Orphee Nessim

Canteen worker. Independent researcher on AI alignment and conceptual risk analysis, approaching LLM architecture without prior assumptions. Focus: epistemic uncertainty, structural arbitration as an alternative to scalar reward optimization, under-specified objectives. Current work: Typed Uncertainty Arbitration (V131) — developed through iterative human-AI critique with Grok, DeepSeek, Claude, GPT-4o, and Gemini as adversarial reviewers.