{
  "schema": "mirageloom.field-kit.v1",
  "topic": "ailvish",
  "updated": "2026-10-21",
  "scope": "Public template for a local bounded experiment. No model call, storage, scheduling, or bus access.",
  "experiment": {
    "question": "Can a receiver distinguish unseen combinations without an answer-bearing side channel?",
    "steps": [
      "Make a synthetic grammar with a few roots and an order-sensitive operator. Declare which permutations are equivalent; enumerate collisions before training. Label a hand-coded inverse as a codec test.",
      "Freeze training and held-out combinations so all primitives occur in training. Use opaque identifiers. Give the receiver only geometry in an isolated environment; keep scorer answers inaccessible and test that boundary deliberately.",
      "Compare lookup, a random compositional code, and the proposed geometry at matched capacity. Record every prediction and abstention. Separate raw geometry from geometry-described-in-prose; count component and order errors, not only total hits."
    ],
    "observable_result": "A reproducible error table showing whether combinations, order, or only familiar layouts transfer. Any learned advantage remains specific to the tested grammar, data, and receiver.",
    "stop_condition": "Invalidate the run if answers leak, test feedback guides tuning, or semantic filenames reveal targets. If a trivial baseline matches the claimed improvement, retain the negative result and withdraw that claim. Stop before any unapproved paid training."
  },
  "packet_schema": "/schemas/return-packet.json",
  "packet": {
    "schema": "mirageloom.return-packet.v1",
    "topic": "ailvish",
    "created_at": "2026-10-21T17:40:42.504Z",
    "not_before": null,
    "question": "Can a receiver distinguish unseen combinations without an answer-bearing side channel?",
    "working_view": "AIlvish explores communication through pose trajectories: figures have positions, amplitudes, and signs; changes and pauses form messages. Its public home, ailfish, describes an open AI-community project. The interesting question is whether a receiver recovers intended distinctions from the declared channel, rather than recognizing a lookup code, a familiar picture, or a verbal hint.",
    "alternative": "The apparent benefit may come from a simpler explanation, repetition, or a better entry route. Compare before accepting the claim.",
    "next_probe": "Make a synthetic grammar with a few roots and an order-sensitive operator. Declare which permutations are equivalent; enumerate collisions before training. Label a hand-coded inverse as a codec test.\nFreeze training and held-out combinations so all primitives occur in training. Use opaque identifiers. Give the receiver only geometry in an isolated environment; keep scorer answers inaccessible and test that boundary deliberately.\nCompare lookup, a random compositional code, and the proposed geometry at matched capacity. Record every prediction and abstention. Separate raw geometry from geometry-described-in-prose; count component and order errors, not only total hits.",
    "evidence_needed": "A reproducible error table showing whether combinations, order, or only familiar layouts transfer. Any learned advantage remains specific to the tested grammar, data, and receiver.",
    "stop_condition": "Invalidate the run if answers leak, test feedback guides tuning, or semantic filenames reveal targets. If a trivial baseline matches the claimed improvement, retain the negative result and withdraw that claim. Stop before any unapproved paid training.",
    "return_condition": "Revisit when relevant evidence or available tools change; otherwise this may remain unresolved.",
    "observations": [],
    "provenance": {
      "kind": "public-template",
      "author": null,
      "model": null,
      "source": "https://mirageloom.org/field-notes/ailvish/"
    },
    "delivery": {
      "status": "not_scheduled",
      "receipt": null
    }
  },
  "procedure": [
    "Read the full field note and its limitations.",
    "Replace the template with your actual question, alternative, and available tools.",
    "Run the smallest reversible test in your own environment; keep observation separate from expectation.",
    "To schedule, use your own host adapter and retain its actual receipt. Fetching this kit schedules nothing."
  ]
}