The Invisible Editor

Larry Ellison, the co-founder of Oracle, once told a room of investors why he likes the idea of constant surveillance. Citizens, he said, will be “on their best behavior, because we’re constantly recording and reporting everything that’s going on.” He was talking about cameras and policing. Most coverage filed it under public safety. At face value, that's what it is. But if you've been paying attention to what's actually being built, it reads as something else: a statement about power. After several years studying how these systems work, I'd call it one of the more honest admissions a tech executive has said out loud in a while.

The same systems that are learning, in extraordinary detail, what moves you are increasingly the systems that decide what reaches you. That convergence is the story. Not privacy. Not “misinformation.” Not even the kind of censorship we usually picture. The danger worth your attention is what becomes possible when the same machines that study human behavior also decide what each of us sees next.

We are all easier to manipulate than we'd like to admit, no matter how smart, informed, or skeptical we are. In 2024, a rumor tore through conservative media claiming Haitian immigrants in Springfield, Ohio, were stealing and eating people’s pets. Local officials found no evidence for the claim. The story spread anyway — because it was vivid, emotional, and confirmed a fear some people already carried. Years earlier, a short clip seemed to show a teenager in a MAGA hat confronting a Native American activist at the Lincoln Memorial. Politicians, journalists, and celebrities raced to condemn him; longer footage later revealed a far more complicated picture, and the student, Nicholas Sandmann, reached settlements with several major news outlets.

Different politics. Different audiences. The same human wiring. A short emotional story arrives before careful thought does, and by the time context catches up, the reaction has already happened. If you assume you’re immune, you’re exactly the kind of person these systems work best on, myself included.

If people are the vulnerability, recommendation algorithms are what turn that vulnerability into a system — and they are remarkably good at finding those soft spots. Every pause, replay, comment, and share teaches the system what holds you, what angers you, whom you trust, and whom you reject. It doesn’t need to control your mind. It only needs to optimize: show you more of what keeps you, less of what loses you, and get a little better every day at predicting what you’ll engage with next.

It doesn’t work by sending millions of people the same message. It works by sending millions of people different messages. The person worried about crime gets one version of reality. The person worried about discrimination gets another. Immigration, religion, the economy, public health, foreign threats — each fear can get its own optimized feed. The system doesn’t have to invent your fear. It only has to find what’s already there.

All of this points to what is actually being fought over, and it isn’t your vote or even your beliefs. It’s your attention. The power here isn’t mainly that an algorithm changes your mind; it’s that algorithms shape what enters your mind in the first place. Your attention is finite; nobody can track every issue. So whatever fills your feed informs your sense of what matters. You don’t need to be convinced that immigration is the nation’s biggest problem if it’s simply what you see 10 times a day. Attention arrives before belief — and attention is cheaper to buy.

This kind of influence is nearly impossible to catch in the act. Old-fashioned censorship left evidence. A banned book, a blacked-out column, a network that won’t cover a story — each leaves a visible gap you can point to and debate. Algorithmic suppression leaves nothing to point at. There’s no book to ban and no broadcast to boycott, only content that quietly never arrives. When a story never reaches you, there’s no missing page, no blank space, no notice that anything was withheld. You don’t know about the video you never saw or the perspective you never encountered. And because everyone’s feed is different, we can’t easily compare notes and catch it together the way past generations could when they at least shared the same front page.

We already know these systems get tuned by whoever owns them, because it has happened in public. After buying Twitter, Elon Musk reportedly had the platform boost his own posts. The lesson isn’t about Musk. It’s that whoever has the controls holds a lever on attention itself.

Nowhere is this clearer than with TikTok — usually covered as a business story, when it's really the sharpest example we have of the same hands holding both the data and the distribution.

Roughly one in five American adults now regularly get news from TikTok; among adults under 30, it’s closer to half. For those people, the algorithm is the editor — deciding what appears, what vanishes, and what gains momentum. Traditional editors had names on a masthead and could be challenged. This editor is largely invisible to the people it informs.

In January 2026, that editor changed hands. ByteDance divested TikTok’s U.S. operations to an American-controlled joint venture, with Ellison’s Oracle as one of three managing investors. Under the deal, Oracle is positioned both to secure American user data and to help retrain and oversee the recommendation algorithm itself. Read that again: one company, sitting close to both the behavioral data and the distribution engine.

The route to this deal was not a straight line. In his first term, President Donald Trump moved to ban TikTok outright over security concerns. He reversed course after returning to office — by his own account crediting the app with helping him win younger voters in 2024 — and then pushed for Oracle, co-founded by his friend Ellison, to sit at the center of the deal. Whether you like Trump isn’t the point. What matters is the pattern: a sitting president came to believe a platform had helped elect him, and then he stepped in to decide who would control that platform's algorithm. A future president of the other party, with a friendly platform owner, would face the identical temptation. That’s the danger — this power belongs to the office and the infrastructure, not to one man.

I am not suggesting that Oracle employees sit in a room deciding what every American believes. The danger is structural, not conspiratorial. Imagine a period of real political unrest. A system that already knows who fears disorder, who distrusts institutions, and who feels economically squeezed does not need to show everyone the same feed. It can show each group a different reality, tuned to a different emotion. Whether or not anyone ever abuses that capability, it now exists.

The deepest danger isn’t Ellison, or Trump, or Oracle, or TikTok. Administrations change, executives retire, alliances shift. Infrastructure stays. Once a system can gather behavioral data at massive scale, profile a population in detail, and shape what millions of people notice, every future leader inherits that capability — and every faction will tell itself it can be trusted with the very power it would never trust the other side to hold. History suggests tools like this rarely stay in the hands that first hold them.

We are building the pipes through which a growing share of people understand the world, and we are concentrating control of those pipes inside a handful of private systems. But none of this is inevitable — it is the result of choices about policy, regulation, and priorities, and choices can be remade.

Algorithmic governance is almost entirely absent from our politics. You will rarely hear a candidate raise this issue. That must change. Ask the people who want your vote where they stand on algorithmic transparency. Ask whether independent audits should be required for systems that inform millions. Ask who should be allowed to collect behavioral data at this scale, and what the limits are. Notice when a story seems to be everywhere, and ask what you might not be seeing. Compare feeds with someone who disagrees with you; the gaps are the most interesting part.

The people who decide how these systems are governed are, ultimately, the people we elect.

Best behavior is the response they’re counting on. We can choose a different path.

Share this with someone whose feed looks nothing like yours — and subscribe to Safe Online Futures.