Research
Research on agentic AI, evaluation, persistent change, and the evidence and authority structures required for trustworthy autonomy.
Research thesis
Persistent agent change creates a governance problem: who controls the evidence, who judges the result, and who authorizes action?
Systems that propose changes should not automatically gain authority to select their evidence, certify success, erase contrary outcomes, or authorize external action. The research program therefore treats evidence lineage, adjudication, policy, and action authority as distinct governance surfaces.
Current program
A two-phase research program investigating how persistent agent change can be governed through evidence lineage, deterministic advisory policy, append-only human adjudication, and explicit separation between recommendation and external-action authority.
Agentic AI · Trustworthy AI · AI Evaluation · Governable Autonomy
Phase I is anchored exactly to v0.10.0. Phase II begins from v0.11 or an explicitly versioned successor and remains future work.
Research areas