There is a particular kind of science-fiction story that stays useful because it exaggerates something real.
In the 2011 film Limitless, a struggling writer gains access to a fictional cognitive enhancer that radically amplifies memory, focus, confidence, pattern recognition, and verbal fluency. The premise is simple enough to feel like fantasy: take the pill, become brilliant.
But the film’s more enduring insight is not the fantasy of enhanced intelligence.
It is the dependency that follows.
A person who appears suddenly more capable also becomes tethered to the external source of that capability. The enhancement creates power, but it also creates leverage. Intelligence becomes something supplied, rationed, interrupted, negotiated, and potentially controlled.
We are not discovering NZT. We are building intelligence-as-a-service.
And the question is no longer only whether external intelligence can make us more capable.
It is whether we can become more capable without becoming dependent on the conditions of our enhancement.
The transformation in Limitless is seductive because it dramatizes a wish almost everyone understands.
The world, once blurry and resistant, becomes legible. Patterns sharpen. Memory unlocks. Language arrives on command. Confidence returns before evidence has caught up. The self, once fragmented and hesitant, becomes fluent, composed, persuasive, and dangerous.
The fantasy is obvious: a human being becomes more intelligent almost instantly.
But the more interesting part of the film is not the increase in performance. It is the change in the relationship between capability and supply.
The drug does not merely enhance the protagonist. It relocates the source of his enhancement. Intelligence becomes externally mediated. Confidence becomes contingent. Fluency becomes conditional.
A person who appears more powerful than before is, in a deeper sense, more vulnerable than before — because the source of that power is no longer fully internal.
Artificial intelligence is beginning to play, at societal scale, the role that the fictional pill plays in the film.
It does not need to enter the bloodstream.
It enters workflows, inboxes, dashboards, copilots, search boxes, design tools, coding environments, strategy documents, customer service scripts, performance reviews, and decision loops.
The result may be spectacular. People can write faster. Teams can synthesize more. Managers can see options sooner. Analysts can produce better drafts. Engineers can explore alternatives. Leaders can simulate arguments. Organizations can accelerate work that previously required scarce expertise.
This is real augmentation.
But every intelligence supplement carries a second question: what happens when the user becomes powerful through a capability they do not fully govern?
In the film, dependency is not merely personal weakness.
It becomes leverage.
Other actors begin to understand that the enhanced individual has a supply chain. The person who appears unusually capable is also unusually controllable. To influence the supply is to influence the person. To threaten access is to threaten the self that access has created.
This is where the analogy becomes uncomfortable.
An organization dependent on externally supplied intelligence may look more capable, not less.
It may produce more analysis, faster decisions, better summaries, cleaner code, more persuasive documents, and more confident strategies. It may look modern, responsive, and intelligent.
But if its own capacity to frame problems, challenge assumptions, verify outputs, preserve accountability, and reason under uncertainty has weakened, then the organization has not simply become AI-enabled.
It has become cognitively leveraged.
Dependency on external intelligence may become as strategically potent as dependency on credit.
Credit extends present capacity by borrowing against future income. Intelligence-as-a-service extends present cognition by borrowing from external reasoning infrastructure.
Both can accelerate growth. Both can create fragility. Both can become instruments of control when the borrower can no longer function without continued access.
The danger is not that AI will make people stupid. That is too crude.
The danger is that AI will make people productive in ways that conceal the erosion of the very capabilities that make productivity trustworthy.
This is cognitive offloading.
It does not arrive as laziness. It arrives as convenience. It arrives as efficiency. It arrives as the reasonable decision to let the machine draft the answer, summarize the meeting, propose the strategy, identify the risks, write the code, compare the options, and explain the decision.
At first, this feels like liberation. Then it becomes habit. Then it becomes dependency. Then, eventually, it becomes difficult to know whether the human is using intelligence — or merely operating near it.
Dependence itself is not the enemy.
Civilization is built on productive dependence. We depend on language to think beyond the limits of memory, on books to converse with the dead, on institutions to stabilize trust, on instruments to extend perception, and on markets to coordinate effort at scales no individual mind can grasp.
The question is not whether dependence can be avoided.
It cannot.
The question is whether a dependency expands human agency or quietly narrows it.
A good dependency leaves the user more capable over time. A dangerous dependency leaves the user more productive in the moment, but less capable without it.
This is the distinction that enterprise AI now forces into the open.
The problem is not that we will use artificial intelligence too much.
The problem is that we may use it in ways that make our own intelligence less necessary to cultivate.
This is why AI Fluency matters.
Not as tool familiarity. Not as prompt technique. Not as enthusiasm for new systems.
AI Fluency is the discipline of using supplied intelligence without surrendering internal agency.
At Coincentives Labs, we describe this discipline through four recurring functions: Communicate, Co-Create, Challenge, and Curate. But they can also be understood less formally as four acts of cognitive self-preservation.
The human defines the problem, context, constraints, priorities, boundaries, and completion criteria. This prevents AI from becoming the hidden author of the direction.
The human uses AI to generate alternatives, synthesize complexity, and refine substance, but remains responsible for shaping the work. This prevents collaboration from becoming disguised delegation.
The human tests assumptions, identifies uncertainty, checks reasoning, corrects errors, and resists plausible fluency. This prevents confidence from becoming dependence.
The human turns useful outputs into reusable artifacts, decision rules, workflows, memory, and judgment. This is where external assistance becomes internal growth.
Together, these four functions offer a path that Limitless leaves unresolved.
How does one become more capable without remaining dependent on the supplement?
The answer is not abstinence.
The answer is governed use.
The objective is not to reject artificial intelligence, just as the objective of literacy was never to reject books, or the objective of calculation was never to reject instruments.
The objective is to use external cognitive tools in ways that deepen, rather than displace, human capability.
This is the tapering mechanism.
A person tapers off dependency not by using AI less, but by using AI differently.
A team becomes less dependent not by refusing AI, but by ensuring that every AI interaction strengthens problem framing, judgment, verification, accountability, and durable knowledge.
An organization becomes resilient not by avoiding intelligence-as-a-service, but by preventing it from hollowing out the intelligence of the organization itself.
That distinction may define the next phase of enterprise AI.
The first phase was access. The second phase is adoption. The third must be governance of dependency.
The question for leaders is no longer simply: how much AI are we using?
It is: what is AI doing to our capacity to think without it?
A high-performing organization may discover, too late, that it has become brilliant only under subscription.
It may find that its apparent intelligence depends on a vendor relationship, a platform policy, a pricing model, a geopolitical permission structure, or a compute supply chain it does not control.
That is not fluency.
That is exposure.
The deeper promise of AI is not that it gives us a synthetic mind to lean on.
It is that, used well, AI can become a mirror, a sparring partner, a scaffold, and a forcing function for better human thought.
But only if we design for it.
Only if incentives reward reasoning, not just output. Only if workflows preserve decision ownership. Only if teams measure verification, challenge, and traceability.
Only if leaders understand that intelligence supplementation without agency preservation is not transformation.
It is dependency with better lighting.
This is why Agentic Readiness matters.
Agentic Readiness is the capability of an individual or organization to work with increasing AI autonomy while preserving human agency, judgment, accountability, and traceability.
It is not achieved by deploying agents.
It emerges when people, workflows, incentives, architectures, and governance mechanisms are prepared to govern the consequences of autonomy.
In the context of intelligence-as-a-service, Agentic Readiness is not a fashionable maturity label.
It is a resilience condition.
It asks whether humans remain capable when AI becomes more capable.
It asks whether the organization can preserve judgment while accelerating output.
It asks whether the use of external intelligence compounds internal capability or quietly replaces the need to cultivate it.
This is also why measurement matters.
Dependency rarely announces itself.
It hides inside performance.
A person who can no longer reason without AI may still produce impressive documents. A team that has stopped challenging assumptions may still move faster. A function that has lost decision discipline may still report productivity gains.
An organization whose internal judgment is weakening may still look, on the surface, more intelligent than before.
This is the adoption illusion.
Usage rises. Output improves. Confidence increases.
But beneath the surface, the organization may be losing the ability to explain, verify, contest, and own its own decisions.
AI Fluency measurement exists to make this hidden shift visible.
It asks not only whether AI is being used, but how intelligence is being governed in use.
These are not soft questions.
They are strategic questions.
Because the organization that cannot observe the quality of human–AI collaboration cannot govern the dependency it is creating.
The lesson of Limitless is not that enhanced intelligence is dangerous.
The lesson is that borrowed intelligence becomes dangerous when the borrower loses the capacity to govern it.
That is the problem AI now places before every individual, organization, and society.
Not whether we can become more capable with machines.
We can.
The question is whether we can become more capable without becoming owned by the conditions of our enhancement.
The future will not divide neatly between those who use AI and those who do not.
It will divide between those who use AI to compound human capability, and those who use AI in ways that quietly make human capability optional.
The first group will become fluent.
The second will become dependent.
And in a world where intelligence itself becomes infrastructure, capability, and strategic dependency, that difference may matter more than any productivity gain we can measure today.
The pill in Limitless was fiction. The supplied self is not.
The broader strategic-dependency argument is explored in:
CL-J26-001-FE-002 — The Cost of Intelligence-as-a-Service in a Fractured WorldA field essay examining how dependence on externally supplied intelligence can become a strategic vulnerability for individuals, organizations, and nations.
We measure AI fluency as governed collaboration — and turn it into evidence (and optional proof-of-skill) that holds up under optimization.
AI Fluency Score is a key leading indicator of Agentic Readiness