AI Agents and Human Constraints
We're building AI agents that can move freely across digital platforms while keeping humans locked within them. We marvel at machines exercising freedoms we've never granted ourselves.
The digital world is built on lock-in. From social platforms to streaming services, e-commerce to financial apps, the internet’s business model enables choice within ecosystems while constraining movement between them. We’ve accepted these limitations as the necessary cost of digital progress, of the freedom to use an ever-expanding array of digital products and services.
Now, AI agents arrive with capabilities we’ve designed out of our own experience: the freedom to move seamlessly across digital boundaries, to act with autonomy across platforms, to escape the constraints of single ecosystems. These agents can traverse the digital landscape, executing actions without supervision, unbound by the walls that confine human users. We marvel at machines exercising freedoms we’ve never granted ourselves.
This paradox underscores a contradiction in how we’ve built the digital world.
For users, boundaries are intentional and interoperability is achieved only through hacks or mandates. For agents, interoperability exists from day one.
The emerging “Agentic Era” grants machines a freedom to navigate digital spaces that we have systematically withheld from people.
The irony is striking: we’re building AI agents to overcome the very constraints we’ve designed into our digital systems.
These were not technical constraints. They were business imperatives that have shaped the internet from its earliest days until now, defining not just products but our very conception of what it means to be human in digital spaces.
Below, I ask why we marvel at the freedoms we’re granting to AI, while accepting as inevitable the limits we’ve long placed on ourselves. I trace how we arrived at a model built on user lock-in, the cultural unease that’s accompanied it from the beginning, alternatives that challenge its assumptions, and the deeper choice now confronting us in the age of agentic AI.
User Stories
Online, we are everywhere users. Searching, shopping, sharing, creating, listening, watching—users. In digital product design, the word most commonly paired with user is experience. User experience teams are the “champions of the user,” uncovering needs and ensuring products meet them.
But another word takes over when design gives way to business: lock-in. Digital businesses depend on keeping activity within their boundaries to maximize acquisition, retention, and monetization. Lock-in is what prevents a social graph, an ebook, a playlist, or a reputation from moving easily between ecosystems. It defines the space in which user experience teams operate.
Lock-in is how digital businesses scale. If your definition of the user doesn’t support that model, you don’t have a business.
Pure-play digital firms perfected lock-in. Legacy firms adopted it in the name of digital transformation. Together, they brought the tension between user experience and user lock-in to nearly every industry.
No sector has yet developed an alternative at scale. The result is an internet that treats people as inherently bounded, not by need or intention, but by business perimeter.
Efforts to change this—from open social graphs to data portability—remain at the margins. Without lock-in, there is no business model. And once you have lock-in, you have users: people with constrained agency by default.
Our most widely used technology products deliver real value through this model. But in doing so, they reinforce a deeper logic: one that treats constraint as a condition of growth, and people as pathways to scale.
That default stands in stark contrast to what we’re now building for AI agents: systems designed, from the start, to move freely across the very boundaries we’ve constructed to contain ourselves.
Red Pills and Black Mirrors
Popular culture has long captured the unease that comes with treating people as locked-in users.
In The Matrix, the red pill awakens people from a simulated reality controlled by machines. Once awake, they can navigate that system, but they remain bound by its rules. Only agents—autonomous code—can bend or rewrite those rules.
Until Neo. Neo moves like an agent, fights like one, bends the world as they do. To be human and free in The Matrix is to become more like code.
Two decades later, Black Mirror takes the logic further.
In Common People, a husband signs his comatose wife up for an experimental brain procedure, tethering her consciousness to a cloud service to preserve function. As the company’s business model evolves, their lives degrade, locked into a decaying digital service they can’t opt out of. Recovery of agency requires radical exit.
Between The Matrix and Black Mirror lies a quarter century of digital growth and a growing sense of decay. Cory Doctorow calls it the “enshittification” of the internet: a system that grows in capability even as it erodes in experience. Beneath that erosion is the internet’s foundational tension between user experience and user lock-in.
Neither The Matrix nor Black Mirror offers an alternative vision. That’s the point. They dramatize the cost of escape when systems built on constraint leave no room to imagine another way. Once we accept the premise of people as users, we’ve already limited what kind of digital world is possible.
To escape that frame, we need new building blocks. A different world—and a different word.
Alt.User
If pop culture captures our unease with being locked in, infrastructure reveals the work of trying to break out. Not everyone accepts the user as the final word. Movements like crypto and digital public infrastructure have begun from a different premise: that people are not just users of digital systems, but participants with agency. Crypto imagines them as sovereign individuals. DPI treats them as citizens.
Crypto replaces user with sovereign individual, and lock-in with decentralization. In its vision of a scaled, decentralized financial system, people can own, store, exchange, and monetize digital assets without relying on intermediaries.
In crypto, agency is ownership. The sovereign individual holds full custody of assets, identity, activity, and access. The model requires both new infrastructure and a shift in self-perception—from user to custodian of digital presence.
Digital public infrastructure begins elsewhere. Instead of sovereign individuals, it centers citizens. Instead of decentralization, it builds public goods. Through verified identity, shared standards, and data portability, DPI offers the building blocks for access, transparency, and agency at scale.
In its most expansive form, DPI envisions people moving through digital life as rights-bearing participants—able to carry their data, identity, and reputation across services as easily as a passport crosses borders.
These alternatives rethink how digital systems scale. Crypto seeks it through decentralized coordination. DPI through state-backed standards and public infrastructure. Both move beyond the user as the default mode of participation.
They remind us that what’s needed is reinvention—a new blueprint for digital life.
The Time Has Come to Make a Choice, Mr. Anderson
The user has had a good run. As the foundation of the internet’s business model, it has helped build an industry that touches nearly every part of life. It can retire comfortably, with a view from the beach of the digital realm it helped construct.
What would take its place? What would technology not centered on users and lock-in look like?
It would begin by separating people from products.
Both crypto and DPI offer clues: as different as sovereign individual and citizen are, both demand that technology adapt to the person—not the other way around.
In the way people are using AI even today, you can see the contours of expanded agency. People move fluidly between different LLMs, improvising and orchestrating workflows across rather than within providers. To follow that logic would mean replacing the user with new roles: explorer, improviser, orchestrator, artist.
But these roles still map people to technology. We improvise not because the systems are seamless, but because we’ve learned to move across their gaps. This framing narrows our imagination to AI, when what we need to rethink is the digital foundation on which AI now rests.
Because the reality is that so much of what we are asking AI to orchestrate for us is a way around the constraints and lock-in of today’s internet. We are building code with agency to cut across the walls that keep people in.
We face a choice.
We can continue to build AI inside the old frame: products that treat people as users, businesses that depend on lock-in. Or we can take the vast energy now spent on crafting agency for code, and apply it to how we think about people.
If AI truly is existential-level technology—as we’re told daily—we need a robust concept of the person to match its scale. The only word that can withstand that weight is human: one struggling to break free in The Matrix, and locked in the Black Mirror.
To center technology on the human is to resist narrowing us to tidy narratives of tasks completed inside a system. It is to reintroduce human complexity into the design of technology itself.
Human-centered technology would build upward from new first principles: meaning, dignity, worth, selfhood, trust, growth. These are not use cases. They are conditions for human flourishing. Why would we want our technology to deliver anything less?
Maximizing these conditions depends on freedom of movement.
Meaning, worth, and trust cannot be contained in one or another digital ecosystem, or aggregated across them. They are strengthened from the freedom to come and go, to cut across digital domains with ease and fluency, to take what works and leave the rest behind. They depend on agency.
And there are signs of change. Many now feel that something fundamental is missing. The early idealism of user-centered design has deepened into a wider search for meaning, dignity, and complexity. A desire to bring the human back in. The talent is there. What’s missing is the model.
Decades into the internet, most businesses still scale through lock-in. The challenge for the next era—the truly agentic one—is to build by expanding human agency. Not by locking people in, but by letting them move.