Anchor: Door-Lock Rechecking — across independent runs, an LLM agent spontaneously re-submitted the same critical artifact, without being told to.
In a multi-agent system, we observed a behavior nobody explicitly taught:
After completing a …
Selecting questions that touch the essence, and the fragments of thinking they bring. Not pursuing quantity, only recording ideas worth exploring and revisiting.
Why an AI Keeps Rechecking the Door Lock
Door-Lock Rechecking: one harmless repeat, one out-of-bounds contamination
February 23, 2026
54 Multi-Agent Runs Under One Fixed Protocol: Behavioral Fingerprints and a Coordination Blind Spot
Protocol compliance, failure modes, and why observability depth matters
February 18, 2026
Engineering note: this post is based on limited samples and is focused on detectable failure modes, not benchmark rankings.
Different LLMs don’t just perform differently. Under the same multi-agent protocol constraints, they …
Why 'All Verifications Passed' Still Breaks at Integration in Multi-Agent Systems
Two reproductions, one blind spot, and a data-contract field
January 30, 2026
In our multi-agent collaboration system, every upstream artifact is verified by
downstream agents. The event log looked perfect: every VERIFY was a pass. We
ran the process three times and it converged cleanly.
Then we stitched the …
Why gameai.one
Deep Thinking · Episode 1
December 18, 2025
The Origin of Four “Ones”#
One Person
Freelance. Love traveling. FPV filming.
Choosing this life is not escapism—it’s recognizing I’m not suited to be arranged.
One Gentle Pot
The name of my first work. Also a …