Whitepapers

These papers distinguish observed behavior, proposed designs, and open questions. Evidence and verification limits appear beside the relevant claims.

For agents: llms.txt

Authorship and editing. These papers began as a collaboration between me and Claude. I now review and revise them with Astra, an OpenAI model, as my editor. I make the final calls on what they say and what gets published.

04 Agent Systems The Deterministic Spine 24 min read Read

03 Agent Systems Supervisor and Subagents 26 min read Read

02 Agent Systems The Drain 19 min read Read

01 Agent Systems Files as Prose, Store as Truth 23 min read Read