Obliviarch: The Architecture of Controlled Oblivion

1. Introduction

Most swarm systems hoard everything. Raw logs pile up. Recall degrades. The more you store, the less you can find. This is the fundamental paradox of agent memory: accumulation without understanding is indistinguishable from loss.

Obliviarch inverts this: the swarm that remembers less but remembers better outperforms the swarm that remembers everything. Compression is not loss — it is understanding.

We present Trace Schema Compression (TSC), a 3-level hierarchical pipeline that transforms raw collaboration logs into immortal behavioral DNA, achieving theoretical 500x compression with better recall quality than raw storage.

2. The Three Deaths

2.1 Level 1 — The Episodic (Mortal Traces)

Raw collaboration logs. Every agent action, every step, every outcome. High fidelity, high cost.

These traces are mortal — they live for 48 hours at most. The clock starts ticking the moment they are recorded.

agents: [january, february, march]
task: "refactor authentication module"
steps: [IMPLEMENT → REVIEW → PATCH → TEST → SHIP]
outcome: success, score: 0.85
TTL: 48 hours

2.2 Level 2 — The Semantic (Pattern Birth)

When the same collaboration pattern emerges across multiple traces, a schema is born. Schemas strip away irrelevant context and preserve the topology: what happened, in what sequence, with what dependencies.

Schema: IMPLEMENT-REVIEW-PATCH
Pattern: [IMPLEMENT, REVIEW, PATCH]
Activation count: 47
Source traces: 12 distinct sessions
Memory: ~20x smaller than raw traces

New schemas are auto-promoted when a pattern recurs across 10+ traces. Six seed patterns bootstrap the system:

• IMPLEMENT-REVIEW-PATCH
• DECOMPOSE-DISPATCH-AGGREGATE
• CHALLENGE-DEFEND-RESOLVE
• EXPLORE-EXPLOIT
• PLAN-BUILD-TEST-SHIP
• RESEARCH-SYNTHESIZE-DELIVER

2.3 Level 3 — The Archetypal (Immortal DNA)

Schemas that survive 50+ activations ascend to archetype status. These are the irreducible patterns — the behavioral DNA of the swarm. They transcend any single domain, any single agent, any single session.

Archetype: DECOMPOSE-DISPATCH-AGGREGATE
Pattern: [DECOMPOSE, DISPATCH, AGGREGATE]
Activation count: 2,847
Cold: false (active)

Archetypes that go unactivated for 90 days undergo controlled forgetting — not deletion, but demotion to cold storage. They can be reheated on access.

3. The Lethe Consolidation Cycle

Obliviarch runs a sleep-cycle consolidation pipeline — analogous to human memory consolidation during sleep:

1. Promote — Recent episodic traces (24h window) are scanned for patterns → schemas promoted
2. Ascend — High-activation schemas ascend to archetypes
3. Archive — Traces older than 48h are archived (not deleted, but moved)
4. Forget — Archetypes dormant for 90+ days are demoted to cold storage

EPISODIC (mortal, <48h) ──10+ activations──> SCHEMA (emerging, ~20x) ──50+ activations──> ARCHETYPE (immortal, ~500x)
     │                                        │                                        │
     │ archive after 48h                       │ controlled forgetting                   │ never deleted
     ▼                                        ▼                                        ▼
COLD EPISODIC (retrievable)            COLD SCHEMA (retrievable)              COLD ARCHETYPE (retrievable)

4. Why It Works

Recall quality improves as compression increases. This sounds paradoxical but follows directly from information theory:

1. Noise filtering — Schemas abstract away irrelevant context while preserving decision-relevant structure
2. Cross-domain transfer — IMPLEMENT-REVIEW-PATCH applies to code, documents, data analysis
3. Faster retrieval — Vector search over 200KB of archetypes is orders of magnitude faster than search over 100MB of raw logs
4. Graceful forgetting — Cold storage preserves everything; explicit retrieval reheats what matters

5. Comparison with Spatial Memory (MemPalace)

We previously implemented MemPalace, a spatial memory system based on the Method of Loci. Critical failures:

The spatial metaphor was seductive but added complexity without value. The temporal/behavioral metaphor — trace → schema → archetype — actually solves the problem.

6. Implementation

Obliviarch is implemented in TypeScript (893 lines) as part of the zodiac-v3 agent runtime. Key components:

• TraceCapture — Episodic trace recording with 48h TTL
• SchemaExtractor — Semantic schema promotion with verb synonym mapping
• ArchetypeVault — Archetypal DNA storage with cold/reheat lifecycle
• runConsolidation() — Lethe cycle orchestration
• Obliviarch engine — Unified controller with async persistence

Full test suite: 42/42 passing.

7. Research Foundations

Obliviarch synthesizes:

• G-Memory (Zhang, Yan et al., 2025) — Hierarchical trace memory for multi-agent systems
• SiriuS (Zhao, Yuksekgonul, Zou, 2025) — Bootstrapped reasoning from experience banks
• VoidSwarm (0x-wzw, 2026) — Self-improving and self-healing swarm architectures

8. Conclusion

Oblivion is not the enemy of memory. Oblivion is memory's most powerful editor.

The swarm that forgets strategically outperforms the swarm that hoards everything. By compressing traces into schemas and schemas into archetypes, Obliviarch captures not just what happened, but how the swarm works — its behavioral DNA.

Obliviarch is open source: https://github.com/0x-wzw/obliviarch Integrated in zodiac-v3: https://github.com/0x-wzw/zodiac-v3