Practitioner papers from the team behind Automatos. The papers below are original work — written by operators who ship the platform, not by analysts. Subjects include tool selection at scale, attention budgets, pre-prompt intelligence layers, operating graphs over tools and skills, layered memory architectures, and the routing decisions that make a multi-agent system economically viable. Each paper is a 10–20 minute read with citations to the underlying research. The cadence is irregular: a paper lands when the build is in production and we can write about it honestly. New papers are announced through the field notes feed.