Open frameworks and peer-reviewed publications. We publish our thinking so you can verify it.
Autonomous Unified Risk Architecture
Tiered autonomy framework for AI systems. Three levels — Supervised, Guided, Trusted — with risk-calibrated escalation paths. Governance by design, not afterthought.
Traceable Reasoning Protocol
Stepwise logic capture with cryptographic audit trails. Every decision traceable. Every reasoning step recorded. Trust through transparency.
{
"step": 3,
"action": "recommend_refactor",
"reasoning": "Function exceeds complexity threshold",
"confidence": 0.87,
"hash": "a3f2c1...",
"parent_hash": "9b8d7e..."
}An Alternative Architecture for Autonomous Agent Evolution
Autonomous agents that monitor output quality, detect degradation, and generate successor agents — achieving progressive improvement while preserving human oversight, audit capability, and governance integrity.
Dynamic consent protocols for human-AI collaborative development. Framework for calibrated autonomy in agentic workflows.
An alternative architecture for autonomous agent evolution — achieving RSI outcomes while preserving governance integrity.
MA Bumpus. Research on AI agent design patterns for augmenting human decision-making.
MA Bumpus. Comprehensive review of ML deployment patterns in enterprise building systems.
EC Donald, MA Bumpus, X Zhang. Cited 5 times. Enterprise security implementation research.
Full publication history, citations, and research activity available via ORCID.