I work independently on AI governance and system design, with a focus on how large language models behave under ambiguity, pressure, and extended real-world use. My current work centers on E5, a constraint-first architecture developed through repeated empirical iteration in response to observed failures such as hallucinated policy, overconfidence, and authority drift.

I come to this work from practical use rather than theory. After encountering recurring governance failures while trying to rely on LLMs for real planning and professional tasks, I began rebuilding my approach around explicit constraints, refusal behavior, and fail-closed design. I prioritize systems that know when to stop, narrow, or defer rather than confidently proceed.

I support my family through hourly work and conduct this research independently outside of formal institutional backing. I’m seeking time-limited support to document, validate, and responsibly prepare this work for broader scrutiny without weakening its safety properties.