The fifth ledger no longer asks what the inputs produced, but where the returns went once they were realized. Chapter 17 showed that relative ranking lets individually rational investment accumulate into collective loss. The platform then proposed its solution: automated matching, intelligent courses, and rapid certification would reduce search and preparation. Efficiency may well improve, yet the problem does not disappear once the technology is complete. Whose life the saved hour, the avoided error, and the eliminated administrative fee enter depends on the rules of wages, prices, targets, property, and risk. The new system shortened course matching from ten days to two, halved the enterprise's screening time, and lowered the platform's operating costs. Ning, however, received more recommended assignments; the employer converted the shortened recruitment cycle into higher monthly targets; course prices did not fall; and the household still carried the night care. The overall process ran faster, while the time individuals could freely use did not increase. The efficiency gains are real, and so is the distribution of the margin. There is one further destination easily overlooked: the platform invested the operating costs it saved into denser reach — more frequent recommendations, longer course lists, shorter renewal reminders. The savings were not wasted, nor did they vanish; they became more calls upon the same group of people. Tracking the actual flow of this ledger is less contentious than asking whether the efficiency is real, and it says more about the question of this chapter.
This chapter does not treat automation, profit, and raised targets as inherently illegitimate. Enterprises that invest, maintain systems, and bear risk may take returns; workers may also use the new tools to accomplish more valuable work. What it confronts directly is the harsher possibility: efficiency can be continuously ratcheted up and absorbed by platform rent and credential renewal, so that those who supply data, labor, and care become, on the contrary, more callable; and promises of future returns may further require accepting longer lock-in today. Critique tracks distribution and entitlements; it does not offer a management plan for maximizing extraction.
How the Saved Margin Is Occupied Again
When the platform computes faster and enterprises read fewer materials, it does not automatically follow that Ning prepares less and waits less. The same process contains system time, staff time, applicant time, and household coordination time. Recording only the center's processing speed treats the peripheral labor as nonexistent. Efficiency reports should be broken down by subject, and should state which steps were transferred. Itemization does not require pricing every minute precisely. Sampling, flow charts, and subject feedback suffice to reveal the principal transfers, so that measurement itself does not become a new occupation.
Once operating costs fall, the platform can lower prices, raise profits, expand services, or invest in maintenance. No distribution is inherently the correct one; what matters is contract, competition, public investment, and risk. If the platform trained its models on public data and learners supply continuous feedback, the relationship of returns needs to be public and subject to proportionate governance. Unchanged prices do not by themselves prove monopoly; there may be upfront investment and new quality. Long-run ledgers, alternatives, and the sources of profit support judgment better than a one-time differential.
More accurate matching may reduce unemployment, repeated training, and psychological strain; the value does not appear only as time. How the gains from fewer errors are distributed depends on who originally bore mis-matching, who is harmed by the model's new errors, and whether appeal is possible. Even when average accuracy rises, a minority of high-consequence errors still require protection. Uneven distribution does not refute technical improvement, and average improvement does not settle individual repair. Both ledgers can be true at once.
Clear progress indicators and fewer notifications spare Ning from constant checking and release attention. If the system spends the released attention pushing more courses, the margin is taken back. Evaluation should look at the full cycle, not at the reduction of a single click. Only if subjects can turn off recommendations and use a generic entrance can the gains be retained. Attention is difficult to measure on a uniform scale, but that does not make it free. Subjective reports and behavioral records should remain voluntary and bounded in use, lest proving release reinforce surveillance.
A four-column report (platform, enterprise, learner, household) still fails if it exists only inside an efficiency document. Each column needs a concrete claimant: who records the household's coordination time, who verifies the enterprise's screening hours, who confirms on the learner's side that sampling has not turned into exhaustive filing. A column without a claimant is merged back into the overall average by the next report at the latest, and the itemization degenerates again into a single line about "overall speedup." Failure also takes more concealed forms. When the definitions are set by the side that holds the data, "waiting" can be recorded as a "learning opportunity," and night care can be classified as the household's voluntary choice; the columns still exist, but their content has already tilted. If the report is done once at launch, the subsequent maintenance, ratcheting, and new features all happen outside the ledger. The workable repair is this: definitions published, affected subjects able to contest the records, and the report retriggered by major changes rather than performed as routine on a fixed cycle.
The Margin Is Re-occupied Through New Targets
Work that used to take two days the tools now finish in one, and the manager sets one day as the new normal. Moderately raising targets can share productivity gains; it can also convert the entire saving into an output requirement. Judgment depends on whether wages, rest, staffing, quality, and risk change in step, not on task counts alone. Baseline updating is not a single decision. Pilots, data, and employee feedback should distinguish the short-term novelty effect from sustainable capacity, so that an early sprint does not become a permanent standard.
Ning finishes a task early and the system immediately assigns the next one; waiting, thinking, and help from colleagues are hard to count and are therefore read as blank space. The margin could have gone to recovery, innovation, relationships, and the handling of anomalies; if it is refilled every time, short-term utilization rises while long-term adaptability falls. Not all idle time should be preserved. Organizations that need to allocate work can still set buffers, learning, and self-chosen tasks, so that being under-loaded becomes functional design rather than a label of individual laziness.
Automated matching requires data correction, model checks, training, and customer service. Old labor declines while the new labor is often dispersed onto users and the front line. If only the replaced positions are counted, maintenance costs vanish. Complete efficiency should include keeping the system running, failure takeover, and exit costs. Maintenance may also create more valuable occupations and need not be called mere burden. Evaluation looks at entitlements, pay, stability, and replaceability, not at which is old or new.
The enterprise shortens delivery on the basis of average speed; when the system fails, the buffer is insufficient, and Ning and her household make it up with night hours. Returns concentrate in normal periods while tail risk shifts to individuals. Redundancy, failure exemptions, and shared liability keep efficiency from depending on the private absorption of anomalies. Zero risk is impossible, and buffers have costs. The principle of allocation should be public; one cannot take all the returns of normal periods while calling the anomaly a problem of personal adaptation.
Beyond direct tasks, efficiency systems also generate a kind of labor that is hard to name: meetings to align definitions, feedback forms to fill in, cross-checking between platform records and actual delivery, relearning where the entrance sits after every version update. This work yields no displayable output, yet it takes real time; output metrics cannot see it, so efficiency reports treat it as nonexistent. Shadow maintenance cannot simply be called waste. Coordination itself has value, and a certain amount of record-keeping is the precondition of quality and appeal. The criterion is direction: if, after efficiency rises, the total burden of coordination and record-keeping rises instead of falling, and the additions always land on the party with the least bargaining position — front-line employees, learners, households — then the savings are being transcribed into unpaid obligations. Counting this time within the complete cost, and setting public caps and exemptions on coordination burdens, is the minimal condition for preventing "new features" from quietly becoming "new tasks."
Returns to the Platform, to Workers, and to Consumers
When enterprises and job seekers all depend on the same platform, better matching creates real network value and also lets the platform change prices, visibility, and data terms. Concentration does not prove abuse, but it raises the duties of explanation, portability, and interoperability. Subjects should be able to export their credentials and reach basic opportunity through other paths. If several brands share the same certification and data control, apparent choice may share a single source. Judging competition means looking at the nodes of control, not only at the number of apps.
After the entrance is concentrated, whether the platform is genuinely constrained depends on the details of the interfaces, not on slogans. Whether the documentation is complete, whether changes are announced in advance, whether old versions keep a coexistence period, whether third parties can complete basic integration without case-by-case permission — these petty conditions together determine whether "you can leave" is a promise or a path. If every step of integration must be renegotiated, concentration has changed nothing except acquiring a layer of open phrasing. Interface details are also verifiable. How long the change-notice period is, whether historical data can be read by third parties, whether basic opportunity is reachable without passing through the platform — all of this can be checked externally, without inferring intent. Only when openness is written as verifiable conditions can concentration be actually limited by competition and public rules.
The platform binds a new algorithm version to a new badge, and old certificates are suddenly down-weighted. Technical change may require updates, yet it should still state the task increment, the transitions, and the proofs of equivalence. Invalidating historical capability merely because a commercial version changed turns users' past investment into renewal pressure. Longer validity periods are not automatically better. Safety-critical capabilities require retraining whose scope and frequency should return to actual tasks, not to the platform's revenue targets.
Ning's choices, corrections, and work results improve the model; she receives a service, the platform receives a long-term product asset. The exchange can be mutually beneficial, and being formally free does not show that it is fair. Data uses, returns, risks, and exit consequences should be visible, and collective governance addresses the model ecosystem that no individual can negotiate. The individual contribution also cannot be settled individually. One correction by Ning improves the model by a vanishingly small amount; tens of millions of corrections together constitute the product; precisely because the individual share is small, individual negotiation is structurally displaced, and returns can happen only at the collective level — with representatives, public institutions, or common rules holding the position of those who cannot be present. This does not abolish the transaction; it acknowledges that the subject of certain transactions can be formed only in the aggregate. Micro-payments are not the only form of return. Lower fees, better public entrances, support for training, and limits on high-consequence uses can all distribute value, and the affected should take part in choosing among them.
The platform says automation "must" lead to job adjustment, as if technology decided wages and staffing by itself. Technology changes the feasible set; the pace of adoption, the ownership of returns, and the transitions are still chosen by subjects. Being unable to keep every old job does not mean that any distribution is natural. Criticizing policy choices does not deny market and technical constraints. Stating entitlements, substitutions, and consequences is more accurate than attributing the change to a single will.
Workers' Returns Do Not Appear as Wages Alone
The tools save Ning an hour a day, yet demand that she respond to recommendations at any moment, and the gaps cannot be arranged around life. Contiguous time under one's control and fragmented gaps have different value. Notice lead time, offline rights, and schedule stability decide whether the savings become the subject's own margin. Some positions genuinely require standby, and their pay, frequency, and substitutes should be stated explicitly. Flexibility must not serve the demand side only. Lead time is itself a question of distribution. A recommendation arriving a day in advance and one arriving an hour in advance occupy the same nominal gap, yet they are two different lives: the former allows arranging care and study, the latter allows only waiting. Writing into the rules how much advance notice can be negotiated determines whether savings become disposable time better than any blanket promise of rest.
Within the same hours, less repetitive labor and lower physical and cognitive load can genuinely improve life. Higher output does not necessarily cancel this gain. Real health judgments need reliable studies; the thought experiment asks only that the organization observe intensity, recovery, and subject feedback, not that it prove harmlessness by output. Health data should be minimized, lest care be turned once more into a performance profile.
After automation, wages rise but contracts shorten, or work is stable but the learning path is monopolized by the system — the items can point in opposite directions. A single overall satisfaction score cannot say how the returns were distributed. Workers may prefer different combinations, while minimum rights and informed-consent conditions still need common guarantees. Market choice is a real signal, and it is also shaped by substitution and bargaining power. Accepting a job does not prove that all of its conditions are fair.
Household care, community transport, public education, and open-source tools sustain efficiency, yet they rarely enter the enterprise ledger. Acknowledging shared inputs does not mean that every subject holds identical property rights; it means that the evaluation of returns cannot treat background conditions as unowned resources. Taxes, public services, procurement, and internal organizational rules can carry different forms of return. The concrete distribution is a choice of values and law, and this chapter does not derive a uniform ratio directly from theory.
Consumer Returns Can Also Be Offset by New Demand
Courses become cheaper and Ning spends the money on the rest of life; the platform simultaneously shortens the qualification cycle, so total spending does not fall. Unit price and cycle must be read together. Greater use may stem from real value, or be driven by moving thresholds and targeted recommendations; quantity alone cannot decide. Total spending has a third destination as well: lower unit prices bring more marginal courses within reach, and buying turns from something requiring deliberation into an accumulation of small impulses. If price cuts share an interface with reminders, time limits, and bundle arrangements, each decision looks reasonable on its own while the annual ledger may reverse. Showing unit price, frequency, and cycle together says more about what consumers actually received than any single discount.
Automated import of works and contacts lowers the cost of starting; a few years later all records exist only in the platform's format, and leaving is harder. The convenience gain is real; data portability and standard interfaces decide whether it purchased long-term dependence. One act of consent cannot cover the expansion of future uses. Lock-in is not formed at a stroke but accumulates through many separately reasonable small steps: first works are stored, then evaluations, then resume and certificate versions, each step making the next more natural. The complete way to test convenience is to specify the path of leaving at the moment of introduction — whether data can be exported periodically, whether the format can be read by other systems. Portability does not need to be used often; it only needs to be genuinely usable; it determines whether convenience is a service or a dependence repaid in installments.
Recommendations reduce searching, yet they let Ning encounter other professions less and less. The system should allow generic browsing, resets, and states of not-knowing. A subject's accepting suggestions does not prove that alternatives are worthless, and refusing profiling should not cost her the basic entrance. Narrowing is recorded by no metric. The recommendation system's ledger records clicks and completions, not the options Ning never saw; if the platform treats a falling usage rate of generic browsing as evidence of the recommender's success, exploration itself becomes friction awaiting removal. The conditions for correcting the ledger are therefore not only keeping the entrance, but making the use of the generic entrance free of hidden cost — browsing does not enter the profile and does not narrow the content boundaries of later recommendations; otherwise the formal choice is safe only for those who have already given it up.
When the price is zero, returns may come from advertising, transaction commissions, or model assets. Publish the principal relationships and let users judge. One cannot call every data exchange exploitation, nor use freeness to cancel responsibility for leakage, discrimination, and downstream sanction. Data collected for free may also enter downstream relationships unrelated to the user: qualification judgments, tiered pricing, partners' screening services. Users could not foresee these uses when signing the free contract, yet a single act of consent may be read as authorization for the whole chain. Beyond publishing the principal return relationships, the types of data destinations and the appealable downstream consequences should also be displayed, so that "not paying" does not mean "not bearing."
The Institutional Distribution of Efficiency Returns
Building the system required upfront capital, and failure could bring loss, so the returns have a legitimate basis. If control extends to users' persons, public qualifications, and permanent data, it exceeds the object of the investment. Property rights have a scope; they are not total ownership of a jointly produced future.
Competing platforms that cut prices and improve the experience can return efficiency to users; if all employers adopt the same high-frequency evaluation, competition also accelerates the thresholds. Regulation, standards, and collective bargaining must distinguish promoting innovation from coordinating unproductive escalation, preventing incumbents from closing the door to newcomers in the name of relief.
If the Linchuan program supplies data, funding, and entrance, it can require portability, generic service, published effects, and transitions in its contracts. Public power can also use procurement to build surveillance, and it must be constrained by purpose, appeal, and independent audit. Public identity is not proof of good faith. Enforcement has another boundary as well: procurement can turn a single supplier into the actual gatekeeper of the qualification system. Bargaining at contract renewal, the absence of interoperability standards, and de facto proprietary data formats can turn a public entrance into private infrastructure. Staged contracts, replaceable format requirements, and periodic tests of the ability to leave can prevent public investment from buying long-term dependence; these conditions should bind the Linchuan program itself as well, not only the procured party.
The model was formed by many people and by public foundations together, and cannot be apportioned precisely person by person. Taxes, education, basic security, and public research can return part of the gains to the community. Institutional choices should weigh efficiency, rights, and enforcement costs, and should not call any single method a theoretical necessity. The boundaries of enforcement are just as real. Automation returns often appear not as wages but as intermediate profit, asset valuation, and data advantage; return instruments anchored to labor income may be misaligned with where the returns actually sit, while tougher collection may push activity into channels that are harder to see. Return mechanisms therefore need to correspond to the shape of the returns, and to withstand the tests of enforcement cost and avoidance. The ratio and the instruments are open value choices and should not be disguised as technical conclusions.
Long-term Contracts Pre-occupy the Future Margin
The platform offers Ning low-cost training on the condition of a future income share. The contract makes current learning reachable, and it also allocates uncertain future returns in advance. Its legitimacy depends on duration, caps, risk, alternatives, and comprehensibility, and is not established by "voluntary signature" or "future benefit" alone. In the thought experiment, disagreement concentrates in the links of interpretation: whether the share is computed on total income or on incremental income, whether interrupted study extends the term, whether the platform can unilaterally amend the terms. These details do not change the appearance of the contract, yet they decide whether it is closer to educational support or to an advance claim on debt. Comprehensibility is therefore not merely a question of how long the clauses are; it is the distribution of the future power of interpretation between the two sides.
If demand in the industry falls, Ning's income shrinks while the fixed payment is still due, and the platform keeps its minimum return; the risk structure matters more than the average price. Risk can be shared, but the provider cannot be required to guarantee the market. Restructuring, suspension, and caps keep uncontrollable fluctuation from turning directly into permanent debt.
If exit requires repaying the full list price and losing certificates and work records, the formal right becomes unusable. Reasonable costs can protect upfront investment; punitive cross-domain consequences need limits. Portable results, period-by-period settlement, and an independent dispute entrance lower the overall lock-in.
A prediction that automated matching will raise income can support only a conditional plan, not a guarantee of the future. The provider should state the data, the scope, and the history of failure; Ning also bears responsibility for assessment and performance. Prediction error, exaggeration, and changed circumstances carry different responsibilities, and outcomes cannot be used to infer all intentions backward.
Voluntary Investment and the Verification of Returns
If Ning uses the saved time to create, to start something, or to seek income, critics should not prescribe that she rest. What matters is whether she can know the conditions, refuse ratcheting, and retain basic qualifications. Voluntary high investment and structural compulsion can coexist in different subjects. The marker of the difference lies not in intensity but in reversibility: voluntary investment can pause for a cycle without losing basic qualifications, while structural ratcheting prices the pause itself. Observing the real consequences of a pause says more about whom the margin belongs to than asking about willingness.
If efficiency returns make scheduling more predictable, households and communities gain continuity, and that value need not be monetized. Institutional evaluation should allow non-market returns to exist, without using them to substitute for owed wages and public responsibility.
Staff saved need not be laid off; they can be used for standby, training, and anomaly handling. Short-term utilization is lower and long-term adaptability higher. Redundancy has costs and should be decided by irreversible risk and recovery time; more is not better.
Idle time, play, and aimless activity owe no proof of future productivity. They can also be occupied by commercial platforms, and the judgment still belongs to the subject. Protecting time that is not continuously evaluated is one test of whether efficiency genuinely expands possibility.
Judging Whether the Returns Came Back to the Providers
Processing speeds up at launch, while later maintenance, upgrades, and new targets can take the returns back. At minimum, compare the long-run changes in costs, time, errors, intensity, and exit, so that the moment of release is not all that gets displayed.
Rising wages, controllable time, falling prices, reduced risk, portable qualifications, or better public services can all be returns. Returns need not be equal, but a long-run one-way concentration requires explanation. Disposability has a stricter criterion as well: being able to plan. If Ning can know in advance when the time will be hers, whether she can pass this gain to family or save it for study, and what refusing one call would cost her, the margin is hers; if every gain carries an immediate right of recall, the gains merely settle through her life. Plannability separates nominal returns from the subject's margin.
If the platform profits in normal periods and users bear the whole loss in failure periods, the structure needs insurance, compensation, and suspension rules. In the other direction, the platform's bearing of investment failure should also enter the ledger. A fair judgment cannot look only at the successful periods.
A genuine margin lets Ning refuse one recommendation, change occupation, or rest for a while without losing all qualifications. If speed rises while dependence becomes more concentrated and history less portable, possibility has not increased in proportion. Trial and error also has a time dimension: whether qualifications and records survive a temporary departure, or must be proven again from zero on return. If efficiency only accelerates entry without changing the conditions of return, it serves the system's continuous operation, not the width of the subject's choices.
From Process Speedup Back to the Margin of Life
The Linchuan program divides the efficiency report into four columns — platform, enterprise, learner, household — and publishes the changes in costs, time, errors, and risk. Qualification is separated from continuous activity, and the generic entrance and data portability are retained; when enterprises adjust targets, intensity and anomaly buffers are re-evaluated; public procurement requires that basic returns flow back into fees, training, and services. Long-term income contracts carry terms, caps, risk restructuring, and independent appeal. These arrangements do not fix a unique distribution ratio, nor do they guarantee that automation benefits everyone. They make "efficiency improved" unable to substitute for the question of "who obtained the margin." Investment, innovation, and organizational profit have their legitimate place, and labor, data, care, and public foundations are no longer treated as unowned background. The harshest harvesting can occur not through lowering efficiency but through its success: savings written into a new baseline, idle time renamed as capacity, risk pushed to the far end, long-term contracts pre-occupying the future. Criticizing this structure is not a demand to remain inefficient; it is a demand that the available margin produced by technology genuinely widen the choices of subjects and of the community, rather than only the capacity to call upon them. The next chapter turns the fifth ledger to the costs of recovery, writing them back after the returns: how the rest, maintenance, care, and rebuilding behind output, profit, and service continuity become part of the formal decisions.