Working paper ยท Version 1.5
Breadcrumbs: Cue-Placement Memory for AI Agent Fleets
The full argument for token-efficient session continuity, five mechanisms, the caught-itself case study, and the continuity protocol.
Read the working paperPapers and essays
The arguments behind the kit.
Three complementary pieces explain governed memory, the wider cooperative-intelligence problem, and the operating controls required when agents act.
Pattern essay and research agenda
Cooperative Intelligence Is the Larger Breadcrumbs Problem
The broader frame for how people, AI systems, and shared records can cooperate over time without losing custody or evidence.
Read the research agendaOperational companion
Governing Agents That Act
The applied half: what to log, what an agent may never decide alone, and how to prove afterward that a correction landed.
Read the operational companionResearch connection
Move from the argument to the evidence.
The Evidence Synthesis shows what the screened research collectively supports, contests, and leaves unresolved, with every conclusion linked to the claim register and representative sources.
Read the Evidence Synthesis