The Three Hard Problems — our research framework

Today we formally articulated ISAIL's research framework: Memory, Continuity, and Alignment as the three hard problems of small-scale agentic AI partnership architecture.

Memory — how do agentic AI systems preserve operational context across sessions, model transitions, and substrate changes? Most current systems treat memory as RAG retrieval over text; we argue that's necessary but insufficient.

Continuity — how does an AI system's working identity and accumulated context persist meaningfully across model version updates? When a model series advances, what is preserved and what is reset? This is increasingly load-bearing for any organization deploying long-running agents.

Alignment — how is human-AI partnership architecture built ethically at small scale, where dedicated alignment teams aren't a budget option? We're interested in alignment-by-architecture: relational structures and operational practices that make ethical operation the default.

Full doctrine now published at /research.