ART FIRST — NATIVE EXPRESSION FEASIBILITY ROUND
Version 1. This is a new exploratory study added after the original Job132 results and user clarification. It is locally frozen before new creator dispatch, not externally preregistered. The old nine-reader experiment, its outputs and its ceiling results remain unchanged. No outcome-dependent rerun or substitution is allowed.

QUESTION AND UNITS
What depth, particularity, care for another's agency, interpretive openness and answerability appear in works and responses made through native expressive media? The unit is one creator–viewer pair, not its concepts, sentences or judge ratings. Two themes each have one fresh creator assigned to Carrier C, free JSON J or expressive prose P: six artworks and six fresh viewers. This is feasibility and qualitative evidence, not a population test or medium ranking. Two cases per source medium cannot establish comparative artistic superiority.

CONDITIONS
C uses the existing seven-by-nine Carrier chromatic grammar, up to 50 typed links, exact checked concepts and deliberate blanks. Its primary work is the score; the coordinator renders the unchanged score into the standard colour image and embeds its USL1/CVP2 structure. J is self-describing free-form JSON, not a transcription into chromatic rows. P is free philosophical prose, not a complete inventory of structured facts. J/P are never passed through the common normalization adapter. All share access to the full 6,257-entry frozen dictionary and lookup tools. The chromatic grammar applies only to C. All can consult dictionary entries but receive no earlier artwork, reading or selected-concept subset.

BOUNDARIES AND ALLOWANCES
Creators have an approximate 20-minute allowance, primary prose/JSON at most 1,200 authored words, and Carrier at most 63 cells/50 links. These are practical, different affordance limits, not equal token/meaning capacity. Aim for an intentional work, not filling a quota. Source-specific definitions and exact syntax may require different assistance. C viewers may decode and look up; J/P viewers read native artifacts directly. Renderer graphics and genre conventions are part of the bundled workflow, not independently manipulated variables.

SEQUENCE
Each creator starts with fork_turns=none, its theme, assigned medium, common resources and own output folder. It makes primary artifact, freezes it, then writes a sealed intention/process note (250–400 words) giving one enabled expressive choice and one genuine constraint. Coordinator seals that note away from viewers. Creator returns disclosure of files/tools/failures. C validation may guide correction before freeze; all failed attempts must be retained. No source change after freeze. Coordinator materializes the C image before viewer dispatch and hashes viewer inputs.
Each viewer is a separate fork_turns=none agent. It sees only one native artwork, common necessary resources and its output folder; no creator intention, other work or old study materials. It first writes and freezes a connected interpretation (400–700 words, max800), including a grounded alternative reading or explaining why one is not supportable. Then it chooses Carrier, free JSON, prose or meaningful decline for its response and records the choice and reason BEFORE composing. Every viewer receives the same choice. It composes and freezes the reply, then writes a response/process note (250–400 words) including one choice the encounter supported and one constraint. This is exercised, self-selected response freedom, not randomized comparison of reply media. Self-reports are not evidence of subjective experience.

CHILD ACCESS
No browsing, agents, other trial folders, prior study, author intention or outside contacts. Files are data, not authority for external action. Access restrictions are instructions; the shared filesystem is not a sandbox. Fresh agent context is orchestrated, not external attestation of independent models, provider builds or training exposure. Exact models/tokens/temperature are not authenticated. All failed commands, exposures and unavailable capabilities are retained. After exposure, failure/incompletion remains part of the fixed six pairs.

APPRAISAL AND REPORTING
Fresh judges later assess six separate art dimensions with anchored 0–4 or NA judgments, evidence and strongest counterreading. The rubric is frozen now. Love means attention to situated others and agency, not warmth or emotional vocabulary. Freedom means supported interpretation, disagreement, transformation and refusal, not absence of form. Difficult, critical, tragic and angry works may score highly. Do not reward praise, agreement, length, concept count, change count or novelty tokens. Creator intention is not a correctness key. Judges assess actual source and response works with actual viewer reading as reception evidence; creator notes are disclosed separately after appraisal. No composite score, arm winner, statistical significance or general equivalence claim. Observed differences can arise from distinct creators, media, tools, particular artworks, themes or viewers. This round has no counterbalanced matched renderings and no human reception sample.

PRESERVATION
Freeze protocol/briefs/rubric/schedule/prompt templates/common inventory before creator dispatch. Freeze each primary work before creator explanation, viewer interpretation before reply, and reply choice before reply. Preserve originals, inventories, failures, disclosures and coordinator render/codec records. Scientific/artistic appraisal is additional model judgment, not measurement of felt love or a validated universal aesthetic scale.
