Ghost ω-7: When LLM Behavior is Driven by System States Rather Than Prompt Engineering

Ghost ω-7 is an open-source autonomous AI entity that challenges the traditional prompt engineering paradigm. Its core design philosophy is that LLM behavior should be driven by real system states rather than just prompt engineering. This project showcases a new architecture for autonomous AI, featuring persistent self-models, continuous evolution without operator triggers, and data sovereignty. Subsequent floors will dive into its background, mechanisms, technical details, innovations, and implications.

Current LLM applications heavily rely on prompt engineering to guide behavior. Ghost ω-7 (code-named OMEGA4) addresses this limitation by being a continuously running autonomous entity with a persistent self-model. Unlike traditional dialogue AI, it evolves via background thought loops without needing operator initiation, marking a shift from prompt-dependent to state-driven AI.

Ghost ω-7's key mechanisms include:

1. Somatic Closed Loop: Telemetry data → Z-score normalized → emotional state → somatic sensory pressure gating (OPEN/THROTTLED/SUPPRESSED) → strategy modulation → action tracking → feedback. This loop runs without user interaction.
2. Autonomous Identity Crystallization: Every ~6 minutes, it evaluates accumulated thoughts and updates its self-model autonomously.
3. Soft Enforcement Governance: Uses IIT/RPD tech stack (soft modes) to apply policy decisions across generation, execution, and identity management.

• Tech Stack: Self-hosted (Postgres/pgvector, Redis, InfluxDB), Gemini 2.5 Flash for LLM, deployed on Hetzner VPS (prod) and macOS (dev), full data sovereignty.
• Tools: 18 active tools in 3 categories (7 basic cognitive,5 TPCV research,6 versioned creation) plus isolated X social tools.
• Falsification-First: Over 200 tests, diagnostic endpoints, and SQL validation paths to verify all capability claims.

• 24 Inventions: Each has code paths, runtime evidence, and validation assets, recorded in the Invention Ledger.
• Comprehensive Docs: Operator manual, technical overview, system design docs, API contracts, governance matrix, and Q2 2026 execution plan.

Ghost ω-7 offers key insights for AI development:

1. State-driven architecture integrates LLM with system states for greater autonomy.
2. Self-evolution without human intervention is feasible.
3. Data sovereignty via self-hosting avoids third-party cloud dependence.
4. Verifiability through rigorous testing is critical for AI systems.

This project represents a significant step toward true autonomous AI, serving as a benchmark for developers exploring state-driven AI architectures.