Preserving Human Intelligence

September 2026

As artificial intelligence becomes more capable, preserving and developing human intelligence becomes more important—not less.

This may initially seem backward. If machines can retrieve information, summarize arguments, generate language, and solve increasingly complex problems, then human beings should be freed from much of the intellectual labor those tasks once required.

And they will be.

But intelligence is not only the ability to produce correct answers.

It is also the ability to form judgments, sustain contradictions, notice assumptions, revise convictions, develop intuitions, and arrive at thoughts that are genuinely one’s own. These capacities are not simply stored within us. They develop through exercise.

AI can inadvertently weaken that exercise precisely by being so useful.

Every unresolved question can now receive an immediate answer. Every ambiguity can be clarified, every unfinished sentence completed, every difficult text summarized. The interval between encountering a problem and receiving a response is becoming shorter.

That interval, however, was not always empty.

It was where thought happened.

A person struggling to articulate something may discover that the struggle changes what they are trying to say. Someone forced to remain with conflicting demands may develop a judgment that neither demand could have produced alone. An intuition that cannot immediately explain itself may become more discriminating when it is questioned without being dismissed.

These are not inefficiencies surrounding thought. They are some of the processes through which thought acquires depth.

When assistance becomes substitution

The danger is not that AI will make people unintelligent. Nor is the answer to reject tools that make intellectual work faster and easier.

The more subtle danger is that assistance can become substitution.

A person may continue producing sophisticated work while exercising less of the judgment that once made that work meaningful. The output remains intelligent, but the intelligence increasingly belongs to the system that generated its structure, language, and conclusions.

This creates a new problem. We may become extraordinarily productive while becoming less practiced at determining what deserves to be produced, what should be believed, and what genuinely belongs to us.

The distinction matters because judgment cannot be reduced to information processing.

Judgment determines which facts are relevant, which distinctions matter, when a rule applies, when an exception is warranted, and what must remain unresolved. It is formed not only by encountering answers, but by undergoing the difficulty of arriving at them.

If AI removes every form of intellectual friction, it may also remove the conditions under which judgment develops.

The value of productive friction

Not all friction is valuable.

Confusing interfaces, unnecessary steps, inaccessible language, and administrative obstacles do not deepen thought. They merely exhaust attention. Technology should eliminate that kind of friction wherever possible.

But there is another kind: the friction produced when reality does not fit our existing understanding.

A conviction encounters evidence it cannot easily absorb. Two legitimate obligations cannot both be satisfied. An intuition resists explanation but refuses to disappear. A familiar description of oneself begins to feel inadequate.

This is productive friction.

It interrupts automatic judgment and makes the structure of thought available for examination. It creates the possibility that a person will not merely defend or abandon an existing position, but develop a more adequate one.

Depth emerges through this contact.

Depth is not verbosity, complexity, or time spent thinking. It is the extent to which thought has encountered what resists it. A judgment becomes deeper when it has recognized its own limits, admitted what it previously excluded, and changed without losing everything it was trying to preserve.

Productive friction produces depth. Depth strengthens human intelligence.

AI that preserves the exercise of thought

The question is therefore not simply whether AI should assist human thinking. It is what kind of thinking the assistance leaves us capable of doing.

Most AI products are designed to reduce the distance between intention and output. The user asks for something, and the system supplies it. This can be extraordinarily valuable. But it is not the only relationship humans can have with intelligent systems.

AI can also be used to construct encounters that return people to their own thinking.

Instead of answering a difficult question immediately, a system can help expose the judgments already operating within it. Instead of resolving a contradiction, it can give the contradiction enough structure to become intelligible. Instead of replacing an unfinished thought, it can help the person perceive what that thought is trying to become.

This is the category Negativa occupies.

A Negativa Encounter does not ask AI to tell someone what to believe. It places opposing judgments into a structure and asks the participant to remain in contact with both. Their language, distinctions, emotional responses, and symbolic choices become evidence of how their judgment is currently organized.

The resulting Reading is not intended to end the thinking. It gives provisional form to what the participant may have been thinking without yet being able to say.

The technology performs interpretation, but the participant remains responsible for recognition, revision, and judgment. The aim is not to outsource thought. It is to make thought more available to the person whose thought it is.

This is AI used to preserve human intelligence rather than replace its exercise.

Intelligence worth preserving

Human intelligence has never consisted solely in outperforming machines at tasks.

Its distinctiveness lies partly in our capacity to be transformed by what we encounter: to discover that our categories are inadequate, to experience competing claims as legitimate, and to create forms of judgment that did not exist before the contradiction made them necessary.

These capacities will not become obsolete as AI improves.

They will become more consequential.

A world filled with generated answers will need people capable of determining which questions matter. A world of frictionless production will need people who can recognize when friction should not be removed. A world of increasingly persuasive language will need people capable of distinguishing fluency from truth and borrowed articulation from genuine conviction.

The purpose of philosophical technology is not to preserve an earlier world in which thinking was slower because tools were weaker.

It is to ensure that more powerful tools do not leave us with less powerful minds.

AI should not only expand what human beings can produce.

It should help preserve the intelligence through which anything we produce becomes meaningfully our own.