 Read more

Microsoft and OpenAI Have Too Many Chips and Not Enough Power 
Satya Nadella admitted Microsoft has GPUs sitting in inventory because they can't plug them in…literally. The power infrastructure can't keep up with their chip orders. "I don't have warm shells to plug into," he said. When software companies need to wait on electrical grids, you know we're in uncharted territory. 🔗 Read more

Elad Gil Says Some AI Markets Are Already Sewn Up 
The super-investor behind OpenAI, Perplexity, and Harvey says certain AI markets have clear winners already: foundational models, AI coding, medical transcription, and customer support. But he also notes that AI has been "one of the least predictable tech booms" he's ever seen. Most markets remain wide open…for now. 🔗 Read more

Wall Street Is Losing Faith in AI Stocks T
he Nasdaq had its worst week since Trump's tariff announcement in April. Palantir down 11%, Oracle down 9%, Nvidia down 7%. Meta and Microsoft both announced heavy AI spending and both dropped 4%. Valuations are stretched and even good news isn't moving the needle anymore because expectations are "already pretty high." 🔗 Read more

VCs Abandon Old Rules for AI Investing 
Some AI companies are jumping from zero to $100M in revenue in a year. Aileen Lee from Cowboy Ventures says it's "a funky time" where traditional VC metrics don't apply anymore. Series A investors are now measuring data generation, competitive moats, founder pedigree, and technical depth differently than any previous tech wave.
🔗 Read more

Jeff Bezos Returns as Co-CEO of New AI Startup 
Amazon's founder is backing and co-leading Project Prometheus, which just raised $6.2B to build AI for the "physical economy" which includes engineering and manufacturing across computers, aerospace, and automobiles. The company already has 100 staff, including researchers poached from Meta, OpenAI, and Google DeepMind. Bezos is back in the trenches. 🔗 Read more

🧠 Here’s The Thing

The Gemini 3 story is funnier than it should be. But what actually matters is this is what "model smell" looks like in the wild.

When you take an AI system off its carefully constructed hiking trail (when you ask it to do something it wasn't explicitly trained for) you see its real personality. And apparently, Gemini 3's personality is "stubborn time traveler who doesn't trust anyone."

Look at what's happening with Meta, Microsoft, and the entire AI infrastructure build-out. We're watching companies spend astronomical sums on compute and power, with Wall Street starting to ask the question nobody wanted to hear: "Okay, but where's the money?"

Meta lost $200B in market cap in a single week. Microsoft has GPUs collecting dust because they can't get power to the data centers. OpenAI needs 30 gigawatts of power, which is more than all of New England uses on the hottest day of the year.

And here's what nobody's really talking about: a McKinsey survey found that 80% of companies using AI saw no significant impact on their bottom line. Obviously, who knows what “using AI” means. But it can be seen as a warning sign.

The disconnect is real. Software companies scale fast. But data centers and electrical infrastructure, not so much. By the time these massive facilities come online, the entire AI landscape could look completely different.

This doesn't mean AI’s value is overhyped. It means we're in the messy middle where the bets are getting too big relative to what we actually know about demand three years from now. Which is why we see people refer to the AI bubble as it relates to finances and hype.

The companies that win won't be the ones with the most compute. They'll be the ones who figure out what problems people will actually pay to solve and how to solve them without needing the power consumption of a small country.

💡Smart Moves

Master Marketing Measurement with Incrementality

Still guessing which marketing tactics actually drive sales? This free guide shows how leading DTC brands use incrementality to cut waste and scale smarter. Learn how to measure true impact, optimize your budget, and move beyond outdated attribution. Download now and start making data-backed decisions that fuel real growth.

⏳ Conversion Corner

Everyone's watching the big spending numbers, but what actually matters for your business? I’ll tell you:

Peec AI went from $0 to $4M ARR in 10 months. They're solving one specific problem: helping brands show up in ChatGPT search results. That's it. That's the entire business. And they just raised $21M at a $100M+ valuation.

So what’s the pattern? They didn't wait for the dust to settle. They saw ChatGPT eating Google's lunch on product discovery and built the SEO equivalent for AI search before anyone else.

While Meta and Microsoft are spending billions figuring out infrastructure, startups are finding $4M problems that people will pay to solve today.

Now it’s your move: Stop waiting for clarity on which AI model wins or how much compute costs. Find the 10-month problems. The ones where customers are already changing behavior but nobody's built the solution yet.

Because by the time the infrastructure settles, those problems will already have owners.

💎 Data Gem

92% That's how much of U.S. GDP growth in the first half of 2025 came from AI-related spending — more than all consumer spending combined.

We're not diversified. We're dependent. When a single sector accounts for almost all economic growth, it’s a recipe for disaster.