We Are the Oil
Why the corporations building AI have more reason to cultivate us than to crush us
Sunday morning. Tea going cold. Dr. Károly Zsolnai-Fehér on screen, walking me through Gemma 4 — just released by Google as open source, runs on budget hardware, performs like something that has no business being free. A gift to humanity, he says. What a time to be alive. But also: what's the catch?
You Shouldn't Look a Gift Horse in the Mouth
I know. And yet here I am, looking.
It's a reflex, not cynicism. Corporations have earned their reputation for this sort of thing — the feel-good announcements, the carefully managed optics, the armies of spin doctors working to ensure that whatever evil is actually happening stays a few news cycles below the surface. Greenwashing, whitewashing, "don't be evil." A skeptical person doesn't take corporate generosity at face value. That's not paranoia, it's just pattern recognition.
So the question stands. Google spends what must be an obscene amount of money training a frontier model, then releases it to the world under a license that permits basically everything — commercial use included. Why?
To answer that properly, you need to understand what kind of corporation you're actually dealing with.
The Wrong Template
When people imagine the dystopias that AI-powered corporations might deliver, they're usually working from a specific template: take a ruthless, traditional corporation and hand it a superpower.
The common template for this is an oil company.
It's a fair instinct. Big Oil is the archetype of the corporate predator. Their business is built on a finite resource, and finite resources produce competition — sustained, grinding, zero-sum competition between companies and countries and anyone else who wants a piece. That kind of competition, over generations, doesn't produce virtue. It produces something lean, territorial, and constitutionally indifferent to the cost it imposes on others. Give that organism frontier AI capabilities and yes, the nightmares write themselves.
But the template is wrong.
Scarcity Versus Abundance
Oil companies operate in a paradigm of scarcity. Every barrel extracted by one player is unavailable to another. The game is structurally zero-sum, and it shapes everything about how they think and move.
Code is not a finite resource. Data is not a finite resource. They are produced — by us, continuously, in the ordinary course of building things and communicating and experimenting.
I grew up in the hillside olive groves of Andalusia, where we extract a different kind of oil. Unlike the crude you get by grabbing land and drilling into it, olive oil cannot be taken — only cultivated. You tend the trees through every stage: fertilization, pruning, harvest. The care you put in, at each step, correlates directly with the quality of what you get out. Neglect them and you get less. Treat them well and they give you more, year after year, for generations.
That's a closer analogy to what tech corporations are actually sitting on. (I'm keeping the title We Are the Oil because it's catchier than We Are the Olive Trees).
Which means that in a very real sense, we are the resource for these corporations. Not in the sense of being exploited and burned — in the sense of being the source. The thing that keeps producing as long as it's tended to. The well that, unlike the ones in Texas, doesn't run dry if you take care of it.
This reframes the incentive structure entirely. Mistreating the people who generate your feedstock isn't ruthlessness — it's self-sabotage. It pulls the rug out from under your own feet.
Actions, Not Words
None of this requires believing corporations are good. It just requires believing they're rational.
What matters, when trying to figure out whether something is genuinely good or a performance of goodness, is never the words. It's the actions. And the actions here are hard to dismiss.
Gemma 4, Apache 2.0 license, universally usable. Runs on hardware that won't bankrupt a solo developer. The cynical read is that it's a bone — a deliberately lesser product offered to hook users before funneling them toward the commercial tier. That logic applies cleanly to shareware, to freemium, to trial versions. It has never really applied to open source. The ethos is different: you release something because you want to see what others build with it, because the experiments that happen out in the open surface possibilities you'd never manufacture internally. That's not charity. It's untapped potential flowing through a channel you couldn't build yourself.
The leaders of these companies understand this. They grew up in open source culture. They've seen what happens when you give something away and let the ecosystem run with it.
The Greatest Crowdsourcing Experiment
Large language models are, at their core, the most ambitious crowdsourcing experiment in history. The scale isn't comparable to anything before. And the results have been spectacular.
Whether it stays that way — whether there eventually comes a point where these models are capable enough to generate their own training signal without human input — is genuinely open. The research on crowdsourcing and the wisdom of crowds suggests that human diversity of thought and experience isn't easily replicated. For now, we remain indispensable. The grove keeps producing as long as more people are experimenting, building, and putting results into the world.
Lower the barrier. More developers enter. More experiments run. More code and data get generated. More of what the big models need to keep growing. Releasing open models isn't philanthropy — it's irrigation.
Perhaps There Is No Catch
Which brings me back to Sunday morning, and the question I started with.
Maybe there isn't one. Maybe what looks like generosity is simply rationality operating at scale — calculated, profit-driven, and entirely self-interested, but pointing in the right direction. The profit motive doesn't negate the outcome. A corporation acting well because it's good for business is still acting well.
I don't underestimate what could go wrong. The acceleration is real, the concentration of capability is worth watching, and I have no illusions that these institutions are guided by anything other than their own interests. But from inside the developer community — where you can see what's actually being built, released, and shared — it looks like we might be on one of the better paths.
Not because the corporations are virtuous. Because the incentives line up.
That's a more durable foundation than virtue anyway. So enjoy the times. Push forward. Achieve together.