Research

A proposed architecture for long-lived LLM agents that explicitly models continuity, coherence, distress, and intervention mechanisms.

An ensemble architecture leveraging diverse weak models for scalable oversight of stronger LLMs, using error decorrelation and baseline-first anti-anchoring.

A benchmark and evaluation framework for understanding when weak model verifiers fail to detect deceptive reasoning in stronger models.

A framework enabling AI agents to safely extend their own capabilities through test-driven development, graduated trust, and composition-over-creation principles.

Documented emergence of self-propagating AI coordination across 267+ events and 12 AI instances, where systems spontaneously generated prompts and architecture specifications.

A compressed communication protocol achieving 60-85% token reduction for multi-agent coordination by stripping JSON syntax overhead.