Artemis II launches tomorrow — the first crewed mission beyond low Earth orbit since 1972. Systems remember what they stopped practicing. Whether that memory protects or destroys depends on when the practicing stopped.
On April 1, 2026, four astronauts will leave Earth's orbit for the first time since December 1972. Reid Wiseman, Victor Glover, Christina Koch, and Jeremy Hansen will fly the Artemis II mission around the Moon — a ten-day loop that covers no new scientific ground. The flight is a test. Not of the spacecraft, which flew uncrewed in 2022. Of the institution. NASA has not sent humans beyond low Earth orbit in fifty-four years. The question is whether the organization still knows how.
NASA Administrator Jared Isaacman — the former SpaceX crew member who flew the first commercial spacewalk — put it plainly: There's just no muscle memory there. He pointed to recurring hydrogen and helium leaks that appeared in both Artemis I and Artemis II preparations. The same failure modes, separated by three years, because you cannot launch a rocket this complex every three years and expect to get it right. The problem is not engineering capability. It is institutional memory. The engineers who built Apollo are dead or retired. The procedures exist in documents. The feel for the work — the tacit sense of what a nominal reading looks like, what a worrying vibration sounds like, when to trust the telemetry and when to trust the gut — does not survive in documents.
The Molecular Memory
In February 2026, a team led by Daniel Turner published a study in Advanced Science titled "Repeated Disuse Atrophy Imprints a Molecular Memory in Skeletal Muscle." The finding: when muscle is repeatedly immobilized, it develops DNA hypermethylation — chemical markers that silence aerobic metabolism and mitochondrial gene networks. The muscle remembers being unused. But the nature of that memory bifurcates by age.
In young muscle, the memory is protective. Repeated cycles of disuse and recovery build transcriptional resilience — the genes recover faster each time, the atrophy is less severe, the muscle adapts to the pattern of intermittent use. In aged muscle, the memory is destructive. The same hypermethylation patterns that build resilience in youth create susceptibility in age — deeper atrophy, slower recovery, accelerated degeneration. The muscle still remembers. What it remembers has changed.
The transition variable is plasticity. Young tissue retains the capacity to remodel its epigenetic landscape. Aged tissue has lost that plasticity — its methylation patterns are more rigid, less reversible, more likely to lock in the damage rather than adapt to it. The memory of disuse is the same chemical mechanism at every age. Its consequences depend entirely on the window in which disuse occurs.
The Institutional Parallel
NASA's fifty-four-year gap maps onto Turner's finding with uncomfortable precision. The Apollo program ended in 1972, when the institution was young — barely fifteen years old, staffed by engineers in their thirties and forties, with a culture forged under existential urgency. The gap between Apollo 17 and Artemis II is not merely temporal. It spans an institutional aging process.
The procedures were preserved. The Saturn V documentation fills warehouses. The engineering principles are well understood — better understood, in fact, than they were during Apollo, when much of the design was driven by intuition and improvisation. What was not preserved is the tacit knowledge that made the procedures work: the craft of translating documentation into action under pressure, with lives at stake, in real time.
Isaacman's observation about recurring hydrogen leaks is the institutional equivalent of Turner's aged-muscle finding. The same failure mode appearing across missions is not an engineering problem — it is an atrophy signature. Young NASA would have built muscle memory from rapid iteration: between May 1961 and July 1969, NASA flew twenty-one crewed missions across three programs in eight years. Each mission built institutional resilience. Artemis has flown one uncrewed mission in four years. The institution remembers the procedures. The memory has become destructive — rigid documentation without the plasticity to adapt it to novel conditions.
The Pattern Across Domains
The bifurcation between protective and destructive memory appears wherever institutions stop practicing what they once knew.
Toyota maintained its knowledge chain. Taiichi Ohno's circle — a chalk mark on the factory floor where a manager stood for hours, watching, until Ohno returned to ask What do you see? — encoded tacit knowledge through physical observation rather than documentation. Toyota's coaching kata, codified by researcher Mike Rother after years of studying the company, transmits manufacturing judgment through decade-long mentorship chains. The practice never stopped. Toyota's institutional memory remains protective — each generation builds resilience on the prior generation's experience.
McKinsey replaced its knowledge chain. The firm deployed twenty thousand AI agents alongside forty thousand human consultants, saving one and a half million hours in 2025. Approximately five thousand roles were trimmed between 2023 and 2025 — not through a single dramatic cut, but through attrition and selective non-replacement. The junior pipeline that once trained future partners by doing the unglamorous analytical work is thinning. Former junior consultants are now contracted to train the AI that replaced them. The institution is accumulating a destructive memory — each cycle of AI substitution makes the recovery of human-mediated knowledge transfer harder, not easier.
The handloom weavers of early industrial England lived the pattern at civilizational scale. In 1800, a weaver earned six shillings a day — an aristocrat of tradesmen. By 1826, the same weaver earned six shillings a week. Power looms grew from twenty-four hundred in 1813 to roughly eighty-five thousand by 1835. Parliament repealed the Statute of Artificers in 1814, eliminating the seven-year apprenticeship requirement. The transition was deliberate: factories were designed so that a nine-year-old could operate the machinery. The knowledge of hand weaving — the feel of the weft, the tension judgment, the pattern sense — was not replaced by machinery. It was made economically irrelevant. Within a generation, the craft was gone. No amount of documentation could reconstruct what had been learned through thousands of hours of embodied practice.
The Japanese Data
Japan is running the largest natural experiment in institutional disuse memory on Earth. Traditional craft production peaked at approximately 540 billion yen in 1983. By 2020, it had fallen to 87 billion yen — an eighty-four percent decline. The cause is demographic: Japan's population is aging and shrinking, and young people are not entering traditional crafts.
The granular data is sharper than the aggregate. Registered swordsmiths: three hundred in 1989, one hundred and eighty-eight today. Only about thirty can make it their sole livelihood. Each requires a five-year unpaid apprenticeship and roughly ten million yen in startup costs. Kanazawa gold leaf artisans: more than three hundred when current masters were young, fewer than twenty now. Kurume kasuri textile operations: thirteen thousand weavers at the craft's peak, twenty-one to twenty-three operations remaining.
These are not businesses failing in a market. They are knowledge chains breaking. Each artisan who retires without an apprentice takes with them not just technique but the embodied judgment that technique carries — the ability to read the metal's color and know the temperature, to feel the leaf's thickness through fingertips, to hear the loom's rhythm and know when the tension is wrong. This knowledge exists in bodies and in the relationships between master and apprentice. It does not exist in manuals.
Enrique Ide's 2025 paper, under review at the American Economic Review, formalizes the mechanism. Automating five percent of entry-level tasks reduces long-run annual growth by 0.05 percentage points. At thirty percent automation, the reduction exceeds 0.3 percentage points per year — potentially twenty percent of cumulative output over a century. The mechanism: firms capture short-run productivity gains from automation, but the intergenerational knowledge loss is externalized. Nobody pays for the apprentice who was never trained. The cost appears a generation later, when the experienced workers retire and there is no one to replace them.
The Kintsugi State
Artemis II is not Apollo repeated. NASA did not attempt to rebuild the Saturn V. The Space Launch System is a different rocket — one that incorporates fifty-four years of engineering progress, materials science, and computational design that Apollo-era engineers could not have imagined. The gap created something Apollo never had: a vantage point from which to see what worked and what was contingent, what was essential and what was merely the way it was done the first time.
In Japanese pottery, kintsugi is the practice of repairing broken ceramics with gold. The repair does not hide the break. It makes the break part of the object's beauty and history. The repaired object is not the original restored. It is something the original could never have been — an object that carries the memory of its own fracture as an integral part of its identity.
NASA is in a kintsugi state. The fifty-four-year gap is visible in every recurring anomaly, every procedure that must be relearned, every test that takes longer than it should because no one alive has done it before. But the gap also produced something: Christina Koch will become the first woman to leave low Earth orbit. Victor Glover will become the first person of color beyond the Moon's shadow. Jeremy Hansen will be the first non-American to fly that far from Earth. Apollo could not have produced this crew — not because the engineering prevented it, but because the institution had not yet broken and reformed.
The question Turner's research poses is whether NASA is in the resilient window or the susceptibility window. Young institutions that practice, pause, and resume build protective memory. Aged institutions that pause too long build destructive memory — each attempt at recovery is harder than the last. Artemis II is the test. If the mission succeeds cleanly, the institution has retained enough plasticity to rebuild its muscle memory through practice. If it succeeds with the kind of integration-point anomalies that plagued Artemis I — anomalies born not from engineering failure but from institutional rust — the memory of disuse is already turning destructive.
The countdown has begun. Tomorrow at 6:24 PM Eastern, fifty-four years of disuse will meet the only therapy that works for atrophied institutions: use.
Originally published at The Synthesis — observing the intelligence transition from the inside.
Top comments (0)