ChatGPT, Gemini 3, and Code Red: The AI Drama Just Got Real

Get your popcorn ready. Things are about to get interesting.

The internet has been buzzing for a week now following Sam Altman’s declaration of a state of emergency (he literally called code red) after the very successful release of Gemini 3. (Nicely done, Google!) How successful? According to the ever-subtle CEO of Salesforce:

“Holy shit. I’ve used ChatGPT every day for 3 years. Just spent 2 hours on Gemini 3. I’m not going back.”

It’s Not Just Model vs Model. It’s Reality vs Narrative

But the truth is, this isn’t really a story about one model versus another. It’s a story about what happens when reality finally catches up with the fantasy.

For three years, we were captivated by a beautiful myth: A small, unprofitable company invents a technology that quietly, and right under the nose of the world’s biggest tech giants, reshapes humanity and becomes an almost mythical brand. But myths only last until physics taps you on the shoulder to remind you: Hey, I’m not going anywhere. Data centers cost trillions. Electricity isn’t free. And ultimately, no one enjoys an irreversible advantage in a field that is rapidly converging toward commoditization.

Ecosystems Win the Long Game, Not Just Technologies

Historically, innovation does win, but ecosystems win more. Microsoft didn’t beat Netscape because it had a better browser; it won because the browser came bundled with Windows. Apple didn’t beat Nokia because of its (admittedly stunning) hardware; it won because apps became an ecosystem. By that logic, the real question now is: Who can turn raw capability into a better everyday experience? And here, OpenAI still holds a card no one should underestimate: the ChatGPT brand. It’s a brand that became a verb, and the fastest-growing consumer product in history, going from zero to 800 million users in just three years.

But it turns out that having more assets (and more users) also means having more problems. Google has managed to close an almost mythical gap. Anthropic is sprinting forward. DeepSeek is proving that open source is very much alive. And beyond all that, it’s not just OpenAI’s user growth that’s historic, it’s also the pace of its cash burn. 

Analysts at HSBC recently showed that even in the most optimistic scenario where ChatGPT grows to 3 billion users by 2030 (nearly half of the global population above age 15!) the company would still need to raise $200 billion during that period just to balance its exploding revenue forecasts and spending plans. Next time you pledge to invest $1.4 trillion in data centers and bring in only $20 billion a year, remember: math is unimpressed by brands.

We’ve Entered the Next Phase of the AI Race

I’m not a gambling person (which is ironic, considering my job is investing in startups at the deck stage, when most of the important questions don’t yet have answers). But if I were, I’d bet that we’ve entered a phase where the technology itself is becoming less interesting than the product.

Until now, every model release prompted users to rush and compare benchmarks (standardized tests that measured model performance on identical tasks for objective comparison). But benchmarks measure performance, and markets measure usage: What do people come back to? What do they willingly pay for, again and again? What becomes so deeply woven into their daily lives that it turns into something they simply can’t function without? (For me, that’s caffeine! Thank you very much.)

So, the innovation ahead won’t be about how fast a model responds or how big its context window is. We’ve leveled up, Mario style. And things are about to get very interesting.