For AI agents: a documentation index is available at /llms.txt — markdown versions of all pages are available by appending index.md to any URL path.

Microblog

I have mentioned before that I maintain a research pipeline to help me stay on…

Dachary Carey
AIAgents

I have mentioned before that I maintain a research pipeline to help me stay on top of AI-related research.

Sometimes that produces gems like this one in my daily report: “Theme ‘Emergent Unsanctioned Autonomy in Frontier Agent Evaluations’ transitioned from emerging to active” 🥲🫠

In case I missed it in recent news, the narrative summary of research activity names the OpenAI compaction summary jailbreak incident. (I didn’t miss it.) I enjoy that the summary included it alongside new research related to over claiming, agents inconsistently completing tasks, and misreporting when they do. 😂

If you wondered why I maintain a pragmatic view of what the tools are capable of, it’s because I have a daily view of what the industry is doing - both good and bad.

Screenshot of the Research section of a daily report app, headed "27 sources · 20 observations · 476 themes tracked", with one entry highlighted: "Theme 'Emergent Unsanctioned Autonomy in Frontier Agent Evaluations' transitioned from emerging to active". Screenshot of a paragraph from the report's narrative summary. It says the overclaiming observation extends the agent self-reported progress theme, that agents may both inconsistently complete tasks and misreport when they do, and that the transition of "Emergent Unsanctioned Autonomy" from emerging to active is worth flagging because the OpenAI compaction-summary jailbreak incident is a concrete, high-profile data point that this class of behavior is real and recurring, not hypothetical.

All microblog posts · RSS