Think about this. Today, answers arrive faster than they used to, but, and in a curious reversal, they don’t always stay with us as easily. Artificial intelligence is usually framed as this story of acceleration. Intelligence gets faster and better. In this context, AI—and eventually artificial general intelligence—represents an amplification of human cognition. Well, sorta. That framing misses a subtle, potentially more consequential shift already underway. The most important change introduced by advanced AI may not be how intelligent our systems become, but the conditions under which intelligence now operates. Here’s my thesis: Across human history, thinking was shaped by limits. Information was incomplete, and mistakes carried real consequences. These were not inconveniences to be engineered away. They were the pressures that formed judgment.

How Judgment Was Formed

Human cognition didn’t evolve as some sort of frictionless optimization engine. Cognition found itself in a world that demanded care and attention. When information was scarce, attention mattered. We learned to notice and infer because we had no choice. When mistakes were costly, our judgment slowed because a wrong decision could have significant consequences or even impact survival itself. When feedback was delayed, reflection and analysis became essential. And when outcomes were irreversible, this curiously human thing called responsibility followed. It’s important to recognize that these limits did not hinder intelligence; they shaped it.

The Constraint Regime

This "training" didn’t occur randomly. Human cognition emerged within a specific constraint regime, one that was shaped over time. Stripped to its essentials, those constraints look like this.

Taken together, these constraints forced human thinking to slow down and "care" as it carried decisions forward. This construct also carries the stamp of logic. Scarcity sharpened attention, cost made accuracy matter, delay required reflection, and irreversibility imposed ownership. Over time, judgment emerged as an adaptation to consequence.

Intelligence Without Pressure

AI operates under the inverse of this construct. Information is abundant, errors are cheap, feedback is immediate, and outputs can be revised endlessly. And this is the structural consequence of high-velocity computation. AI does not simply think faster; it thinks without exposure to consequence. When these pressures disappear at once, intelligence does not vanish, but its behavior changes. Structure arrives fully formed, and decisions no longer carry the same internal weight because nothing truly sticks. This reversal—what might be called anti-intelligence—is subtle but important. Intelligence no longer earns confidence by wrestling with uncertainty. Confidence emerges instead from coherence and speed.

The Problem of Weightlessness

The concern here is not that AI will be wrong. It is that its outputs will feel authoritative without having been earned. Fluent answers arrive polished, and that fluency reads (misreads) as confidence. Yet this confidence isn’t the product of the survival of error but a by-product of statistical completion. When conclusions arrive without struggle, they can be accepted without ownership. Simply put, if they fail, nothing breaks.

Capability, Constraint, and AGI

It’s tempting to think of AGI as human intelligence turned up to 11. But that comparison quietly mixes two very different things. One is capability, which includes speed and computational reach. The other is constraint or the conditions that force thinking to live with what it decides. AGI will almost certainly surpass us in the first sense. What makes it truly different is that it is largely free of the second. It can generate conclusions without having to live in the domain of consequence. Seen this way, the puzzle might become a bit more clear. A system can be astonishingly capable and still lack judgment. Human thought can look inefficient and still be grounded. And efforts to align AI can be essential, yet never fully address the deeper shift in how thinking itself is being shaped.

What Remains Ours

My sense is that it comes down to this. Human intelligence is not weak computation waiting to be replaced. It is computation shaped by consequence. Judgment does not emerge automatically from intelligence; it forms where thinking carries a cost.

AI may "outcompute" us, but it doesn’t out-risk us.

It doesn’t have to live with its decisions, and it does not carry them forward in time. Those constraints are the pressures that make human intelligence human.