Field notes and technical insights from the frontier of production AI systems.
Why transcripts cannot safely coordinate agent workflows, and how explicit state, version checks, and durable checkpoints close the gap.
A practical contract for agent handoffs: ownership, task, context, authority, and outcomes, with telecom and travel examples.
Your agent is only as good as what it sees. A practical tour of context assembly: memory, retrieval, token budgets, and the pipeline that joins them.
Single agent first; escalate only on evidence. A decision procedure for topology, specialist boundaries, tool classification, and cross-cutting concerns.
Every agent has an orchestration layer deciding what happens next. The five questions it must answer, and when one agent stops being enough.
Running the decomposition framework on ecommerce support, and landing on one agent. When the answer flips to multi-agent, and where the framework strained.