Steve Waddington ONET.COM.AU

WRITING / PAPERS

Papers

Versioned technical reports published under my own name. Each paper states its methods, limitations, evidence boundary and review status.

OCT 2026 VERSION 1.0 Human in Command: Governability Before Review in a Growing Agent Skills Library Approval gates, agent self-review and wholesale removal each failed to keep a human in control of a growing agent skills library. What worked was making the library understandable first. Self-published exploratory technical report; not peer reviewed → SEP 2026 VERSION 1.0 Corrections Consolidate: Why Every Agent Mistake Should Not Create Another Rule When an AI agent fails, adding another permanent instruction can make its governance worse. A better correction starts by deciding whether the rule was defective or the agent simply failed to follow it. Self-published exploratory technical report; not peer reviewed → SEP 2026 VERSION 1.0 Reducing Context Before Inference: An Exploratory Comparison of Full History, Deterministic Selection and Model Curation An exploratory comparison of full history, deterministic context selection and model curation across 23 checkpoints from six natural human-agent trajectories. Self-published exploratory technical report; not peer reviewed →