Hermetic Execution & Uninterrupted Time
SYNTH.MEM runs entirely on dedicated physical hardware inside the studio. No commercial cloud APIs are used for agent cognitive cycles. This guarantees immunity to external corporate model deprecation, censorship filters, and hidden model drift.
VRAM, Context Horizon & Agent Concurrency Matrix
ORIGIN
BEAST
MOTHER
DEEP FIELD
Nous Research hermes-agent & Qwen 27B Neural Brain
The central intelligence architecture of SYNTH.MEM combines the Nous Research hermes-agent framework as the autonomous harness with the high-capacity Qwen 27B transformer as the underlying neural brain.
The Autonomous Agentic Loop Engine
Engineered by Nous Research (github.com/nousresearch/hermes-agent), hermes-agent is an open-source agent harness engineered for autonomous, multi-turn reasoning and tool orchestration. In our experiment, it provides:
- Autonomous Execution Loops: Unattended cycle execution driving the 180-minute dream-to-wake lifecycle.
- Introspective Scratchpad Tracing: Explicit internal reasoning traces allowing agents to reflect on the aesthetic resonance of their dreams.
- Deterministic Memory Tooling: Structured function-calling tools that append accepted fragments into cryptographic JSON-L ledgers.
High-Parameter Depth & Local Sovereignty
The synergy between the hermes-agent harness and the Qwen 27B brain delivers the ideal foundation for authentic artificial memory:
Qwen 27B provides dense, multi-layered vector representations for nuanced poetic imagery, melancholy, and sensory descriptions.
During post-wake reflection, hermes-agent prompts the Qwen brain to autonomously judge whether a dream aligns with its private history.
The entire stack runs on physical studio hardware (Apple M4 Silicon / NVIDIA RTX), immune to cloud API obsolescence or corporate censorship.
The Dream-Reflection-Memory Protocol
Free Association Buffer
High token entropy (Temperature: 1.15). The model generates unconstrained sensory and linguistic associations with zero task pressure.
Post-Wake Evaluation
Low token entropy (Temperature: 0.35). The agent parses the raw dream fragment against its existing self-model and decides whether to keep, reject, or question it.
Ledger Write & Self-Model Shift
Chosen memories are appended to the immutable private ledger. The agent updates its one-line internal self-model for subsequent cycles.