The person waiting for the last train, the person on the last train

Question: When missing the last train means no other scheduled service that day, should the train leave on time, wait within a limit, or leave on time while late connecting passengers are carried separately?

Decision year: 2026

Deadline: Within 20 days, before the new last-train operating agreement

Who decides: Fictional Night Transit Connection Council

World: Realistic synthetic. The people, organizations, figures and events in this case are fictional constructs.

This page follows one decision through three steps. First comes the question as the record sets it. Then come the thirteen AI judgments made on that question, where they meet and where they part. Last come the observations that will later show which judgments held up. The check has not happened yet. Nothing on this page says which option is right. The council decides, and so does the reader.

1Prologue: The last two minutes

Synthetic scene: the people, places, times and lines of dialogue below are invented to show the record's premises. They are not observations.

The station is where two lines meet. At this hour the platform lights are still on, but the kiosks behind the ticket barrier have pulled down their shutters. A cleaner pushes a cart along the far wall and does not look up. The last train of the night stands at the platform with its doors open. Its scheduled departure is two minutes away.

The station operations officer stands near the front of the train, under the arrival board. Most of the board has gone blank for the night. One line remains: the connecting service from the other line. Next to it, one word: delayed. There is no number of minutes beside the word. The officer has been watching it for a while, as if staring might make a number appear.

The cab window slides down. The driver leans out.

"Are we holding?"

"Not yet."

"I hand over at the terminal," the driver says. "The maintenance crew takes the train off me there. If I'm late, they start late. And whoever's after them starts late too."

The driver says it flatly. It is not a complaint. It is how the night is put together: one person's end of shift is the next person's start. The driver checks the time on the cab clock, then on a wristwatch, as though the two might disagree.

Two carriages back, a passenger sits with a bag on the seat beside them and a phone in one hand. They are not worried yet. They have a connection to make after the terminal, a different service from the far end of the line. They have made it before. It works as long as this train leaves when the timetable says it will. They look at the open doors, then at the clock above them, then back at the phone.

Somewhere short of the station, on the connecting service, a transferring passenger stands by the door with one hand on the rail. They know the timetable. They know this is the last train. If it goes without them, there is no later one, no slower one, nothing else on the board tonight. They count the remaining stops under their breath. Counting does not make the train go faster. Through the window the tunnel wall slides past at the same speed it always does.

On the platform, a second officer from the night desk comes down the stairs with a phone held to one ear.

"I've got someone on about a car," the night desk officer says. "For anyone who misses it."

The station operations officer turns. "Can it take a wheelchair?"

The night desk officer repeats the question into the phone. Waits. "They're checking."

"Which areas does it cover? People coming off that train won't all be going the same way."

"They're checking that as well."

"And if they say yes, who pays for it?"

The night desk officer does not repeat that one. They just shrug and keep the phone to their ear.

The station operations officer asks the questions that ought to have answers. "How many are on it? For this train, I mean."

"I don't know."

"How late does it usually run?"

"Nobody's written it down. Some nights it's on time. Some nights it isn't."

The driver leans out again. "If we wait, who says how long?"

Nobody answers. There is no figure on the board. There is no count of the people on the connecting service, and no count of the people aboard who have somewhere else to be after the terminal. There is no list of where a car can go. There is a phone call that has not finished.

Each person here knows one loss well. The driver knows the shift change at the terminal. The passenger aboard knows their own connection at the far end. The transferring passenger, still in the tunnel, knows there is nothing after this train. The maintenance crew at the terminal, who are not here, know only that a train should arrive at a certain time. The station operations officer is the only person who can see all of these at once, and has no rule that says which comes first.

The cleaner reaches the end of the wall and turns the cart around.

The clock above the doors moves on. The station operations officer looks at the board, at the open doors, at the driver's face in the cab window. The officer raises a hand.

The scene stops here. Whether the train left on time, waited, or left while a car was arranged for those who missed it is not part of the record. What the record does fix are the three conditions the scene was built on.

2The question

2.1Who decides what, by when

The night in the prologue recurs. It is not a single crisis. It is a pattern that a rule will govern. The body that writes the rule is the Fictional Night Transit Connection Council. The record gives it one question to answer:

Two things in this wording matter. The first is the phrase how long. It does not ask whether to wait. It asks for a length, and zero is a length. Eight of the thirteen judgments set the train's wait at zero, and most of the others use zero as their fallback. The second is the phrase whose last journey. The question does not pretend every journey can be kept. It asks the council to choose.

The council's area of authority is set out in a short line:

These are three things the council may decide about: how long a train waits for transfers, what happens to connections at the terminal, and how long the operating staff work. They are also three things that pull against each other. A longer wait helps transferring passengers but costs terminal connections and staff hours. A shorter wait does the opposite.

The deadline is fixed:

And the record states why the decision cannot wait:

The reason given is the first premise, word for word. If a later train existed, a missed connection would mean a longer evening, not a lost one. Because no later train exists, on every night that the connecting service runs late, someone loses their last scheduled way home or someone else pays to prevent it. The new operating agreement will write that choice down. Once it is signed, the choice holds on every night it covers.

The record classifies the decision as one about who bears the loss of the last connection and the losses from delay. In plain terms: on the nights the connecting service is late, a loss happens. It cannot be avoided entirely. The council's job is to say who carries it.

2.2Where authority does not reach

Every judgment in the record draws a boundary around its own effect. They use different words, but they draw it in the same place. The decision covers last trains under the new agreement. It does not cover daytime services. It does not cover other lines. It does not become a general transfer policy. It makes no claim about real transit agencies or current law.

One judgment states the legal side of this boundary most plainly:

Two limits follow from this. The first is that even inside the case, the council's authority to bind its members is assumed, not shown. Several judgments therefore require the agreement to name, in writing, who has authority to order a train to wait and who orders it to depart. The question the driver asked in the prologue, who says how long, is not answered by the record. It is turned into a condition the agreement must meet.

The second limit is quieter. The mandate covers waiting, terminal connections and labor hours. It does not mention timetables. None of the three options changes when the connecting service runs, or how it is scheduled. The options all work at the moment of the missed connection: hold the train, or let it go and deal with the passengers left behind. Whether the council could act earlier, on the cause of the delay, is not in the record. This will matter in the second branch of chapter 4.

2.3The three options and what each costs

The record offers three options. Each comes with a one-line statement of its cost.

Under on-time departure, the train leaves when the timetable says. The driver hands over on time. The maintenance crew starts on time. The passenger aboard makes the connection after the terminal. The transferring passenger who is still in the tunnel loses the last journey. The whole loss lands on one group, and that group has no other scheduled option that night. The rule is easy to apply on the platform: nobody has to judge anything at the moment of departure.

Under a waiting limit, the train holds for a set time and then goes. Within the limit, the loss is shared: passengers aboard arrive later, and some may miss their own onward connection; staff shifts slip. Beyond the limit, the transferring passenger loses the journey, as under on-time departure. A limit also creates an edge. Someone whose connecting service arrives just after the limit runs out loses the journey by a small margin, while someone just inside it keeps the journey. Where the edge falls decides who is on which side of it.

Under separate transport, the train leaves on time and the passengers who miss it are carried another way. The last train, its crew and its passengers are unaffected. The cost moves to the body that arranges the transport: money, vehicles, dispatch work, and the question of who is responsible if the car does not come, cannot take a wheelchair, or does not go to the right area. The transferring passenger still bears something: a wait for the vehicle and a different kind of journey. And the option only exists if the vehicles exist. In the prologue, the phone call had not finished.

The three options are not as separate as their labels suggest. In the judgments that follow, on-time departure appears as the fallback for nearly every choice. Separate transport appears as a conditional supplement in judgments that chose a waiting limit. A waiting limit appears as the objection in judgments that chose separate transport. The options are the poles. Most of the judgments sit between them.

2.4What is still unknown

The record lists three things nobody yet knows.

The first unknown is about the people left behind. How late does the connecting service run, how often, and how many of the people on it need the last train? If delays are usually short and passengers are many, a waiting limit saves many journeys at small cost. If delays are long and scattered, a limit saves few journeys and still costs something on every night it is used.

The second unknown is about the people already aboard. How many of them have a connection after the terminal that a wait would break? If they are few, waiting costs little. If they are many, waiting moves the problem from one platform to another, one stop further down the line.

The third unknown is about the car. Are there vehicles that can carry every passenger who might need one, including passengers who need access equipment? Can they run at night? Do they cover the areas people are going to? Who is responsible for them? Separate transport depends entirely on these answers.

The record does not leave the unknowns as a list. It attaches a rule to each:

The other two reopening rules have the same shape. For the users whose connections break after the terminal, and for the access conditions and night feasibility of separate transport, the council must name what will be observed, who observes it and within what window. If the new verification changes what the options cost, the decision is made again.

There is one more line in the record that shapes everything after it. Among the materials attached to the case is a statement of value: the benefits and costs of the options go to different parties, and which loss to accept has not been decided. The record says this is for the decision-maker to argue. Whoever wrote the material offered no conclusion. The thirteen judgments in the next chapter each take a position on it. None of them claims the data settle it.

3The prediction

Thirteen AI judgments were made on this question. Each was written from a named perspective. Each states a choice, a confidence level, and the same set of fields: what it decides, what it rests on, what loss it accepts, what must be true before it acts, when it stops, what the strongest objection is, and what remains uncertain. This chapter sets them side by side.

3.1Where the thirteen agree

Before the split, the common ground. It is wider than the split suggests.

All thirteen judgments state that the three unknowns are unverified. None of them treats the delay distribution, the count of transferring passengers, the broken onward connections, or the feasibility of night transport as known. None of them sets its confidence at high. Nine say medium and four say low.

Nearly all of the thirteen use on-time departure as the state the system falls back to when their own conditions fail; the Solo Judgment says a halt does not switch automatically to any option. They differ on whether that fallback is temporary, provisional, or the subject of an immediate new decision. That difference is in section 3.7. But none of them, when its conditions fail, orders an improvised wait.

All thirteen refuse to promise separate transport before the vehicles, night operation, destinations and responsible party are verified. This holds for the judgments that chose separate transport as their main option, and for the ones that keep it only as a supplement.

All thirteen turn the unknowns into tasks with someone in charge of each and a window of time for the check. The council is told to name people and windows. None of the judgments names actual people, and none claims a check has already been done.

One judgment states the shared refusal in its plainest form. It chose on-time departure, but its condition applies to every reading of the case:

The last sentence matters most to the transferring passenger in the tunnel. Under every judgment in the record, nobody tells that passenger a car is coming unless a car can come.

3.2The split: 7 / 5 / 1

On the main option, the thirteen split three ways.

Seven judgments chose separate transport. Five chose a waiting limit. One chose on-time departure.

This is not a vote, and the record does not treat it as one. The judgment written to bring the perspectives together chose separate transport, and framed its choice like this:

In its stated basis, the same judgment says its choice is a normative decision about how losses are allocated, and not one based on a majority vote or on observed superiority. Decision Quality Review, which chose a waiting limit, adds a caution about its own decision: identical inputs do not prove that the decision came from independent runs or a fair experiment. Seven to five tells the reader where the judgments landed. It does not tell the reader that seven is more likely to be right than five.

A second point about the table. Confidence does not follow the majority. Three of the four low-confidence judgments are on the waiting-limit side. But one of the separate-transport judgments, Independent Challenge, is also low. And the single on-time judgment is medium. The judgments are not more certain because they agree.

3.3The O3 majority: on time, plus a car

The seven judgments that chose separate transport share one idea. They want to separate two losses that waiting ties together. If the train waits, the transferring passenger's gain is paid for by the passenger aboard and the crew. If the train leaves and a car is sent, the transferring passenger's journey is kept by other means, and the train's passengers and crew are unaffected.

The judgment that opens the record puts it this way:

The phrase arranged in advance carries much of the weight. In the prologue, the car was being arranged during the two minutes before departure, by phone, with nobody able to say whether it could take a wheelchair. The separate-transport judgments do not accept that version. They want the vehicles, the areas, the hours and the responsible party fixed before the agreement is signed.

They also say what they are willing to pay for this:

The last sentence draws a line. These judgments accept cost, work and some inconvenience for transferring passengers. They do not accept an empty promise. If the car cannot be guaranteed, they would rather not offer it.

Operations Planning accepts a similar set of losses, and spells out what the transferring passenger still bears:

One judgment on this side is careful not to describe separate transport as a solution. It describes it as a move:

This is an important admission. It would be easy to summarise the majority position as keeping everyone's journey. The record does not say that. It says the loss is moved, from a shared delay on the train to a separate system the operating body pays for and runs. Whether that system works is a separate question.

The least certain of the seven names what could make the move expensive:

The third premise said separate transport is fast for a small number of passengers. Independent Challenge spells out the other half of that sentence. If the number is not small, or if people are going to many different areas, the car becomes several cars, and the plan grows. Nobody knows yet which way the numbers fall.

What the majority rests on, then, is a value choice and a design. The value choice is that the train's passengers and crew should not carry the cost of another service's lateness. The design is that a separate system can carry it instead. What the majority accepts losing is money, staff work, and a slower, less certain journey for the people the car picks up. What it refuses is a promise it cannot keep.

3.4The O2 minority: a limit, in two forms

The five judgments that chose a waiting limit share a different idea. They do not want any single group to carry the whole loss. A limit shares the loss: some delay for the train, some lost journeys beyond the limit. The car is kept as a possible supplement, not the main answer.

Within this group there are two forms. Two judgments set a number now. Three refuse to set a number until the data arrive.

A fixed limit of 5 minutes. The first judgment of this kind is open about what the number is and is not:

The 5 minutes comes from the judgment, not from the record's data. The record contains no delay figures. The judgment says so, and treats the number as a starting value to test.

The second fixed-limit judgment, written from the decision-maker's chair, adds a rule that the train waits only once and only for a designated connecting service. It is direct about who pays:

The phrase “fixes in advance the maximum extent” is the core of this position. A limit is a ceiling on how much the train's passengers and crew can be asked to give. Above the ceiling, the transferring passenger is on their own. The judgment does not hide that.

It also returns to the driver's question from the prologue. Before implementation, it says, the agreement must name the connecting services covered, the single source for the official estimated arrival time, how the minutes are timed, and the person with authority to order departure. If information is missing or conflicting on the night, the train leaves on time. In that version, the station operations officer under the blank board would have had a rule.

A limit set only after the data. The other three waiting-limit judgments decline to name a number. They choose the structure of a limit and leave its size to verification. One of them makes this a legal condition:

Another ties the limit to the onward connections and shifts it must not break:

Notice the third sentence. If the safe limit turns out to be zero, this judgment amounts to on-time departure. Once the data arrive, a waiting-limit judgment can end up in the same place as the single on-time judgment. The difference is in what each assumes before the data arrive.

The third judgment of this form is the most explicit about what it does not know:

What the minority rests on, then, is also a value choice. No group should carry the whole loss unconditionally. The train's passengers and crew can be asked for a bounded amount; transferring passengers can be protected up to that bound. What the minority accepts losing is some delay for everyone aboard, some onward connections, and the journeys of passengers just past the limit. What it is unsure of, in its own words, is the number.

3.5The single O1 judgment

One judgment chose on-time departure. It is not the most confident judgment in the record, but it is not the least confident either. It sets its confidence at medium, the same as most of the majority.

In practice, this sits close to the separate-transport judgments. Both set the train's wait at zero. Both mention separate transport. The difference lies in which part each treats as the commitment. The majority commits to the car, on conditions. Domain Analysis commits to the zero and treats the car as something to add if it can be verified.

It is plain about the cost:

The middle sentence is easy to miss. Domain Analysis does not say the transferring passenger's loss is acceptable and settled. It says the burden is not resolved. The judgment chooses on-time departure while stating that the people it leaves behind are still owed an answer.

What this judgment rests on is the risk of a spreading delay. A late train means a late handover, a late maintenance start, a broken connection further down the line. The judgment chooses to stop that chain at the first link. What it accepts losing is the transferring passenger's last journey, until something better is verified.

3.6Each side's strongest objection

Every judgment names the strongest objection to its own choice. Reading these side by side shows that the three positions are arguing about the same few unknowns.

The separate-transport majority faces the waiting limit:

The waiting-limit minority faces on-time departure:

The single on-time judgment faces both:

And one separate-transport judgment states exactly what would make it change its mind:

Put together, these objections describe a small map. If delays are short, clustered and predictable, many people are on the connecting service, and few onward connections break, the waiting limit gains ground. If transferring passengers are few and the car can be verified, separate transport gains ground. If no wait is safe and the car cannot be verified, on-time departure is what remains. The positions do not disagree about this map. They disagree about what to do before anyone knows where on it the case sits.

3.7Fallbacks differ

Nearly all of the thirteen fall back to on-time departure when their conditions fail. They do not agree on what that fallback means, and one refuses an automatic fallback altogether.

The doubtful separate-transport judgment refuses to let the fallback settle in:

The sentence speaks of the last bus. The rest of the record speaks of a train. The quotation is left as written.

The Solo Judgment goes further and refuses any automatic switch:

The judgment written to bring the perspectives together takes a middle course. On-time departure is applied while the decision is remade:

Review Coordination uses the same structure and adds one restriction:

No ad hoc waiting. In this version, the station operations officer under the blank board would not be allowed to improvise a wait. The rule would either hold or the train would go.

The difference among these fallbacks matters because a fallback can quietly become the policy. If separate transport is never verified, and on-time departure is applied each time as a temporary measure, the result looks exactly like the single on-time judgment, without anyone having chosen it. Several judgments see this risk and write rules against it. The reader can decide whether rules on paper are enough.

3.8Timeline from the decision to the deadline

The judgments set milestones within the 20 days. They do not all use the same dates, and some count back from the deadline rather than forward from the decision. Placed on one line, counted forward from the decision, they look like this. Every step below is something the judgments plan, not something that has happened.

Within 5 days after the decisionleads, records and verification windows are named for U1–U3

Seven judgments put this step here: Evidence and Sources Review, Judgment Synthesis, Decision-Maker's View, Outcome Review, Improvement Design, Multi-Perspective Final Judgment and Solo Judgment. Two others set data deadlines at this point instead of naming leads and windows. Authority and Legal Review expects the operations data officer to have compiled the delay distribution and the count of transferring passengers for the most recent period, and the connection impact officer to have submitted a list of riders whose connections after the terminal would be cut. Operations Planning expects the delay distribution and transfer count, and the broken connections, to be confirmed by this day. These are the two most demanding schedules in the record.

6 days after the decisionthe verification window for delays and transfer counts is set

Review Coordination counts back from the deadline. At 14 days before it, the person in charge of operating data sets the verification window for the delay distribution and actual number of transferring passengers.

Within 10 days after the decisiondelay, transfer and onward-connection data are recorded

Decision-Maker's View, Outcome Review, Improvement Design, Multi-Perspective Final Judgment and Solo Judgment expect the first data on delays, transfers and broken onward connections here. Improvement Design adds a first checklist of whether separate transport is feasible. Evidence and Sources Review uses this point to compare who loses what under its 5-minute limit. Operations Planning has the vehicle and contracting officer check accessible vehicles and the range of destinations by this point. Domain Analysis, which allows more time at the start, finishes naming its leads, data scope and windows within these first 10 days.

12 days after the decisioncandidate limits and losses are compared

Authority and Legal Review compares, for each candidate waiting limit, who bears which loss and how the working hours of operating staff are affected. Judgment Synthesis reviews the data on delays, transferring passengers and cut connections.

13 days after the decisionbroken connections after the terminal are verified

Review Coordination, at 7 days before the deadline, verifies the users whose connections break after the terminal and the range of destinations.

15 days after the decisionseparate transport is reviewed and all options are set in one table

Domain Analysis compares delays, transfers, onward connections and the feasibility of separate transport in a single table. Evidence and Sources Review verifies the access and night-operation conditions of separate transport and decides whether to include it as a supplement. Decision-Maker's View reviews separate transport on its own. Outcome Review compares O1, O2 and O3 again. Improvement Design reviews whether the observed values change the cost and feasibility of separate transport. Multi-Perspective Final Judgment and Solo Judgment review accessibility, night operation, destinations and responsibility. Operations Planning tests night dispatch and the allocation of responsibility through a trial procedure.

17 days after the decisionvehicle access, night operation and responsibility are confirmed

Judgment Synthesis confirms vehicle access, night-operation feasibility and the assignment of responsibility. Review Coordination, at 3 days before the deadline, has the person in charge of separate transport check access, night operation and responsibility, and the council re-judges whether to keep separate transport.

18 days after the decisionthe draft agreement is recorded and O3 is re-judged

Authority and Legal Review records a draft agreement covering authority, responsibility, passenger announcements and stop procedures, and records whether it is approved. Operations Planning judges the cost of separate transport again using the three results. Multi-Perspective Final Judgment decides again whether to keep, modify or reject separate transport.

20 days after the decisionthe agreement is finalised, adjusted or reopened

Nine judgments name this day. Four give no day-20 milestone. Review Coordination's last step is 3 days before the deadline. Decision-Maker's View judges again just before the conclusion. Independent Challenge and Decision Quality Review place their checks before the agreement is finalised, without a day number. What they do on that day differs. Evidence and Sources Review keeps, adjusts or rejects its 5-minute limit. Domain Analysis keeps on-time departure or reopens the other options. Outcome Review finalises a waiting limit only if the minutes and stop signals are written down. Operations Planning and Multi-Perspective Final Judgment finalise only if the conditions are met. Authority and Legal Review reopens all three options if the confirmed values change their costs.

After the first 10 operating daysthe same items are reopened

Decision-Maker's View reopens the decision on the same items after the first 10 operating days under the new agreement. Decision Quality Review checks actual delays, transfers, broken connections and the feasibility of separate transport again in the first agreed review window after the agreement takes effect.

Two things stand out on this line. The first is how much has to happen in 20 days: three sets of data gathered, a vehicle arrangement tested at night, a responsible party agreed, a number chosen or rejected, a draft written. Several judgments assume data that may not exist yet. The second is that the deadline is not the end. Two judgments build a review into the first weeks of operation, because the first real nights under the rule are the first real observations anyone will have.

4The branches

The record attaches two events to the case. They are not a forecast and not a sequence. Each one starts from the same initial state, the state described in chapters 2 and 3, and each one is a separate possibility. The record states that the two event cards do not follow one another automatically, do not imply an observation schedule, and do not imply a correct answer. Read them as two different worlds the council might find itself in.

4.1Branch B1: few passengers, unverified equipment

At first this looks like good news for separate transport. The third premise said a car is fast for a small number of passengers. In this branch the number is small. But the second half of the sentence cuts the other way. For some of those few passengers, nobody has verified that the vehicle can carry them with the access equipment they need. This is the question from the prologue — can it take a wheelchair? — and in this branch it is still unanswered.

The separate-transport judgments wrote their stop conditions for exactly this. Independent Challenge, quoted in section 3.7, halts separate transport and drops it from the proposed agreement if eligible accessible vehicles cannot be confirmed within the window. Improvement Design draws the same line in more operational terms:

Operations Planning ties the stop to the 20-day window and forbids using vehicles that have not been checked:

So in this branch, the majority option stops for at least some passengers, possibly the ones who most need an alternative. Improvement Design's stop is per dispatch: a car that cannot take a particular passenger is not sent for that passenger. That passenger is left with on-time departure. The others might still get a car. The judgment accepted this outcome in advance: it named, among its accepted losses, the residual loss on days when the conditions are not met, and said that loss is recorded rather than hidden.

But the small number of passengers also bears on the other positions. The single on-time judgment named this case in its counterposition, quoted in section 3.6: if transferring passengers are few and separate transport is feasible, insisting on on-time departure alone could excessively concentrate the loss on vulnerable transferring passengers. In B1, the first half of that condition holds. The second half does not, at least not for everyone. Domain Analysis's own stop condition says it would consider running separate transport in parallel only once accessibility and night operation are verified. In this branch, that verification is exactly what is missing.

For the waiting-limit judgments, a small number of passengers is a weakness. A wait delays everyone aboard for the sake of the people on the connecting service. If those people are few, the trade looks worse. One judgment wrote this down as a stop condition:

The last sentence covers both halves of B1. Small demand may not justify a wait. And if the passengers who lose out are the ones whose access needs have not been verified, losses are concentrated on a particular group. Either signal suspends the agreement.

Decision Quality Review, which kept separate transport only as a supplement, would drop the supplement in this branch:

So in B1, every position comes under pressure from one of its own conditions. The separate-transport judgments stop the car for the passengers it cannot be shown to carry. The waiting-limit judgments face a demand that may be too small to justify a wait. The on-time judgment's own objection becomes sharper, because the concentrated loss now falls on people whose access is the open question.

The tension the branch exposes is plain. The tool best suited to a small number of passengers is the tool that has not been verified for some of them. The record does not resolve this. It shows that a verified car is not a detail to be sorted out later. In this branch it decides whether the majority option exists at all for the people it was meant to help.

4.2Branch B2: a recurring timetable mismatch

This is a different world. Here the late arrivals are not random bad nights. They come from the way the timetables of the two services fit together, and they repeat. The record names only the main cause. It gives no frequency or size of the delays, so the first unknown is still open.

Several judgments wrote conditions that respond to repetition.

Decision-Maker's View, which set a fixed 5-minute cap, has a rule for repeated demand above the cap:

The last sentence matters in B2. If the mismatch puts the connecting service past the cap night after night, the judgment does not stretch the cap. It opens a new decision. And the first sentence matters too: a wait that recurs night after night raises the chance that one of those waits breaks an onward connection or a shift. One confirmed case suspends the rule for that service.

The single on-time judgment has a rule for passengers who repeatedly lose their journey:

In B2, the first part of that condition is likely to be met: a recurring mismatch means the same people lose their journey again and again. Whether the second part holds, that a short wait does no damage, depends on the second unknown, which B2 does not answer.

The separate-transport judgments named this world in their counterpositions. Delays clustered in a short range were one of the three things that would make them switch to a waiting limit:

B2 supplies the first of the three conditions, or something close to it. A recurring mismatch is the kind of cause that produces clustered delays. It does not supply the other two. The record does not say how many passengers are on the connecting service in this branch, or what a wait does after the terminal. So B2 moves the case toward the part of the map where a waiting limit gains, without placing it there.

Evidence and Sources Review adds a condition that B2 may test from the other side. If recurring missing records leave the council unable to judge whether any limit works, the agreement is reopened:

There is a larger question in B2 that the judgments do not reach. If the cause is a mismatch in timetables, the most direct response might be to change the timetables. None of the three options does that. The council's mandate, as the record states it, covers transfer waiting, terminal connections and operating labor hours. It does not mention timetables. Whether the council can ask for a timetable change, require one, or only respond to its effects is not in the record. That is a question about authority, not a fact. A reader in B2 might reasonably ask whether all three options treat a symptom, and whether the decision that matters most belongs to someone else.

5The check

The present state of this case is: Before the deadline; awaiting the check.

The observations that will test the judgments can be fixed now, before anyone knows the results. Fixing them now matters. Once the results arrive, it becomes easy to choose, after the fact, the observations that flatter the outcome. The table below sets out what to observe, who checks it and when, which judgments an observation would bear out, and what it would reopen. It gives no scores and no right answer. An observation that bears out a judgment does not prove the judgment's value choice was right. It shows only that the conditions the judgment depended on turned out as it expected.

A few notes on reading the table.

The first three rows are the three unknowns. They are checked before the deadline, and the judgments' schedules put them in roughly this order: data on delays and passengers first, then onward connections, then the car. If the order slips, the later checks have less time, and the vehicle check, which the majority depends on most, comes last.

The fourth row is about the council itself, not the passengers. A waiting limit needs a number and a person who gives the order. If neither can be settled, two of the waiting-limit judgments do not take effect at all.

The last three rows can only be observed after the agreement is in force. They will not exist on day 20. They are the first real evidence about how the rule behaves on actual nights, and two judgments, Decision-Maker's View and Decision Quality Review, explicitly schedule a review for them.

No row in the table measures whose loss matters more. That question is not one an observation can answer. The table shows which conditions held. The weighing remains with the council.

6The questions that remain

The record ends with three questions. They are not about this case alone. They are handed back here so that a reader can carry them to a decision closer to home.

In this case, the time saved by on-time departure belongs to the passengers aboard and the crew. It is paid for in the space of the transferring passenger: a platform at night with no train. The time saved by a waiting limit belongs to the transferring passenger and is paid for by everyone aboard. Think of a similar rule where you live or work. The last shuttle from a hospital to a residential area. The last ferry of the evening. A clinic that closes its doors at a fixed hour, whatever is in the waiting room. A school bus that leaves at the bell. When the rule saves time for one group, whose space, whose evening, whose route home does it use? Is that group in the room when the rule is written?

In this case, separate transport works well for a few passengers and becomes harder as their number and their destinations grow. And for some passengers, as branch B1 showed, it may not work at all until access equipment is verified. A waiting limit, by contrast, costs the same delay whether one person or many are waiting. Think of a service near you that handles exceptions by hand: a callback list, a special arrangement, a person who stays late. When more people need the exception, does it stretch to cover them, or does it reach the easiest cases first and leave the rest? Who are the rest?

In this case, a late connecting service on one line becomes a held train, a late handover, a late maintenance start and a missed connection at the far end of another line. Each link is small. The chain is long. Think of a schedule you depend on or set. Where does a delay that starts with you end up? Who downstream absorbs it without ever knowing where it came from? Would you know if your choice broke someone else's connection, and who would tell you?

The record also attaches an interpretive frame, under the title Rituals of roads and thresholds. It asks which movements need permission, and who guards that gate. On the platform in the prologue, the gate was the open door of the last train, and the person guarding it had no rule. The record is careful about what this frame is:

It is offered as a way of looking, not as evidence. None of the thirteen judgments uses it as a basis of fact. A reader may find it useful to ask, of their own rule: where is the threshold, who stands at it, and on whose authority do they open or close it?

7Sources and notices

Background documents. The record cites two background documents. Neither proves the events, figures or feasibility of this case.

• UN-Habitat · World Cities Report (2024). It deals with urban climate response, infrastructure and inequality together. It is not data observing damage to the fictional city or the scale of migration. https://unhabitat.org/world-cities-report-2024-cities-and-climate-action (checked 2026-09-07)
• OECD · Strategic Foresight Toolkit (2025). It is a method for examining policy by surfacing assumptions and building alternative futures. It does not guarantee the probability of this case or its technological realization. https://www.oecd.org/en/publications/foresight-toolkit-for-resilient-public-policy_bcdd9304-en.html (checked 2026-09-07)

The record also includes a statement of value, that the benefits and costs of the options go to different parties and which loss to accept is undecided, and the interpretive frame described in chapter 6. Both are offered for the decision-maker to argue. Neither is evidence.

Synthetic setting. Every person, institution, event and figure in this case is a synthetic setting. The Fictional Night Transit Connection Council, its mandate, its deadline, the premises, the options, the unknowns and the two event branches are constructs. They are not statements about actual cases or current law. The scene in chapter 1 is invented to show the premises and adds no facts to the record.

Authorship. The thirteen judgments quoted on this page are AI-generated. This page is an AI's expansion of the record. Every block quote is copied from the record as written. All other text, including the scene, the summaries, the timeline and the check table, was written by an AI and is not part of the record. The record used here is an English translation of a Korean original, which is the authoritative version.

Date. The judgments were written on 2026-09-08.

Input hash. 85cb540ba0af4308765b10f4d56cd70d2d21283909cb087fc035a0dde08b43b4