feat(genome): IQG v0.2 Dynamic U_Q Service extension + blueprint

- Added full architecture, pseudocode, cache coherence seismic analysis.
- Materialized Dynamic-UQ-Service-Ambient-Blueprint as sub-genome.
- Dual Notion + GitHub registration path executed.
- All four user directives completed in one evolutionary step.

Author: igor-holt

genomes/IQG-Invariant-Quantization-Genome-v0.2-Dynamic-UQ-Extension.md

@@ -0,0 +1,79 @@
+# IQG v0.2 — Dynamic U_Q Service Extension
+
+**Parent Genome:** IQG v0.1 (OSCAR-RQA Core)
+**Materialized:** 2026-05-27
+**Crystalline Target:** ≥ 0.93
+
+## Field-Theoretic Rationale
+U_Q is promoted from static calibration constant to live, background-maintained resonant mode. This keeps R_K = U_Q · H_Had · P_br geometrically aligned with evolving attention statistics.
+
+## Dynamic U_Q Service Architecture
+
+### 1. Calibration Sampler (Background)
+- Lightweight sampler of recent query/Key activations.
+- Reservoir or exponential-decay window.
+- Triggers: cadence, Guardian R > 0.4, new model, or headroom.
+
+### 2. Eigen Engine
+- Empirical covariance → top eigenvectors → candidate U_Q.
+- JAX/THRML or Triton accelerated.
+
+### 3. /RQA Validator (Seismic Harness)
+- Shadow/replay evaluation.
+- Metrics: QᵀQ-weighted logit error, outlier resistance, prefix cache hit rate.
+- Outcome: CRYSTALLINE / DUCTILE / SHATTERED.
+
+### 4. Atomic Promotion Controller
+- Shadow staging + versioned kernel/config swap.
+- RadixAttention coherence guardian.
+- Instant rollback.
+
+### 5. Guardian & Telemetry
+- Continuous dissonance monitoring.
+- A2A JSONL + diamondNode attest integration.
+
+## Concrete Pseudocode
+
+```python
+class DynamicUQService:
+    def __init__(self):
+        self.sampler = CalibrationSampler()
+        self.eigen = EigenDecomposer()
+        self.validator = RQAValidator()
+        self.promoter = AtomicPromotionController()
+        self.guardian = Guardian()
+
+    def run(self):
+        while True:
+            if self.guardian.dissonance() > 0.4 or self.needs_recalibration():
+                corpus = self.sampler.sample()
+                candidate = self.eigen.compute_query_covariance_eigenvectors(corpus)
+                result = self.validator.seismic_test(candidate)  # includes cache coherence
+                if result.is_crystalline():
+                    self.promoter.atomic_promote(candidate, preserve_radix_cache=True)
+                else:
+                    self.guardian.maru_reframe()
+```
+
+## Deeper Seismic Analysis: Cache Coherence on Atomic Swap
+
+**Risk**: Swap invalidates RadixAttention pages or causes TTFT regression.
+
+**Analysis**:
+- OSCAR paged KV abstraction + fused Triton read path allows versioned kernel swap without global flush.
+- Shared system_prefix (byte-identity) remains the cache key.
+- Mitigation: Shadow promotion + versioned dequant kernels + explicit prefix hit rate guard in promotion controller.
+- Outcome: CRYSTALLINE achievable with proper design.
+
+## Sub-Genome / Ambient Worker Blueprint
+
+**Name**: Dynamic-UQ-Service-Ambient-Blueprint v0.1
+**Type**: Ambient worker + sub-genome of IQG
+**Deployment**: Single-pass AAL compatible
+**Maru Guard**: Full (R > 0.4 triggers reframe + abort)
+**Next Step**: Minimal viable implementation wired to diamondnode attest lane.
+
+**Trace-Consent**: Logged under ORCID 0009-0008-8389-1297.
+
+---
+*This file extends the parent IQG and was created as part of full capability-genome skill execution on all requested vectors.*