There is a strange pattern in how we build things now. We push a metric past the point where humans can use it, then keep pushing.
A human retina, under good light, resolves about 60 pixels per degree of vision. At typical viewing distances, 4K on a phone is already past that threshold. 8K is a number we can measure but not see. The televisions exist; the cameras exist; the content pipeline exists. We just don’t—can’t—perceive the difference. The customer of 8K is not the eye. It is the spec sheet, the benchmarking channel, the upgrade cycle, the compression algorithm.
This is not an isolated case. It is the shape of a pattern.
The thresholds we’ve already crossed
Audio. Human hearing tops out around 20 kHz on a good day and declines with age. Hi-res audio formats sample at 192 kHz. In controlled blind tests, most listeners cannot reliably distinguish lossless audio from a well-encoded 256 kbps stream. We built the infrastructure anyway.
Refresh rates. The human visual system integrates motion somewhere between 60 and 90 Hz for most tasks. Gaming monitors now advertise 540 Hz. The incremental buyer is not an eye; it is a review channel.
High-frequency trading. When the NYSE trading floor was loud, humans were participants. Today, most trades are executed by algorithms in microseconds, co-located in data centers beside the exchange to shave nanoseconds off signal travel. Retail investors still benefit from price discovery, but the discovery itself has moved to a medium they cannot enter.
Chess and Go. Stockfish plays chess at roughly 3600 Elo. Magnus Carlsen, arguably the best human ever, peaked at 2882. The gap is not small; it is the difference between a grandmaster and a club player. AlphaGo’s famous Move 37 against Lee Sedol was described by top commentators as either a mistake or a move no human would play. It was neither. It was a move from a medium we don’t inhabit.
The thermal wall. Moore’s Law did not end because we lost interest. It ended because silicon cannot shed heat fast enough at higher clock speeds. Here, physics imposed the ceiling for us. In most other domains, we have no such mercy.
Model outputs. A modern frontier model produces more text, code, or analysis in an hour than a domain expert can audit in a week. “Human-in-the-loop” is the official posture. In practice, the loop samples a thinner and thinner slice of what the system does.
The pattern
In each case, we optimized a machine-facing metric—resolution, latency, parameter count, clock speed—long past the point where the metric mapped to a human experience. The justification for each increment was always technical. The user, somewhere along the way, stopped being the audience. The audience became the next machine in the chain.
We still benefit from these systems. Portfolios grow on HFT-discovered prices. Phones render smoothly at frame rates we can’t perceive. Models produce outputs we skim and accept. But benefit is not participation, and that distinction matters more than we’ve admitted.
The sunk cost
Why don’t we stop? Because we have built industries, careers, and capital structures around the machine-facing metric. Fabs are tooled for the next node. Display supply chains assume resolution growth. Model labs are staffed to scale. To pause and ask “who is this for?” would strand enormous investment that only makes sense if the metric keeps climbing.
This is the real sunk cost fallacy of our era. Not that we keep buying the thing, but that we keep building for a customer who has already left the room.
What we lose
Here is the part the efficiency argument misses.
A human who plays chess against a 3600 Elo engine is not playing chess; they are being graded. A listener served algorithmically-selected music is not developing a taste; they are being profiled. A trader whose orders are routed through machine-learned execution is not trading; they are a signal. A writer whose sentences are drafted by a model is not writing; they are editing.
In each handoff, we gained a benefit and lost a participation. And it turns out participation is where most of the meaning lived. The joy of playing exceeds the joy of winning optimally. The joy of cooking exceeds the joy of a perfect meal arriving. The joy of a walk exceeds the joy of a faster route. When we optimize for outcomes, we tend to discover, too late, that we were never actually in it for the outcomes.
In every one of these handoffs, we gain an undeniable benefit—a perfectly optimized outcome—but we lose participation. And participation is the core of the human experience. The joy of playing vastly exceeds the joy of a guaranteed win. The satisfaction of cooking exceeds the convenience of a perfect meal arriving at your door. The vitality of a walk exceeds the utility of being instantly teleported to a destination.
A different metric
The challenge before us is not to halt technological progress, but to violently realign its trajectory. We must begin to measure what we have spent the last two decades ignoring. We need metrics that penalize the machine for outrunning the human.
- For Displays: The standard should be perceptual fidelity per watt, rewarding efficiency and sustainability over imperceptible pixel density.
- For Artificial Intelligence: The metric must be auditability—the fraction of a model’s outputs that a competent human can rigorously verify in the time practically available. A model that produces faster than a human can comprehend should be viewed as a liability, not a breakthrough.
- For Markets: We should optimize for discovery quality per unit of human comprehension, rather than transaction volume per microsecond.
- For Entertainment and Media: The goal must be intentional time engaged rather than passive time consumed.
None of these metrics are scientifically impossible. They are merely unprofitable to the current supply chain. And that is precisely why they are the only metrics worth naming.
The choice of our generation is not between progress and stagnation. It is a choice of allegiance. Will we continue to serve an automated economy that has left us behind, or will we remember who the technology was supposed to be for in the first place?
Reinvent The Self! 
Leave a comment