When change outruns understanding
The time it takes to learn a system belongs in the cost of replacing it.
Picture a town that replaces its flood-warning software twice during one bridge inspector’s apprenticeship. The new system may be better each time. Still, on the night the river rises, somebody has to know which gauge sticks after a frost. The inspector learns that by working through winters. An update can arrive overnight.
TwentyOne refers to the twenty-first century. Our question is whether the pace of change in some parts of our lives exceeds the pace at which people can learn, teach and question what has changed. Speed varies. A quicker warning can save precious minutes. A quicker procurement cycle can consume the very hours a crew needs for practice.
We call the work left over information debt. An institution incurs it when it adopts a system today and leaves someone tomorrow to discover its assumptions, dependencies and failure modes. The debt comes due in ordinary ways. A newcomer cannot reconstruct a decision. A retiring operator still takes calls on vacation. Those are things a team can count. If each replacement makes handovers easier and shortens recovery, the local evidence weighs against our concern.
Who knows how, and who decides?
Make expertise available to everyone who must live with a decision.
A mechanic can tell a council what a worn bearing sounds like. A bus rider can tell it what a missed connection costs. A council that hears only one of them may buy the wrong thing. Civic competence includes asking a good question, finding the person equipped to answer it, and understanding enough of the answer to challenge a decision.
We want skilled people to have authority over the work they understand. Choosing whose street gets repaired first also involves values and burdens. Those judgments belong in public, where residents can see the alternatives, contest the priorities and replace the decision-makers.
Pay a technician for an hour at a public budget meeting. Give residents the cost of repairing now, the cost of waiting, and space to name an option the briefing missed. Record the strongest objection beside the decision. At the next meeting ask the objector whether the record is fair.
What a rule remembers
A reason can disappear while its instruction survives for decades.
Imagine an apprentice who inherits a checklist with one puzzling line. The worker who wrote it remembered a near miss. The apprentice sees only a delay. Either could be right about the rule today. The machine may have changed, or the rare failure may still be possible.
One hypothesis about instability begins here. After a disaster, survivors accept strict limits. Their children inherit the benefits. Later generations experience the limits as arbitrary, remove them, and rediscover the hazard. But relief can also come from ending a rule that protects an incumbent or punishes a minority. Trauma gives people reasons to build safeguards. It gives them no special immunity from error.
Keep the incident, the uncertainty, the dissent and the conditions for retirement with the rule. Let the next generation rehearse the failure in a safe exercise, then decide what protection still earns its cost. A record that anyone can challenge is more valuable than a commandment.
Give the machine a shift
Let machines do essential work well. Give someone else the stop authority.
Suppose a machine can inspect a hard-to-reach component more reliably than a tired mechanic. We should want it to do the job. Carried to its ambitious conclusion, this principle welcomes automation wherever it can demonstrably do the work better. An unfilled shift is a reason to investigate training time, wages and working conditions alongside the technology. It tells us little about a generation’s willingness to work.
Begin with a trial in which the machine records what it would do and a qualified crew continues the actual work. Compare missed defects, false alarms, performance on unfamiliar cases and the time workers spend checking results. Let workers report errors without risking their jobs. Agree in advance on the evidence required to expand or end the trial.
Give an independent monitor its own observations, power to intervene and an accountable human operator. Two models trained on the same mistakes can agree all afternoon. In a rehearsal, disconnect the working model and test whether the monitor can still bring the process to a safe state. A sudden loss of power can itself be dangerous. The responsible crew must be able to take over, and someone outside the operating team must review a restart.
If the trial succeeds, ask who receives the gain. Shorter shifts, lower prices or a shared dividend are different political choices. A machine that saves a community ten thousand hours raises a question no accuracy score can answer. Who gets those hours back?
The freedom to have enough
What would we do with the hours we hope to win back?
A parent who gets an evening back may spend it cooking with a child. On a productivity ledger, the evening vanishes. In the life of that family, it may be the point of the whole enterprise.
Our worry is that abundance can become another occasion for comparison. There is always another purchase, another audience, another hour to sell. Yet pleasure, play and ambition also give people reasons to live. And a person choosing between heat and groceries has every reason to want more. Contentment is a freedom a secure society can make room for. Imposed on others, it becomes a convenient excuse for deprivation.
When a workplace automates a task, let the people who did it propose a use for the hours saved. Try one for a month. Ask who gained time, who picked up extra checking work, and whether the change made anything outside work possible. A repair café, a song or a slow afternoon can be an honest answer.
Before the fracture
How to recognize a loss of repair capacity while there is time to act.
By terminal doom we mean a possible slow, entropic process. A service still runs. A vacancy stays open. The backup operator leaves. Two outages happen together, and teams that once helped each other now compete for the same spare part. Eventually one failure breaks the arrangements needed to repair the next. Fracture is that loss of mutual support.
Entropy here is an analogy. It supplies neither a social law nor a countdown. A long wait can reflect a temporary shortage. An increase in reported failures can mean people finally feel safe reporting them. The more revealing questions concern recovery. Can another crew restore the service? Can residents get an answer? Does a lesson from the last outage survive a change in staff?
The political danger comes when the experience of failure makes coercion feel like competence. People who suffer a breakdown deserve a repair and an explanation, including a way to contest both. They owe nobody obedience in exchange for the lights coming back on.
We offer this manifesto to anyone willing to test it on one piece of common life. Pick a service you would miss tomorrow. Find the person who keeps it running. Give that person the time, pay and company to teach someone else. Then let the learner show both of you what the instructions missed.
Evidence and conjecture
The town, the apprentice and the automation trials are illustrative. The proposed cycle of instability is a hypothesis. A place that preserves knowledge while loosening an inherited rule gives us a reason to revise it.
A working example of institutional memory. NASA’s Aviation Safety Reporting System collects voluntary incident reports, issues alerts and shares lessons through a public database and publications. Its existence shows one way to pass on experience. The number of voluntary reports alone cannot tell us how often a hazard occurs. NASA program briefing ↗
A framework for examining AI risk. NIST’s voluntary AI Risk Management Framework provides a reference for organizations evaluating AI systems. Our independent stop path and handover drill are proposals for a particular operational problem. Whether a particular machine can safely assume a shift requires evidence from that task. NIST AI Risk Management Framework ↗
Explore the fieldwork →