Pick up a five-dollar bill. Look at it closely.
What you're holding in your hand is a small rectangle of cotton and linen fiber, printed with green ink. That's it. There is nothing in its physical composition that makes it worth five dollars. It is not five dollars because of what it is. It is five dollars because everything that's been built around it that says it's five dollars—the Federal Reserve, the institution of exchange. A collective agreement that says that this particular piece of paper can be traded for five others, or for a coffee, or for a percentage of a bill at dinner. The value is not in the small rectangle of fiber you're holding in your hand. It lives in a vast network of relationships, agreements, and shared frameworks that we mostly never think about.
Philosopher Jay Garfield explored this at length with Sam Harris on episode #282 of the Making Sense podcast. Conceptual dependence, he calls it—the idea that things exist and hold meaning not independently, but through the webs of understanding we build around them. The self, Garfield argues, works the same way. We are not autonomous beings who enter the world and then interact with it. We emerge from it—shaped by language, relationships, and the physical and social environments we're embedded in before we ever have a way to describe any of it. We don't discover our identities. We construct them. And the shape of that construction is a direct byproduct of everything that surrounds us.
It may sound abstract, but this idea has very practical implications for how we're building AI.
Because the dominant model of AI development inherits this assumption without quite meaning to. It learns from what you've done and gets better at anticipating what you'll do next. Impressive, and genuinely useful. But it can only work within the boundaries of what came before it—and what came before it assumed that what you do tells you what you need.
The problem is that humans don't work that way. We—as Garfield, and David Hume before him, in his Treatise of Human Nature, described—are natural artificers: meaning-makers by design. We don't simply have needs; we construct frameworks for understanding those needs, and those frameworks are always in motion, always dependent on context, and always shaped by the relationships and conditions around us. The same action, in two different people, can mean entirely different things. The same purchase. The same search. The same reel that caught the same 30 seconds of attention.
When a system is trained to optimize for (historical) behavioral signals, it gets very good at reading the surface. What it cannot read is (present) orientation—the direction you're moving, the values you're building toward, the gap between who you are right now and who you're trying to become. The result is a loop with no clear exit.
And this is not a minor gap—because ultimately, it's the difference between a system that's designed to serve you and a system that's designed to shape you.
Language compounds the problem.
Language is how we understand ourselves. When we acquire it, we acquire the medium through which we interpret our actions, our relationships, and our place in the world. It is foundational—not just for communication, but for identity. We are, in a meaningful sense, constructed through language.
But language is also a constraint. It imposes categories. It reduces the full complexity of experience into what can be named and what can be measured. And it is not universal—not across cultures, or literacies, or the spectrum of human experience. A system built entirely on language, however sophisticated, will always be limited by what language can hold. Which is not everything.
Consider what sits at the top of Maslow's hierarchy—transcendence and self-actualization. The states that, once you become aware of them, serve as the north star for the entire journey. These are, almost by definition, the places where language struggles most. We spend hours discussing the transformative nature of meditation—a practice whose entire goal is the quieting of the mind. And then, almost immediately, we try to put that silence into words. And we expect AI—a technology built on the foundation of language, trained on the articulation of human experience rather than the experience itself—to help us get there.
It cannot. Not with the tools it currently has.
Conceptual dependence suggests a different path.
If value emerges from frameworks rather than objects, then the solution is not to measure behavior more precisely. It is to expand beyond existing frameworks by integrating a language that transcends numbers and words—one that can meet people in their present state rather than trap them in their past one.
This is the argument behind the Visual Universal Language (VUL) at the core of HumaneAGI. The VUL is not a replacement for language. It is a layer beneath it—a needs-based code built on three channels: Red for the body, Green for the heart, Blue for the mind. It doesn't track what you've done. It reads where you are. And because it's non-linguistic, it works across cultures, literacies, and the conditions where words simply aren't enough—including, crucially, the ones where no words exist at all.
The VUL, at its core, is a framework argument. The premise is the same one Garfield draws out of the five-dollar bill: value is not intrinsic to the signal. It emerges from the framework around it. A behavioral signal—a click, a purchase, a pattern—is just ink on paper until you have a framework sophisticated enough to understand what it actually means. Right now, what exists is a system that reads the ink and calls it value. And the cost of that mistake is significant.
The machine is getting better at knowing us. It can predict what you'll watch next, what you'll buy, when you'll engage. But prediction is not understanding. What it cannot yet do is read the distance between who you are and who you're trying to become—the gap where most of human life actually happens. The gap where your most consequential decisions get made, where your most important relationships form, and where you construct the frameworks through which you understand yourself.
The VUL is designed to operate in that gap. Not by tracking behavior, but by reading orientation—your present state, expressed through color and intensity, interpreted not as a data point but as a field of need. The system responds not by predicting what you'll do next, but by finding what might restore your balance. That is a different kind of intelligence. And it requires a different kind of language.
The framework for that language already exists. It exists in color, sacred geometry, and the patterns nature has always used to organize itself. We didn't invent these things. We're just remembering them. The next step is building technology humble enough to carry what we find inside of those memories.