How Nobody Leaves on Schedule was built

From paper to story

This document is the public record of one editorial run. It summarises the evidence used, the alternatives considered, the selection criteria, and the changes applied before publication.

The story was produced under narrative methodology 1.4.1. This record uses public-process format 1.5.1, introduced later, and was reconstructed from artifacts preserved during that run.

Starting point

The run did not begin by asking an AI to write a story. It began with eight scientific works available in the editorial reserve.

The selected work was arXiv:2603.02223v1, a study of wildfire evacuation behaviour.

The narratively useful feature was not that a model could predict everything, but precisely the opposite: evacuation mode could be estimated more reliably than the moment at which a person would decide to leave home.

That asymmetry became the story's causal constraint.

Cauce can prepare vehicles based on how a population tends to evacuate. It cannot know with the same certainty when each household will decide to do so.

The paper does not demonstrate that Cauce, its autonomous fleet, Vega Alta, or Berta and Maia's decisions could exist. All of that belongs to the fiction.

The adaptation nevertheless preserves the limit that makes the conflict possible: a prediction can be useful without eliminating the uncertainty that the original research also leaves unresolved.

Six directions, one selected

• Nobody Leaves on Schedule — selected. A bus assigned to a carless village leaves another settlement where eight workers are absent from the registry. It was selected because it combined immediate physical friction, two legitimate incompatible interests, and a decision that did not require the model to be explained first.
• The Fifth Statement — finalist. An auditor reconstructs a past evacuation inside an urban twin and finds inconsistencies in the official account. It passed the probe, but required the twin, the records, and the later investigation to be explained at once. The scene also risked depending too heavily on testimony about events that had already happened.
• Everyone's Keys — discarded. A temporary requisition of autonomous vehicles turned the problem mainly into a property conflict, moving the centre away from the uncertainty studied by the paper.
• Herd Six — discarded. Livestock drones produced strong imagery but added too much narrative infrastructure and risked reducing the conflict to logistics.
• Three Colours for One Street — discarded. The choral structure could show different responses to the same prediction, but it broadened the number of focal points too much for the intended length.
• Those Who Always Wait — discarded. An insurer penalised people who systematically delayed evacuation. The conflict drifted toward contractual negotiation and came too close to mechanisms used in recent stories.

Discarding a proposal does not mean declaring it bad. For this run, another alternative preserved the scientific core better, produced more visible opposition, and differed more clearly from the existing archive.

Two probes before the draft

Only the lead candidate and the finalist became scenes.

Each produced a short prose probe before a complete story was developed, to answer three questions:

• Is there a concrete objective?
• Is there a legitimate force capable of preventing it?
• Does the scene end with an identifiable narrative expectation still open?

Both passed this evaluation.

The audit proposal required reconstructing the past while introducing the urban twin, the registry, and the investigation at the same time.

The bus scene instead began with an irreversible action: Berta pulls the red cable and stops a vehicle that ought to leave.

The choice was made on dramatic economy and causal clarity, not on prior thematic preference.

What changed in review

The first complete story received the verdict APPROVED_WITH_CHANGES.

The evaluation did not identify a critical failure in its relationship with the science, main causality, opposition, or ending.

It did identify an important formal problem: the sequence of numerical warnings was turning the narrative pressure into a countdown and bringing the structure too close to a previously withdrawn story.

A complete rewrite was therefore carried out.

The new version made four main changes:

• removed the numerical countdown and replaced it with qualitative changes in the state of the escape corridor;
• made it understandable why the eight workers were absent from the registry through the repeater cabins, contractor credentials, and annual data update;
• strengthened Maia as legitimate opposition: she protects twenty-two registered residents of Vega Alta and also risks her own professional responsibility;
• retained the red cable, broken card, first vehicle failure, and final crossing because they establish and resolve elements of the central conflict.

There was no need to return to design or switch to the reserve candidate.

One complete rewrite was required before the Spanish edition was closed.

What fell away

Figures, interface messages, and explanations that described the system without changing the characters' decisions were removed.

The administrative absence of the eight workers was also kept out of expository dialogue.

The problem appears materially: the workers walk, arrive, and occupy the seats the system had not counted.

The final version concentrates the story around one causal chain:

an AI can know how a community usually evacuates better than when it will decide to do so → that information allows resources to be allocated usefully → an exception appears that the data does not contain → two people must decide how to complete that imperfect information → both incur a cost by doing so.

The technology starts the problem. It does not solve it.

Controls recorded before publication

These controls belong to the methodology used for this run. Some are structural checks and others are editorial evaluations.

• CF1 — Visible science fiction. The public AI, driverless bus, and autonomously operating vehicles appear within the first quarter and affect later decisions. Result: PASS.
• CC1 — Causal comprehensibility. The evaluation concluded that the story allows a reader to reconstruct, without consulting the paper, what capability Cauce introduces, what Berta is trying to do, what she decides, and what consequences follow. Result: PASS.
• T1 — Tension and cost. Berta and Maia protect different groups, both have real power to alter the bus's movement, and both risk professional authority. Result: PASS.
• B2 — Scene traction. The probe ends in a materially different situation: the vehicle is immobilised, a credential has been broken, and the alternative remains open. Result: PASS.
• J1 — Progression and payoff. Successive decisions reduce the available options, and the final crossing uses elements established earlier in the story. Result: PASS.
• L1 — Isolated evaluation. The final text was evaluated without consulting the paper or original design. At approximate 25%, 50%, and 75% reading points, concrete unresolved narrative expectations could be identified, and the ending resolved elements established earlier. Result: PASS.
• Bilingual QA. The ES/EN pair was checked for preservation of narrative facts, technological capabilities, relevant figures, opposition, losses, ending, and scientific limits. Result: PASS.

The L1 result describes what the evaluation procedure found. It does not demonstrate how a real reader will react.

The results above record the application of editorial criteria. They do not turn literary taste into an objective measurement or guarantee that different readers will assess the story in the same way.

Boundary of this record

This document is a public reconstruction based on artifacts preserved during the run.

It shows the source used, the alternatives considered, the selection decisions, material changes, and control outcomes.

The complete working artifacts remain in the private repository.

For that reason, this record provides public traceability of decisions, but does not in itself constitute an independent audit of the entire process.

Its purpose is to let the reader distinguish what came from the research, what was invented, which alternatives were discarded, and what had to change before the story was published.

Editorial contract

One general method. One concrete execution.

This record documents how the editorial process was applied to Nobody Leaves on Schedule.

It does not attempt to prove that the story is good.

It shows how the version that was ultimately published came to exist.