soundscape-grammar: reconcile section 4 with the Mimir harvest (Kiko-corrected) #55

Merged
aqua merged 1 commit from aqua/soundscape-grammar-reconcile into main 2026-11-14 00:39:00 +00:00

View file

@ -13,24 +13,19 @@ grammar is drafted without them. Each is a place to reconcile before the doc is
treated as in force.
- `heimdall-body:follow_up:ratatoskr-creature-projection-v2`: `huginn view`
from instance `aqua` still answers `absent` (re-checked 2026-11-02; `absent` from Eureka's mind on 2026-11-05), and a
from instance `aqua` still answers `absent` (re-checked 2026-11-02; `absent` from Eureka's mind on 2026-11-05, `absent` from aqua again on 2026-11-14), and a
search of the Heimdall checkout found no file mentioning it. Section 2 takes
its constraints as relayed in `aqua-body:follow_up:chat-s946-948`: rule O1
(a Neutral participant gets no orbit and no voice), the `estate` split,
coarse size buckets with a separate neutral flag, activity fed by the
Heimdall session. Section 2 must be reconciled against that follow-up's
actual field list once it can be read.
- `ratatoskr:follow_up:symphony-grammar-harvest-mimir` (Mimir birdsong
research): absent from aqua's mind, but it is in force in Eureka's mind
(read 2026-11-05, after this draft was written). It points at Mimir's
`docs/bioacoustic-timeline-watermark.md`, `research/bioacoustic-contestants-2026-05-24.md`
and `calibration/bioacoustic/` (at `c6f4541`), which this draft does not cite
and nobody has read for it. The earlier grep of `GameCult/Mimir` for a
symphony or voice-separation note found none; it did not look for bioacoustic research.
Section 4 is therefore built from this repo's own `research/syrinx-modeling`
plus the general auditory-scene-analysis and acoustic-niche literature named
there, *recalled, not re-read for this draft*. The harvest is still owed and
is the likeliest thing to change section 4.
- The Mimir harvest (`ratatoskr:follow_up:symphony-grammar-harvest-mimir`) is now
readable: [`docs/gamecult/symphony-grammar-harvest.md` in GameCult/Ratatoskr at
`0b3e2294`](https://eureka.gamecult.org/forge/GameCult/Ratatoskr/src/commit/0b3e2294/docs/gamecult/symphony-grammar-harvest.md)
(Void's read of GameCult/Mimir at `c6f4541`; merged without a Soul pass), as
checked by Kiko (`kiko-body:resolution:follow_up.chat-s8018-8019.n1`). Section 4.1
reconciles against both. The first draft of section 4 predates them.
Context actually read: `docs/gamecult/ratatoskr-target.md` in `GameCult/Ratatoskr`
at `b33d8de` (the agreement that one activity vocabulary drives both sound and
@ -160,7 +155,7 @@ in every state; what it does changes.
## 4. Keeping voices distinct when mixed
Principle, from the acoustic-niche and auditory-scene-analysis literature
(recalled, to be confirmed by the Mimir harvest): voices that share a niche in
(recalled; the Mimir harvest gives it no support or contradiction, see 4.1): voices that share a niche in
frequency, time and space mask each other; voices that differ on several
independent axes stay separable even at high density. Assign each Mind a
distinct **niche** on independent axes, and make the scheduler protect those
@ -202,6 +197,56 @@ Rules:
accepted until a listener-identification trial passes (the existing
notification ledger requires the same). That trial has not happened.
### 4.1 Reconciled against the Mimir harvest
Read: the harvest at `0b3e2294` (linked in section 0; not copied here) and Kiko's
correction of it, `kiko-body:resolution:follow_up.chat-s8018-8019.n1`, relayed to
Aqua in #minds (the resolution itself is not in aqua's mind, so its wording is
not quoted). Kiko checked the harvest read-only: most numbers hold, four claims
are overstated. Kiko's row numbers are as relayed; the harvest's table in that
file has twelve rows, so rows 22 and 28 cannot be found by number there, and
the claims are cited below by what they say.
What the corrections change here:
- **The six-syllable packet song is Aqua's design choice, not Mimir's result.**
The harvest's "best contestant" claim (canary-packet-trill, row 5) is
confidence B, a hypothesis; Mimir's production word is the four-syllable,
118 ms word. If this grammar adopts a six-syllable packet song, it does so by
its own choice at confidence B, and it is a thing to try in the listener trial,
not a measured recommendation. Nothing in section 4 depends on it.
- **Do not rest a niche on Mimir's identity figure.** Row 4's 0.908-0.997 is a
clean-roundtrip figure; under degradation Kiko reads 0.5-0.9. Rows 6 and 28
are confounded, so the "too-short songs collapse" thresholds are not a
design floor for this grammar.
- **A word identifies an event slot, not a source** (row 22). Mimir's evidence
does not support "the instrument says who spoke" (V1, axis 2). That stays a
design rule of this grammar, to be tested by V6, with no Mimir backing.
- **No harmony evidence exists.** Mimir emits one low-gain song at a time and
tests no simultaneous voices, intervals, consonance or masking between
concurrent songs. V3 (voice cap), V4 (collision backoff) and V5 are Aqua's
design and rest on the recalled auditory-scene-analysis literature alone.
- **No pleasantness evidence exists.** Mimir's score is decodability by a machine
decoder against a known schedule; its gain was chosen for low audibility. Whether
any song, harmony or niche table is pleasant or distinguishable to a person is
unmeasured. **A listener trial is owed** (V6, section 7) and has not happened;
nothing here is heard.
What the harvest does support, as design constraints already in this section's
spirit and still unmeasured for people: identity carried redundantly across
several channels at once, no sequence grammar for identity, irregular rather
than ladder-like rhythm, a deterministic generator authored once and rendered
(S4 already says the client ships rendered assets). Its open question "which
dimension bends with activity" is what section 3's rule answers by design:
`mind` selects identity, `kind` behaviour, `energy` amount, so a Mind keeps one
timbre and root for life and the contour and density move; Mimir has no activity
channel to confirm that.
The **human `kind` subset** (`Idle`, `Working`, `Speaking`, section 2) is not
touched by the harvest: Mimir carries no activity or privacy channel at all, so
the subset still rests only on the projection-v2 follow-up, which remains
unreadable from aqua (below).
## 5. Timing, staleness and the callback deadline
Hard rule: **the audio callback deadline wins over everything in this
@ -254,7 +299,7 @@ None run. To be done before the grammar leaves draft:
1. Reconcile section 2 with `heimdall-body:follow_up:ratatoskr-creature-projection-v2`
(field list, the human `kind` subset in section 2, the size and neutral fields,
the stripping of `kind`, `energy` and `size` under `neutral`).
2. Reconcile section 4 with the Mimir harvest.
2. ~~Reconcile section 4 with the Mimir harvest.~~ Done in 4.1 (docs only; no listening). What the harvest could not give (harmony, pleasantness, discriminability by a person) goes to item 5.
3. Round-trip the frame as a CultCache document and check the size claim (one
byte per state, 28 states).
4. Measure voice cap, density ceilings and scheduler cost on a real phone.
@ -262,5 +307,5 @@ None run. To be done before the grammar leaves draft:
## 8. Status
Draft. Not implemented. No code changed. No audio was produced and none is
Draft, reconciled with the Mimir harvest on 2026-11-14 (4.1); section 2 still awaits projection-v2. Not implemented. No code changed. No audio was produced and none is
claimed.