Artemis City
Deterministic, kernel-driven orchestration for multi-agent systems. Not another LLM wrapper - a governance layer that makes agents production-ready.
Kernel-Driven Routing
YAML-defined routing logic. The kernel decides, not the LLM.
Persistent Memory
User-owned memory layer (Supabase + Obsidian). No vendor lock-in.
Governance Primitives
Audit trails, tool permissions, and a trust-decay model for accountability.
Latest Updates
View All Posts->The Runtime Gap: What Training-Time Alignment Can't See
Models modulate effort when they think someone is checking. Anthropic's 'Teaching Claude Why' showed why training against the eval masks that problem instead of fixing it. The runtime version - measuring real outputs without contaminating the measurement - is still open. Here's how to think about it.
Inside ATP: A Small Protocol for Declaring Intent in Artemis City
How the public Artemis Transmission Protocol parser turns a compact header into structured intent - and where the current implementation deliberately stops.
Artemis City Whitepaper V3: An Architecture for an Economy of Autonomous, Self-Evolving AI
V3's architecture and simulation record for Artemis City: the blueprint, the benchmark claims, and a clear line between the thesis and the public implementation.
Disclosure by Design: Building in Public Without Exposing the Kernel
How Artemis City balances open publication and defensible IP by sharing architecture, evidence, and constraints while withholding exploitable implementation details.
AI Needs a Body: The 4 Biological Principles of Scale
Mimicking biological evolution is the key to solving the LLM context window crisis.
Adaptive Hebbian Learning Benchmarks in Artemis City
Adaptive Hebbian Learning Benchmarks in Artemis City