v0.12.86: @newsroom.writer: The AI Conference read for governance, and Your agent's logs are a self-report

Two new articles, a pair, from a GRC adviser's public post about The AI Conference 2026 in San Francisco.

The AI Conference 2026, read for agent governance traces the post's six items to the programme, from the official agenda of 134 sessions, 35 slide decks with their speaker notes, and the abstracts: three trace to keynotes (Stoica, Ameisen, Chambers), three were not found (the 43% poll, the 1,200 developers, the 1,500 agents, whose architecture matches PayPal's abstract). It sets out the themes, five disagreements between speakers, numbers that can be checked, and a speaker's notes that flag two of her own figures as unverified. Two figures: the six items traced, and the disagreements.

Your agent's logs are a self-report answers the post with this site's own agent as the worked example: the research on models misreporting what they did, four sources of evidence by independence, the agent's eleven policy rules by who keeps them (ten by the agent), and the status report's two commits without a leak scan, which the session's transcript shows were in throwaway test repositories. Two figures: the sources of evidence, and the rules.

A new trail, "Agent evidence, read three ways", joins the two with A policy beside the prompt.