A personal research project. Implementation, controlled observations and published evidence remain separately scoped.
Elora Taurus Research, with a human perspective.
Inside the research
Questions we explore.
A research engine built on non-neural machine learning.
Elora is a custom Python research engine that uses non-neural machine learning to detect AI behaviour patterns while governing execution through a deterministic proposal-to-commit boundary and surfacing structured operational decisions.
Detected behaviour signals are scored, linked to policy controls, and captured as structured evidence so allow or block outcomes can be replayed deterministically without regenerating model output.
Threat Cases connect detections to explainable Governance outcomes.
Evidence-backed findings and Case Reports connect attack evidence, recommendations, authorization context, final actions, and Governance decisions without giving Threat Intelligence execution authority.
Elora Decisions turns runtime events into reviewable decision history.
Operational outcomes are grouped into bounded decision stories so preparation, intervention, completion, cancellation, and cleanup can be reviewed as one structured record instead of scattered notifications.
Connected capabilities
Observe. Govern. Understand.
01 / Observe
Runtime Intelligence
Study runtime pressure, behaviour patterns and constrained-compute performance using bounded non-neural and statistical methods.
Explore the research engine and bounded Agent integration: delegated access, controls, task evidence and retained-history investigation. Implementation and deployed acceptance remain distinct.
Explore behaviour research, memory delivery and role-aware selection experiments, with separate delivery, format and content measurements and explicit evidence limits.
