Welcome to the first What Is Essential newsletter. Each month I'll take one of the eleven essentials and test it against something happening in the world. We start with the one that sounds simplest and turns out to be the hardest.
She looks straight into the lens.
She doesn't blink, look away, or fidget. Elizabeth Holmes holds the camera's gaze the way she once held the gaze of investors, board members, and patients.
"I don't have anything to deceive you on. Of course I'm not deceiving you. I'm engaging with you as a human being. Why would I deceive you? There's no reason for me to do that."
Listen to how it's built. The denial comes first, then the appeal to being human, then the question that hands the burden back to you. Come up with a reason, it says. And when you can't, you trust her.
You Can See Everything is built on a promise of total access. Thirty-four days before prison, Holmes lets a film crew follow her everywhere. They keep filming for three years. Nothing is off limits.
And still, you leave the theater not knowing who she is.
Access is not understanding. Seeing is not knowing.
The performance of openness
We are being sold the same promise about artificial intelligence.
The new systems show their work. They walk you through their reasoning step by step. They explain why they answered the way they did. The language is calm, fluent, and sure of itself. It's eye contact rendered in text.
And we read it the way we read a steady gaze. As honesty.
But the explanation is just another output. It is produced by the same system it claims to explain. It is the machine looking into the lens.
The people who build these systems say this themselves. Interpretability, the science of understanding what actually happens inside a model, is still young. Researchers can trace fragments. They can't yet read the whole thing. We can watch every word appear and still not know why it appeared.
That is transparency as performance. A window that turns out to be a painting of a window.
The machine on the table
Holmes sets the Theranos machine on the table. She argues that it can do what it was supposed to do. It just needs the science to catch up.
Think about that claim. The device works. It simply doesn't work yet.
She has collapsed the vision into the fact. Because she can picture how it should function, she treats it as already functioning.
It's ridiculous when she says it. We nod when the AI industry says it.
We are told that artificial general intelligence is close. That we are on the threshold. Yet the same industry can't fully explain what its current systems are doing. You can't claim to be standing at a doorway when you can't read the floor under your feet. Declaring arrival when you don't understand the road you're already on isn't vision. It's the machine on the table.
What transparency actually requires
Theranos did not fall because someone finally saw inside the box. It fell because people tested the blood. Employees compared the results. A reporter checked the claims against reality. The truth came from verification, not access.
That is what the fourth essential means.
Transparency is not a view. It is a test that anyone outside the system can run. In practice, that asks for four things:
- Claims that can be checked. Not "trust our explanation," but results that hold up when someone independent checks them.
- Honest limits. Disclosure of what a system can't do, not just demonstrations of what it can.
- No future tense posing as present tense. What a system does today is reported separately from what its makers hope it will do.
- Consequences when the test fails. Openness without accountability is just a better camera angle.
Transparency feels like trust. It looks like trust. But it can be performed. And the better the performance, the less we think to check.
The Little Prince was told that what is essential is invisible to the eye. That doesn't mean we stop looking. It means we stop treating looking as enough.
Holmes asked why she would deceive us. She was counting on us not having an answer.
So the next time a system explains itself perfectly, will you be checking, or just believing?
What Is Essential is a human-centered AI movement built on one conviction: human judgment, accountability, and moral agency cannot be safely outsourced. Read the Mission and the eleven essentials.
Know someone who would argue with this issue? Forward it to them.