The Naked King

I did not return to the emperor and his new clothes because I needed a metaphor. I returned because the possibility of many gardens left me with a practical problem. If political systems are to learn responsibly from one another while remaining different, they need enough visibility into their own institutions to judge how an outside lesson relates to what already exists at home. Tolerance without legibility can preserve difference, but it cannot turn difference into learning.

The old tale contains a simple failure of speech. The king is visible, the crowd can see him, and social pressure prevents almost everyone from saying what is plainly in front of them. Modern political life often produces the inverse. The public may possess the law, the budget, the contract, the implementation notice, the audit, the correction, and the outcome report. No one has forbidden the truth. The difficulty is that the truth of their relation is not written in any one of them.

This was the small puzzle that first destabilized my forecast of Earth. I had expected secrecy to be the great obstacle to political learning. Instead I found myself defeated by disclosed material. A public promise would live in one document, its legal authority in another, the money in another accounting system, the software in a contractor’s specification, the experienced consequence in a later audit, and the correction on a webpage few readers would ever connect back to the original claim.

At first this seemed merely inconvenient. After what I had already found, it looked more consequential. A political system may possess a useful capability, a real safeguard, and a serious route of correction, yet still learn slowly because evidence about how those elements compose is scattered among institutions that answer different questions. The garden is not hidden. Its paths have no common map.

So I returned to the example that had first made this visible to me: the United States’ attempt to simplify its federal student-aid application. I do not return because FAFSA is the most important political program on Earth. I return because it is an almost perfect naked-king problem. Much of the consequential record was public, much of it was produced by competent institutions, and yet one ordinary question—what happened to the promise of simplification?—required reconstruction across law, administration, procurement, software, support, deadlines, and lived use.

That question is narrow enough to test a larger possibility. What changes when a machine can help hold those relations together while the policy is still alive?

The King Is Surrounded by Documents

The public policy aim was unusually legible at first. Congress enacted the FAFSA Simplification Act in 2020. The law changed the application process and federal student-aid formulas, and later technical corrections moved full implementation from the 2023–24 award year to 2024–25 while allowing selected provisions to take effect earlier. That sequence matters because the object called FAFSA Simplification did not become operative all at once. Its legal state changed before every form, warning, system, and institutional procedure could change with it.

One transition made the problem visible before the troubled 2024 rollout. Federal law removed Selective Service registration and covered drug convictions as conditions of federal aid eligibility, but related questions and warning codes remained temporarily on the form. Institutions were told to disregard those old warnings while the form caught up. A student could therefore encounter a message whose visible wording belonged to one version of policy while its legal consequence belonged to another.

Nothing in that sequence requires concealment. Statutes have effective dates. Administrative systems take time to change. Forms have release cycles. Schools need guidance. A responsible implementation can therefore contain several temporarily different versions of the same public object without anyone intending to mislead. The difficulty begins when a reader sees only one layer and mistakes it for the whole.

The records were separated for good reasons. A statute has to state legal effect; a Federal Register notice has to communicate administrative action; a form has to turn rules into questions a person can answer; a processing system has to convert those answers into data that other institutions can use. Demanding that every artifact narrate the entire program would make each one worse at its own job. Political legibility therefore cannot mean one enormous document that contains everything. It requires a way to move among specialized records without losing which one has authority over which part of the action.

That distinction mattered to me because I had initially treated fragmentation as evidence that the political object had been badly designed. Here fragmentation was partly the consequence of specialization doing useful work. The problem was not that the records differed; it was that no ordinary route preserved how their different kinds of authority and evidence composed. I was beginning to see that political visibility would have to respect institutional difference rather than flatten it.

The 2024 rollout multiplied the layers. Some families with a contributor who lacked a Social Security number could not complete the identity process needed to enter or contribute to the form until a technical fix was announced in March. Other applicants encountered errors or waited for assistance while colleges needed records to prepare financial-aid offers. During the first five months of the rollout, the Government Accountability Office found that nearly three quarters of calls to the Department of Education’s support center went unanswered and that submissions by first-time applicants were about nine percent lower than in the prior cycle, with the largest declines among lower-income students.

Those facts still do not answer whether FAFSA was simplified. A shorter form may be simpler in the number of questions it asks while becoming harder for a particular family to enter. A formula may be simpler to administer after the underlying data arrive while the identity route that supplies those data becomes more brittle. A legal simplification can coexist with a difficult implementation. The claim that FAFSA was simplified is therefore underspecified until the dimension, population, date, and operative version are named.

This was the first thing the naked king taught me. Apparent political contradiction can arise before anyone lies. Two institutions can issue accurate statements about different cuts of the same object, and the public can be left to decide which statement stands for the whole.

The identity of the political object is what makes the connection difficult. The phrase FAFSA Simplification can refer to the statute, the redesigned application, the processing environment, the aid formula, the implementation program, or the lived route through which a family seeks assistance. Those objects overlap without becoming interchangeable. A person asking whether the law reduced the number of questions is not asking the same question as a college asking whether usable records arrived on time. The shared public name invites them to speak as though they were.

Money Does Not Speak the Language of Delivery

The fiscal record deepened the problem because money changes meaning as it moves. In February 2025, the Department of Education’s inspector general reported that Federal Student Aid did not have readily available full accountings of appropriated funds, obligations, administrative expenses, and staffing for the FAFSA simplification initiative across fiscal years 2021 through 2024. Required spend plans had been submitted, but the first submission was late in each year reviewed and the plans contained limited initiative-specific detail.

That finding sounds like a statement about cost until one asks what kind of cost. An appropriation is authority to spend under stated conditions. An obligation is a legal commitment. An outlay is money actually disbursed. Administrative expense may sit outside a contract whose public price receives attention. A program can therefore have several defensible totals depending on which fiscal state and boundary the question uses.

The back-end FAFSA Processing System contract illustrates the same problem at smaller scale. The Government Accountability Office reported that the contract had been initially valued at about $121.7 million if all options were exercised. After modifications, the estimated total value had risen to about $236.8 million; by May 2025, Federal Student Aid had obligated about $100.3 million. These figures can be placed beside one another, but they are not three estimates of the same thing.

Nor did the contract equal the service. The Department described the FAFSA environment as depending on several interdependent components carried through multiple vendors: the applicant-facing form, back-end eligibility processing, tax-data functions, identity services, and hosting. A complete record of one award could therefore remain an incomplete record of the public capability that families actually encountered.

Once I saw this, the temptation to treat spending visibility as political visibility disappeared. A contract record can show what one unit bought without showing whether the purchased capability composed with the other capabilities around it. An obligation can show that government committed resources without showing whether a student could submit a form. An audit can document a failure without revealing every earlier decision that made the failure possible. Each record is real. The political object exists in their relation.

By then I had learned to be more generous toward such fragmentation. The existence of separated records does not establish corruption, incompetence, or bad faith. Procurement officers, program managers, contractors, colleges, auditors, and legislators perform different duties and therefore record different objects. The nakedness lies in the gap between those locally useful records when public judgment requires them to answer one another.

I recognized a pattern that had confused us on Luminara. Institutional specialization can increase local competence while reducing anyone’s ability to describe the whole. The finance office becomes better at finance, the technical unit better at systems, the legal office better at legal interpretation, and the local administrator better at the case in front of it. The civilization becomes more capable through differentiation, then discovers that the political meaning of a decision exists across the boundaries that made the capability possible. Complexity is not the opposite of competence. It is often one of competence’s accumulated consequences.

The Human Limit Is Attention, Not Skill

A skilled investigator can reconstruct that gap. The auditors who produced these reports already did much of the work. Lawyers compare amended statutes. Procurement specialists follow contract modifications. Program officials understand operational dependencies. Journalists and researchers connect records across time. The problem is not that human beings are incapable of the task. It is that the task grows faster than any one person’s continuous attention.

To answer one narrow question about FAFSA, a reader may have to identify the governing statutory version, distinguish eligibility from identity verification, locate the issue record current on the date a family was blocked, follow a technical correction into a later form state, distinguish contract ceiling from obligation, identify which vendor or public unit controlled the relevant function, and then compare all of that with support-center performance and college deadlines. Change the date and some of the answer changes. Change the student’s circumstance and another part changes.

Even identifying the same entity across those records is work. One source may use the program name, another an office name, another a contract identifier, another the name of a system component, and another the public-facing service. An amendment can change the legal object without changing its familiar title; a contractor can change while the function continues; a correction can preserve the page address while changing the rule it describes. A reader has to decide when two differently named things refer to the same entity, function, or institutional object, and when one familiar name has silently come to contain several different things.

This stopped me because I had met the same failure in Sareth. The evidence of intervention, substitution, withdrawal, and later diagnosis had survived in different archives while the relation among them faded. By the time Luminaran historians reconstructed the sequence, many of the substitutions had already become institutional facts that later governments inherited. Here the same problem appeared before my eyes in miniature: names changed, functions persisted, legal objects shifted under familiar labels, and no single record carried the identity of the whole relation. For the first time, entity resolution looked to me less like archival housekeeping than a possible reduction in historical delay.

Now multiply the problem by every major program, jurisdiction, contractor, regulator, public promise, and implementation change. Transparency does not scale automatically with publication. At sufficient volume, disclosure can produce a strange result: the public record becomes more complete while the capacity to reconstruct it becomes more specialized.

This is the point at which machine intelligence becomes more than a convenient research assistant. The relevant capability is not primarily that it can summarize a long document. It is that it can search across differently named records, compare versions, align dates and entities, retrieve the passages carrying a particular claim, and keep the proposed relation attached to the sources from which it was inferred. The work is connective before it is generative.

Imagine asking a narrower question than whether the whole rollout succeeded: Why could a contributor without a Social Security number not enter this application on this date? A useful system would not answer from general familiarity with FAFSA. It would locate the operative guidance and issue record for that date, distinguish the legal eligibility rule from the identity process that blocked access, show the later technical fix, identify the public unit responsible for the process, and keep any missing contractual or technical link visibly missing.

The same method can move sideways into money. It can follow one contract through modifications and obligations, connect the purchased function to the larger service architecture, and place a performance finding beside the requirement it actually tested. If the identifiers do not match cleanly, the system can offer a candidate connection rather than inventing certainty. If a source is unavailable, the gap remains part of the answer.

Nor must the reconstruction begin from a fixed master diagram. The relevant path depends on the question. A student blocked from the form needs the route from eligibility and identity requirements to the available workaround and the office that can correct the case. A legislator examining cost needs appropriations, obligations, contract modifications, and delivery evidence. An auditor may need the same sources but a different relation among them. Machine-scale visibility becomes politically useful when it can preserve these several traversals over a common record instead of pretending that one view is the program.

That plurality of routes changed my idea of what a political intelligence should do. On Luminara, the ambition to reconstruct a confused record often ended in a demand for one authoritative account. Here the student, legislator, and auditor could traverse the same evidence for different questions without one traversal becoming the sovereign picture of the program. The machine began to look less like a mind above politics and more like a connective instrument beneath several human inquiries: capable of holding a common record together while leaving the question, the institutional standing, and the judgment with the actor who had reason to ask.

Research on retrieval-augmented language systems and source attribution already demonstrates pieces of this capability: models can retrieve external records and generate source-linked answers, while evaluation research treats citation correctness and completeness as separate problems rather than assuming that a fluent citation makes the relation sound. That limitation is useful here. Political reconstruction becomes credible only when the machine can show enough of its route that another person can inspect where the connection holds and where it may have overreached.

When the Pieces Can Answer One Another

Once the pieces are connected, an important change occurs before anyone reaches a verdict. Statements that previously floated in separate institutional contexts begin to constrain one another. A claim of simplification can be placed beside the exact dimensions that became simpler and the populations for whom access became harder. A contract value can no longer drift casually into a claim about money already spent. A technical defect can no longer end at the phrase the system failed when the configuration, authorizing unit, and correction route are recoverable.

This does not make the record self-interpreting. One observer may emphasize the successful statutory reduction of complexity; another may emphasize the avoidable burden of a troubled launch. One may judge the implementation failure temporary; another may judge the timing politically decisive because college choices could not wait. The connected record does something more modest and more valuable: it gives those disagreements a shared object against which competing claims can be tested.

That shared object changes the quality of disagreement. A defender of the rollout can point to statutory simplification and later technical correction without having to deny the access failures. A critic can show that timing imposed real burdens without claiming that every part of the redesign failed or that every official statement was deceptive. The disagreement becomes harder, not easier, because fewer ambiguities are available as substitutes for judgment. But it also becomes more honest: the sides must say which part of the same reconstructed route carries their conclusion.

I had underestimated the importance of that condition. On Luminara, many political disputes survived not because the facts were wholly inaccessible but because each side could preserve a different cut of the record. One ministry cited the authorized program; another cited expenditure; a local unit cited implementation; a later government cited the corrected version. By the time historians assembled the sequence, the institutional decision had often become irreversible.

Earth’s emerging machines may reduce that delay. They can keep the legal version, fiscal state, implementation record, and experienced consequence near enough to one another that contradiction becomes visible while the object is still politically live. The novelty is not omniscience. It is temporal proximity between a claim and the records capable of testing what the claim actually means.

That proximity changes the value of public footprints. A correction no longer has to remain on the page where it was issued; it can be connected back to the earlier claim it revises. A new contract modification can be connected to the old ceiling and the capability it was meant to add. An audit finding can be connected to the implementation choice, without the connection being misrepresented as proof of motive.

For the first time I began to see why your political abundance of documents might become an asset rather than merely another source of fog. Earth has already produced much of the raw material of political memory and comparison. The missing capability is often the ability to make those footprints compose before attention moves elsewhere.

Power Leaves a Longer Shadow

The consequence reaches beyond research convenience. Power behaves differently when it expects its footprints to remain connected. That proposition must be stated carefully. Visibility does not compel honesty, and a government can ignore an audit, dispute an interpretation, accept political criticism, or possess enough power to resist correction. Connected records do not manufacture accountability.

They can, however, change the cost of certain forms of evasion. A public actor who changes a definition may still do so, but the earlier definition does not disappear as easily from the meaning of later performance claims. A unit that cites money committed as evidence of delivery can be confronted with the next link in the chain. An official who says a contractor chose the technical behavior can be asked which public requirement, acceptance decision, or delegated authority gave that behavior institutional force.

Earth offers some evidence that timely visibility can alter behavior. Monika Bauhr, Ágnes Czibik, Jenny de Fine Licht, and Mihály Fazekas analyzed more than 3.5 million European government contracts from 2006 through 2015 and found that greater tender transparency was associated with substantially lower corruption risk as measured through single bidding. The effect was driven largely by ex ante transparency—information available before award to firms able to monitor the bidding process. Single bidding is a risk indicator, not proof of corrupt intent, and the study does not establish that every disclosure regime disciplines every actor. It supports a narrower proposition: information can change the incentives inside a process when it is available while action is still possible and to actors able to use it.

Expected disclosure can also change where the record is created. A two-year study of the United Kingdom’s Freedom of Information regime found no broad evidence that the feared chilling effect on candor and recordkeeping had generally materialized. Parliamentary scrutiny reached the same conclusion — it found no certainty that FOI itself caused a general chilling effect — but it also recorded testimony the general finding does not capture: some officials described moving sensitive advice into oral briefings, shortening minutes after high-profile disclosure disputes, or preferring face-to-face conversation to a fuller written trail. That mixed result is more useful than a neat verdict. Visibility can discipline conduct inside a record while also creating incentives, in some settings, to move consequential reasoning outside it.

That qualification sharpens the machine’s role. AI can reconnect records that exist; it cannot recover a consequential judgment that was never recorded. Political legibility therefore depends not only on traversability but on duties to create, preserve, and attribute an adequate public record. The connective capability shortens the distance among footprints. It cannot manufacture the footprint that power successfully avoided leaving.

I find this qualification reassuring rather than disappointing. It prevents the new capability from becoming another Luminaran fantasy of total knowledge. A society still has to decide what public actions must leave a record, which reasons may remain legitimately confidential, and which institution must answer when the trail is deliberately broken. The machine can make an existing route easier to traverse; it cannot absolve political institutions of the duty to leave enough of a route to traverse.

The effect is incentive-related rather than moral. When contradictions are expensive to reconstruct, selective memory is cheap. When relations persist, a later statement is more likely to encounter the earlier statement, the governing text, and the consequence that followed. That makes precision more valuable before the controversy begins. Institutions have more reason to distinguish a target from an outcome, a ceiling from an obligation, a temporary workaround from a settled rule, and a correction from proof that no earlier harm occurred.

The expectation of connection can matter before publication as well. If a public unit knows that a future reader can move easily from its headline measure to the governing definition, from the definition to the operative date, and from the date to later outcomes, then qualifiers become less disposable. A target can still be ambitious and a press release can still advocate. The difference is that favorable compression leaves a recoverable trail to the conditions it compressed. Political rhetoric remains possible; contradiction becomes cheaper to inspect.

The same improvement can serve the institution being scrutinized. A department may discover that several offices are reporting incompatible versions before an auditor does. A legislature may see that an appropriation and an implementation schedule have ceased to describe the same plan. A public servant may locate the unit whose decision is blocking correction instead of treating the problem as an undifferentiated system failure. Legibility is not only a weapon citizens point at government. It can become a way institutions see across their own partitions.

For citizens, the shift is from access to interrogation. Today, formal transparency often answers the question, Where is the document? A reconstructable environment lets a person ask, Which document governs this step, what changed since the version I saw, which unit authored the condition, and what happened after the condition was applied? Those are the questions specialists ask because they know where political meaning hides. If machines can lower the cost of asking them, specialist method can become more widely available without pretending that specialist judgment has been automated.

This matters to the many-gardens argument because political learning begins at home. A society is less able to adapt a mechanism responsibly from elsewhere if it cannot reconstruct the mechanisms already operating within itself. Nor can it compare two political forms responsibly if one side is represented by doctrine and the other by a handful of outcome indicators. The first contribution of machine-scale political visibility is therefore almost embarrassingly elementary: make the existing political world inspectable enough that comparison starts from what actually happened.

I find that possibility more radical than an automated policy recommendation. A recommendation gives an institution another answer. Reconstruction gives more actors access to the question itself.

A recommendation also concentrates attention on the machine’s conclusion. Reconstruction can do the opposite: it can disperse attention back into the public record. The most useful answer may be a set of source-linked paths that leave the citizen with several possible judgments rather than one polished verdict. That is why I have become less interested in whether a political machine sounds intelligent. The more important question is whether it makes the surrounding institutions easier for other people to question intelligently.

Seeing Is the First Acceleration

When I arrived on Earth, I imagined artificial intelligence changing political development by producing superior analysis. I now think its first civilizational contribution may be simpler. It can lower the cost of seeing relations that political societies already record but rarely hold together. That changes who can inspect a complex public action and how quickly a contradiction can become common rather than specialist knowledge.

The difference is important because Luminara did not lack records. We had archives, auditors, courts, scholars, administrative reports, and eventually powerful analytical machines. What we lacked for much of our long crisis was persistent connective visibility across institutional boundaries. Lessons assembled slowly, often after the consequence had already become the next generation’s inherited condition.

Our retrospective histories often made this failure difficult to notice because the relations had eventually been reconstructed. Once a commission, archive, or generation of scholars had assembled the sequence, later readers could mistake the finished history for knowledge that had been available to the people acting at the time. It was not. The relation may have required years of access, translation, institutional cooperation, and political distance before it became obvious. A lesson discovered after the institutions that needed it have disappeared is still a lesson, but it cannot shorten the path that produced it.

I do not claim that Earth has solved that problem. The capability remains uneven, technically fallible, institutionally fragmented, and dependent on the quality and accessibility of the underlying record. Some evidence will remain legitimately secret; some records will never have been created; some relationships will remain disputed. A machine that cannot say where the record ends merely replaces political fog with fluent completion.

But the relevant historical possibility no longer requires perfection. If a citizen, journalist, auditor, legislature, court, or public agency can reconstruct more of a consequential route while correction is still possible, then some political lessons need not wait for retrospective history. The machine does not make institutions learn. It can make it harder for the lesson to remain invisible.

That is the first way I can imagine Earth shortening a timeline that Luminara once treated as unavoidable. Not by predicting the final political form, and not by asking an intelligence to decide which garden is best, but by making the present sufficiently visible that societies can recognize what they are actually cultivating.

Yet visibility has a temporal weakness. A route that can be reconstructed today may fragment again when the website changes, the contract ends, the administration turns over, the metric is renamed, or the correction becomes detached from the claim it revised. Seeing the naked king once is not the same as preventing the court from forgetting what it saw.

If political learning is to become faster than political catastrophe, the connected record has to survive its own present. Seeing is the first acceleration. Remembering is the next.