The Carnival Before The Extraction
On the current AI summer and the coming clouds
Right now, it is genuinely amazing to be a consumer of AI.
I don’t want to dismiss the profound anxieties: the fears about job losses, the erosion of human creativity, and the sheer pace of change. All valid, all very real. But for a moment, let’s step back and appreciate the absurdity (and exuberance!) of our present reality. We have access to the most powerful cognitive tools ever created, for basically nothing.
ChatGPT, Gemini, Claude, Perplexity: all these engines are either free or about the price of a Netflix subscription. This is not normal. This is not sustainable. What we have right now is a “capitalist subsidy.” Big Tech is burning billions of dollars in a frenzy of land grab, and we are the temporary beneficiaries.
It Smells Like 1996 All Over Again
I am a Gen X, and the feeling is eerily familiar. The Internet, the early Web, everything felt possible. Decentralized, exploratory, innocent. You could learn HTML (the basics) in a weekend and put out something that could grow huge.
Then came the ads, the algos tuned for addiction, the industry consolidation. And just like that, in less than a decade, we went from the idealism of “information wants to be free” to the grim economics of “you are the product.”
AI today feels like that same exuberant, open-feeling phase. But the extraction phase is coming soon; in fact, it is already here.
The Extraction is Here
Case in point, OpenAI just made two announcements in the span of a few weeks that tell us everything we need to know about the future.
First, they are rolling out ads in ChatGPT. Sam Altman, who once called the idea of combining ads with AI “uniquely unsettling,” has reversed course.
Second, they are launching “Adult mode” —age-verified access to mature content and erotica.
Why? The reason is simple. Because OpenAI has committed to spending up to $1.4 trillion in infrastructure and data centers over the next 8 years. But their current unit economics simply do not work (optimistically, a mere $20 billion in annual revenue in 2025)
These are not product innovations; they are revenue panic. This is what happens when the bills come due.
Sovereigns and Serfs
And when the bills come due, and they will, we are headed towards an Intelligence Divide. Not just between those who can afford the “Pro” tiers and those who can’t. But between those who will have access to the real frontier AI model versus those with the lobotomized, ad-saturated, and privacy-flawed version.
Two tiers of societal AI:
The AI Sovereign - Entities with private models, compute, and clusters, clean data, and 100x productivity gains.
The AI Serf- The rest of us, fed “lite” models that harvest data, serve ads, and lock us into subscriptions we can’t afford to leave.
The Inequality of Reality
Today we’re playing with AI chatbots. But soon, AI will be baked into education, healthcare, legal aid, and governance.
When that happens, if this technology is owned by the same handful of companies that dominate our digital lives, we’re not just looking at inequality of access. We’re facing inequality of reality itself.
Imagine two parallel worlds.
In one, AI amplifies human judgment, fosters creativity, and expands freedom.
In the other, AI replaces human judgment. Standardizes thought. Optimizes solely for extraction.
This isn’t science fiction. It’s the logical endpoint of business models built on extracting value from user attention and data.
While There’s Still Time
We’re in a rare moment where the future is not yet fixed. The decisions we make today (or not), about model ownership, data rights, and open-source infrastructure, will define the next decades.
Alas, it is all happening while we’re exuberantly distracted by how cool it is that AI can draft our emails, make bland music, or generate meme videos.
We must learn from the errors of the ‘90s.
While we enjoy this AI Summer, let us:
Support Open Source. Robust open-source AI is the strongest counterweight to a closed future. Transparent, modifiable models are the bedrock of any democratic alternative.
Demand data agency. We need frameworks that treat our queries, feedback, and prompts as more than free training data. Data dignity is essential.
Advocate for public AI infrastructure. What does the AI equivalent of public roads or libraries look like?
Stay cognitively independent. Use these AI tools, but never outsource our human judgment. Our discernment is the one thing that can’t be automated. Yet.
Let’s enjoy this AI summer. It’s real. It’s fun.
But fences are being built right now, around us, with our cheerful compliance as we bask in the convenience.
Let’s ask who owns the park. Let’s agree on the rules.
Because when the gates click shut, it will be too late to wish we had fought for a different design.