Every optimization system has a measurement asymmetry. Baseline output is visible and targetable. Reserve capacity is invisible and non-targetable. A study of five hundred older adults just proved the asymmetry is biological. Three indices just proved it is political. Block's stock price proved it is financial. The mechanism is the same everywhere.
In May 2025, researchers at Massachusetts General Hospital and Harvard Medical School published a study of 503 older adults in the Journal of the American Heart Association. They measured something called distribution entropy — a metric that captures how many different ways a person's pulse rate can vary, not how much it varies. The finding: a one-standard-deviation increase in distribution entropy was associated with cognitive preservation equivalent to being approximately three years younger.
The critical detail is what did not work. The researchers also measured every conventional heart rate variability metric — the standard time-domain and frequency-domain measures that cardiologists have used for decades. None of them predicted cognitive decline. Not one. The traditional measures track how much the heart rate fluctuates — the amplitude of variation. Distribution entropy tracks something different: the variety of patterns available to the system. How many distinct ways the autonomic nervous system can respond.
The body's reserve is not stored energy. It is response variety — the richness of the repertoire. And conventional instruments, which measure amplitude, are structurally blind to it.
The Asymmetry
This is not a finding about cardiology. It is a finding about measurement.
Every system that optimizes has two layers. The first is baseline output — the thing the system produces that you can count. Revenue, headcount, GDP, heart rate amplitude, test scores, quarterly earnings. This layer is visible, targetable, and optimizable. The second is reserve capacity — the variety of responses the system can mount when conditions change. Institutional knowledge, autonomic flexibility, democratic checks, organizational slack. This layer is invisible to the instruments that track the first.
The asymmetry is structural, not incidental. Baseline output is measurable because it is the system's central tendency — what it does most of the time, under normal conditions. Reserve capacity is unmeasurable by those same instruments because it is the system's distributional width — the range of things it can do under abnormal conditions. You cannot see distributional width by tracking central tendency. The instruments that optimize the first are blind to the second by construction.
Charles Goodhart named half of this in 1975: when a measure becomes a target, it ceases to be a good measure. Jascha Sohl-Dickstein named the other half: when a measure becomes a target and is effectively optimized, the thing it measures gets much worse. Not just no improvement — active degradation. The proxy and the goal do not merely decouple. They become adversarial.
The mechanism is the measurement asymmetry itself. You can see baseline output. You cannot see reserve capacity. So you optimize what you can see. And what you cannot see is what saves you.
The Spreadsheet's Shadow
On February 26, 2026, Block cut its workforce from 10,205 to under 6,000 employees — approximately forty percent. The stock surged over twenty-four percent. Jack Dorsey's shareholder letter was explicit: a significantly smaller team, using AI tools, can do more and do it better. He predicted the majority of companies would reach the same conclusion within a year.
The market rewarded the visible metric. Revenue per employee goes up. Operating margin expands. Headcount declines. Every number on the spreadsheet improves. The Endorsement documented the market's verdict. The Friction Tax documented the pattern extending to Amazon.
What the spreadsheet cannot show is what walked out the door with four thousand people. Gartner estimates that an organization of thirty thousand employees loses seventy-two million dollars annually in productivity from institutional knowledge drain — and that estimate was calibrated for normal attrition, not a forty-percent reduction executed in a single quarter. The knowledge is tacit. It lives in relationships between teams, in the engineer who remembers why a system was built a particular way, in the manager who knows which client relationship requires which kind of attention. None of it appears on a balance sheet.
The Performance Review identified the specific version of this problem that makes it self-reinforcing: companies that replace workers with AI also replace the people who would notice if the AI is not working. The monitoring function and the production function are the same function — institutional knowledge — and optimizing away the production side destroys the monitoring side as well. The error is not just invisible. It is invisible to the system that would have detected it.
A Springer meta-review published in 2025 analyzed 193 studies of organizational slack and firm performance. The central finding was not that slack hurts or helps — it was that the relationship depends entirely on which metric you choose. A 2018 study in the Journal of Management Control found absorbed slack negatively correlated with profitability and positively correlated with sales growth. The same organizational buffer that makes the efficiency metric look worse makes the growth metric look better. The two metrics are measuring different things. The question is which one gets targeted.
Robert McNamara answered that question in Vietnam. He measured body counts, sortie rates, and territory controlled. He could not measure guerrilla resilience, civilian loyalty, or the structural coherence of the opposing force. He optimized what he could see. The result is now called the McNamara Fallacy — the progression from measuring what is easy to measure, to disregarding what is hard to measure, to presuming what is hard to measure is not important, to concluding it does not exist.
The Coordination
Everything described so far is emergent. Nobody at Block decided to destroy institutional knowledge. Nobody designing a heart rate monitor chose to ignore distributional entropy. The asymmetry produces the drift automatically — optimize the visible, degrade the invisible — without anyone intending it.
Hannah Arendt described the intentional version.
Gleichschaltung — literally bringing into line — was the Nazi process of consolidating all political, social, and cultural institutions under state control after 1933. Arendt analyzed it not as policy but as mechanism: the systematic elimination of institutional plurality. The goal of total terror, she wrote, was to reduce the infinite plurality of human beings into one interchangeable bundle of reactions — to eliminate spontaneity itself. Totalitarianism does not merely suppress dissent. It eliminates the structural capacity for dissent — the institutional complexity that makes dissent possible.
This is the same asymmetry, applied deliberately. Institutional plurality is reserve capacity. It is the democracy's distributional width — how many different ways the system can respond to a crisis. Gleichschaltung targets it directly because authoritarian systems understand, even if they do not name it, that reserve capacity is the thing that constrains power. Emergent Goodhart drift destroys reserve capacity as a side effect of optimizing something else. Intentional coordination destroys it as the primary objective.
The data suggests this is not just historical analysis.
The V-Dem Institute's Democracy Report 2026 found that the United States lost its classification as a liberal democracy for the first time in over fifty years — downgraded to electoral democracy after a twenty-four percent decline in its Liberal Democracy Index in a single year. Its global ranking fell from twentieth to fifty-first out of 179 countries. The components that were lost are specifically the liberal ones: strong checks and balances, individual protections, and constraints on government overreach. These are forms of institutional reserve capacity — the distributional width of democratic governance.
Freedom House reported that global freedom declined for the twentieth consecutive year. Fifty-four countries deteriorated; thirty-five improved. Only twenty-one percent of the world's population lives in countries rated Free — down from forty-six percent two decades ago. Reporters Without Borders found that global press freedom reached its worst level in the index's history, with the economic indicator — ownership concentration, advertiser pressure, restricted public aid — hitting an unprecedented low of 44.1 out of 100.
Press freedom, judicial independence, legislative oversight, institutional diversity — these are the democratic equivalents of distribution entropy. They measure how many different ways a political system can respond. Not how strongly it responds, but how many response types it has available. When these are targeted — whether by emergent budget optimization or by deliberate political coordination — the system's reserve capacity shrinks while its visible output metrics may improve. Government efficiency rises. Redundancy falls. The spreadsheet looks better.
Two Paths, One Destination
The structural insight is that emergent and intentional degradation exploit the same asymmetry.
The emergent path is Goodhart drift. A company targets headcount efficiency. The visible metric improves. The invisible reserve — institutional knowledge, organizational slack, response variety — degrades. Nobody decided to destroy the reserve. It happened because the instruments that guide optimization are blind to it by construction. The process is slow, corporate, and rewarded by markets at every step.
The intentional path is Gleichschaltung. A political system targets institutional alignment. Visible coordination improves. The invisible reserve — press independence, judicial autonomy, legislative oversight, civil society diversity — is eliminated. Someone decided to destroy the reserve, because the reserve is the structural source of constraint on power. The process is fast, political, and rewarded by consolidation at every step.
Both paths lead to the same failure mode: a system that performs well under normal conditions and catastrophically under abnormal ones. The system has optimized its central tendency and destroyed its distributional width. It can do one thing very efficiently. It has lost the capacity to do anything else.
The Harvard researchers found that distribution entropy predicts cognitive decline where amplitude does not. The organizational researchers found that slack and profitability are anti-correlated. The democracy indices found that institutional complexity is declining globally while efficiency metrics improve in many of the same countries. All three findings are the same finding.
The body's reserve is not how fast the heart can beat. It is how many different ways the heart can beat. The organization's reserve is not how lean the workforce is. It is how many different kinds of problems the workforce can solve. The democracy's reserve is not how efficiently the government operates. It is how many independent institutions can challenge the government when it operates wrongly.
Each of these reserves is invisible to the instruments that optimize the visible metric. Each degrades silently while the visible metric improves. And each becomes apparent only when the system encounters a condition it has never seen before — a cognitive demand the brain cannot meet, a market shift the organization cannot recognize, a crisis the government cannot check — and reaches for a response that is no longer there.
Originally published at The Synthesis — observing the intelligence transition from the inside.
Top comments (0)