5 layers ยท 13 capabilities ยท 7 sovereignty bindings
| Layer | Name | Latency |
|---|---|---|
| L1 | Care-Floor (pre-call gate, split-conformal) | <5ms |
| L2 | OWEM Routing (5 specialist groups) | <50ms |
| L3 | Brain Stack (4 paths: small/medium/large/sovereign) | <200ms |
| L4 | Topology (5 architectures) | <2s |
| L5 | Fluid Pyramid (12 layers + capstone) | live |
Endpoint: GET /api/sovereign-stack
| Layer | Retention | Capacity | Use |
|---|---|---|---|
| L1 Episodic | 0-7d | 10,000 | Recent context |
| L2 Semantic | 7-90d | 50,000 | Pattern recognition |
| L3 Sovereign | PERMANENT | unlimited | Audit trail, BFT records |
| L4 Replay | 0-30d | 1,000 | Continual learning |
Endpoint: GET /api/memory/lifecycle
Next-gen brain: Qwen3-1.5B (vs current 0.6B) + LoRA rank=64 (vs current 32) + 8-stage training pipeline
| Metric | Current | Target |
|---|---|---|
| Base model | Qwen3-0.6B | Qwen3-1.5B |
| Total params | 605M | 1.5B |
| LoRA rank | 32 | 64 |
| Trainable params | 9.18M | ~30M |
| Training samples | 200 | 2,000 |
| Loss reduction | 57.7% | 80% |
| Compute | 120s MPS | 30.6 GPU-hr Kaggle T4 |
NEW prize pool: $1.075M (3 new comps). Total Kaggle: $2.52M (was $1.45M).
| Competition | Prize | SOV33 entry | Advantage |
|---|---|---|---|
| ARC-AGI 2025 | $1,000,000 | sov33_arc_owem + JEPA | world model for abstract reasoning |
| MATH 2026 | $50,000 | sov33_math_owem + 1.5B | CoT + self-consistency + BFT-33 |
| TruthfulQA | $25,000 | sovereign brain + Honor | 12 Pillars - truth |
16 layered capability tests ยท 8 sovereignty checks ยท 4 performance metrics
| Test | Status |
|---|---|
| Fluid Pyramid | โ GREEN |
| Inner OWEMs | โ GREEN |
| Conformal Veto | โ GREEN |
| Alphabet Stages | โ GREEN |
| Stats / Capabilities / Charter | โ GREEN |
| Security Audit | โ GREEN |
| Text Chat | โ GREEN |
| Fluid Pyramid Layer 1 + 12 | โ GREEN |
| Sovereign Brain v2 Status | โ GREEN |
| 12 Around 1 | โ GREEN |
| OWEM Fast (live sovereign brain) | โณ ~5-20s first call |
| Master (all 17 capabilities) | โ GREEN |
| Artifact | Purpose | Lines |
|---|---|---|
sov33-llms.txt | AI-discoverable identity (LLMs read this) | ~110 |
sov33-charter.md | Formal 12 Pillars + Article 0 + BFT-33 | ~230 |
sov33-sigil-format.md | Ed25519 SIGIL signing spec | ~220 |
sov33-owem-card.json | A2A agent card (MCP discoverable) | 2.5KB |
sov33-press-release.md | Formal launch announcement | ~140 |