How The House Without Sides was built

This dossier provides public traceability for edition 006. It summarises recorded decisions without publishing the complete working artefacts. The run followed narrative methodology `1.6.1`; the public dossier contract retains format `1.6.0`.

Scientific starting point

The source reports a randomised experiment in which 92 adults spoke with GPT-4o about personal or non-personal topics. Self-concept moved significantly towards the traits exhibited by the AI in the personal condition but not the non-personal condition. The source check separated method, reported results, inferences and limits: an immediate effect, a small US sample, one model, an unmanipulated personality, an indirect mechanism and possible anchoring.

The house, Nara, the spatial maps, and all individual behaviours and causal chains are fictional compositions. The comparison checks fidelity to the source; it does not reproduce the study or constitute independent scientific validation.

Alternatives considered

Six possibilities covered five tension topologies and five progression modes. The selected design translated self-concept convergence into a separating couple whose dwelling can no longer distinguish their rooms. Its positive literary wager was alternating first-person narration in which grammar, objects and architecture make it difficult to separate change, influence and will.

The runner-up placed a conservator before portraits converging on her assistant’s personality; it lost because it made the effect more contemplative and offered less material agency. Other options explored kinship, translation, collective mourning and training, but were rejected for proximity to recent fingerprints or dependence on an overly abstract revelation.

Probes and selection

Two probes were written within the recommended length. Both passed `B1` and `B2`. The house won because the novum altered an immediate need—producing two dwellings—and every declaration had to reappear as a use, choice or cost before a wall could move.

Evaluation and rewrite

The adversarial evaluation returned `APPROVED_WITH_CHANGES`. It found a critical rule concentrated in one explanation, voices that were too similar, too many trait lists, and an ending that risked turning the common room into easy reconciliation.

One full rewrite followed. The rule of difference returned as declaration, test, use and cost; the voices gained distinct textures; terminology was reduced; the central room came to require two keys and shared maintenance for a year; and the ending replaced explanation with the lamp illuminating the box from two sides.

Resulting form and fingerprint

The dominant topology is interpersonal, with internal tension and material constraint. The mode is deterioration-transformation: the house loses its borders, attempts to restore old identities, and finally preserves an unassigned remainder. Vale and Simón alternate in the first person; the architecture changes only when a difference produces observable consequences.

The fingerprint departs from the recent archive through a domestic setting, two protagonists without an authority figure, no clock, relational and property costs, a climax of mutual recognition, and separation with a shared remainder. It does not combine institution, urgency and professional sacrifice.

Recorded gates

• `CF1_PASSED`: unused source and create-only claim.
• `B1_PASSED` and `B2_PASSED`: both probes showed agency, friction and transformation.
• `CC1_PASSED`: isolated reading reconstructed the novum, objective, decisions and costs.
• `T1_PASSED`: separation requires incompatible choices with spatial consequences.
• `J1_PASSED`: each section changes the question about influence, identity and responsibility.
• `L1_PASSED`: cuts sustained concrete expectations and recurrent motifs received a final payoff.
• `D2_PASSED`: essential causal facts return in different functions without critical referent collisions.
• Bilingual QA: facts, causality, ambiguity, motifs, scientific limits and ending match.

The final evaluation ran in a separate context that received only the frozen Spanish text. L1 describes that procedure’s result; it does not measure the desire, behaviour or response of real readers.

Public limit

This document derives only from recorded artefacts. Since the complete materials remain private, it offers public traceability, not independent external auditing or an objective measurement of literary quality.

Read the story · Read the methodology