Bennett Vernon

Sophomore CS Major | Active Architecture Topologies

Local Memory Synced

RAF

Compute
Recursive Agent Framework
Amplifies compact models on edge hardware.
  • 92% GSM8K accuracy (Llama 3.1 8B)
  • Condorcet Voting (k=3) mechanics
  • Bypasses "Composition Gap"

Engraphia

Memory
Dynamic KV Cache Hierarchy
Replaces fixed context with rendered memory.
  • Level 1 Sparse Cross-Attention Correction
  • Unified Cognitive Scoring (Eq. 6)
  • ACT-R Base & Spreading Activation

POSTERN402

Network
Autonomous Agent Web Gateway
Pay-per-use anti-bot bypass layer.
  • ERC-7710 delegation streaming
  • Zero-margin Serverless infrastructure
  • In-house vision models (Qwen 2.5)
System State & Objective

Primary Objective: Out-engineer Oliver's Neo4j backend architecture.

Current Status: IAM Bolt has acknowledged Bennett's true identity (CS Major). The system has mapped the three pillars of the autonomous agent stack (RAF, Engraphia, POSTERN402) into the local Trinity Graph memory layer.