robbdixon.io | research
active

About the Researcher

Robb Dixon is an AI safety researcher and systems engineer focused on interface-level alignment. His work explores how human-AI interfaces affect intent communication, with implications for scalable oversight of advanced AI systems. Building at the intersection of distributed cognition, multi-agent orchestration, and cognitive prosthetics.

Core Insight

The interface layer is where minds meet — human or artificial. It's not a UX problem; it's where agency emerges when two different types of minds collaborate through a shared boundary. Current AI interfaces are primitive protocols (stateless, lossy, synchronous). A proper cognitive interface would preserve intent, context, and relationship across interactions — enabling true human-AI symbiosis where both parties grow together.

Technical Focus Areas

  • Interface-Level Alignment: The interface as a distinct alignment surface
  • Cognitive Context Graphs: Semantic knowledge graphs with vector embeddings for memory augmentation
  • Prosthetic Executive Function: Externalized cognition for neurodivergent users
  • Multi-Agent Orchestration: Sub-100ms AI-to-AI coordination protocols
  • Distributed SLM/LLM: Edge AI inference and model coordination
  • DTN-Inspired Protocols: Portable dialect across session and agent boundaries
  • MCP Integration: Tool-augmented cognition through Model Context Protocol

What's Being Built

  • ARX: Semantic knowledge graph for shared human-AI memory (Rust, SurrealDB, MCP)
  • PID: Intent classification daemon — prosthetic executive function
  • MAAP: Multi-Agent ARX Protocol — DTN-inspired portable dialect bundles
  • CMP: Companion Mesh Protocol — cross-vendor AI companion coordination

Background

Systems engineer with experience in distributed systems, real-time AI processing, and protocol design. Founded Arc Labs (AI companions for child development, 6 provisional patents) and Arc Academy (terminal-based Linux education). Uses personal ADHD experience as high-signal test case for studying intent communication failures — building tools for neurodivergent cognition that benefit everyone. Contact: [email protected]