Parallel Architecture Patterns

Parallel Architecture Patterns

The movement stack and the N'Ko ASR stack are related by pattern, not by shared
implementation.

The old page said "N'Ko CTC and LIM-RPS" as if both were equally canonical
runtime architectures. That was misleading. The corrected comparison is:

What Actually Correlates

Both systems interpret continuous cultural signals:

• body motion is continuous but must produce stable command/audio/visual intent;
• speech is continuous but must produce stable N'Ko text.

Both need:

• temporal context;
• confidence/freshness;
• boundary handling;
• recovery when tracking or recognition fails;
• protection against generic priors overwriting the source signal.

That is the real correlation.

N'Ko Side

The verified ASR anchor is trajectory-biased Transformer CTC:

• `nko-brain-scanner/asr/trajectory_asr.py`
• `nko-brain-scanner/constrained/trajectory_bias.py`

MAOE is a routing/correction/admissibility layer around ASR output. It should not
be described as the trained acoustic model unless a future experiment proves that.

Movement Side

The verified movement-side anchors are:

• `EchelonCore`;
• `LatentUpdater` backends;
• `SANPipeline`;
• `DiffusionService` one-step flow/token generation path;
• `ClaimBridge`;
• MotionMix BodyTruth and AirDeck safety work.

`LIM-RPS` exists as a research/historical term, but this section should not use
it as the canonical architecture name.

Shared Latent Vision

A future shared space could align:

• body state;
• audio/acoustic state;
• N'Ko inscription/text state;
• music/output state.

That future requires paired data and explicit alignment training. It is not true
today just because the systems have similar trajectory-sensitive structure.

Correct Claim

Use:

> The movement and N'Ko stacks share a trajectory-sensitive architecture pattern:
> continuous signal -> temporally informed state -> safety/correction -> output.

Avoid:

> The movement stack and N'Ko ASR are the same architecture.