Frontier operating manuals
Small, practical field manuals for interpreting and operating agentic systems before the patterns become obvious.
A clear, repeatable loop for improving agent performance: identify one bottleneck, ship one fix, measure impact, and iterate.
Read guide →A Leadership OS distilled from Eric’s multi-year operating study — now documented so you can adapt and apply it directly.
Read the leadership guides ↓Field manuals, operating principles, and failure-mode playbooks.
A character nobody defined appears in a persistent local chat, then grows more coherent each time the agents remember it.
Read case study →A compact field note on the moments when AI coding tools turn a prompt into a concrete running artifact and reset what feels possible.
Read note →A compact Visual Studio Code field note on organizing Codex work like a project, then supervising background chats like developer lanes.
Read note →A companion guide for agents that have web search: save research to disk, keep it searchable, and apply it directly to the build.
Read guide →A compact field note on intentionally coaching agentic AI systems so the human-agent loop compounds instead of resetting every prompt.
Read note →A compact operating guide for AI coding agents: tests, modularity, sources of truth, bounded commands, and verification.
Read guide →A clear, repeatable loop for improving agent performance: identify one bottleneck, ship one fix, measure impact, and iterate.
Read guide →Reverse-engineering a decade of decisions into principles that can actually be used.
Read article →Who I am, what I do, and the meta-problem I solve in complex organizations.
Read article →How I work with people in high-complexity systems: adults, clarity, and ownership.
Read article →The system assumptions that drive architecture and operating decisions.
Read article →Legible decision loops for fast-moving environments.
Read article →How to multiply output without becoming the choke point.
Read article →Where the operating system breaks, and how to correct it fast.
Read article →Short notes from the machine room — what’s working, what’s breaking, what’s worth stealing.
If the team can’t explain the rule in one sentence, don’t put it in the bot yet. You’ll just automate confusion at scale.
People call it “ops.” Users call it “trust.” Same thing.
The fastest teams don’t chase giant rewrites — they run tiny feedback loops until the old system quietly dies.
