Project Argus is an autonomous AI platform built on a stateful, high-availability cognitive architecture designed to manage long-term context, temporal patterns, and deep workspace state tracking.

• Layer 0 — Core Execution Bedrock (libargus.cc): An in-process, zero-allocation native AI runtime utilizing OpenJDK Project Panama (FFM API) to bind directly to unmanaged C++ tensor compute graphs. It provides the performant, low-latency foundation of the platform by executing inference directly within the JVM process space.
• Layer 1 — Stateful Cognitive Shell (L-TABB): [In Active Development] An upstream memory orchestration framework that implements stateful working memory caches, spatial thread lane isolation, and continuous background memory consolidation routines.
• Stateful Context Optimization: [In Active Development] An advanced memory-management layer designed to coordinate key-value (KV) states between system memory and accelerator VRAM, minimizing redundant processing overhead across multi-turn and multi-agent operations.