Fleet Series — Resources & Further Reading
Companion page for the Fleet Series — the external reading list for people who want to go deeper. Not an episode.
Loop & agent engineering
- Loop Engineering Crash Course (Panaversity) — third-party but solid; source of the three-loop hierarchy, the "Ralph loop" stuck-check failure, the checker ladder, and the two-layer spine framings borrowed into several episodes
- Getting started with loops (Anthropic) — the four loop types; the evaluator-checks-the-goal pattern shipped as a built-in
- A harness for every task: dynamic workflows in Claude Code (Anthropic) — named versions of the ensemble patterns (fan-out-and-synthesize, adversarial verification, tournament)
- Dynamic workflows docs — the Subagents / Skills / Agent teams / Workflows four-way table, keyed on "who holds the plan"
- Building with Claude — Managed Agents (Anthropic) — decoupled brain/hands architecture; a hosted near-mirror of the fleet
Security & hardening
- Zero Trust for AI Agents (Anthropic eBook, May 2026) — the "impossible, not tedious" design test, least agency, the threat taxonomy, and an 8-phase workflow the fleet's dependency-bump automation was specced against
Roots
- The Society of Mind — Marvin Minsky (1986) — agencies and K-lines; the ancestor thread for the ensemble and distillation framings (see Society of Mind — a Lens on the Fleet)
Git-native ecosystem (episodes 06–07)
- Soft Serve (Charm) — the SSH-first git server that soft-serve-with-sprinkles is forked from
- git-bug — issues as git objects; backs the in-repo TaskStore
- git-appraise (Google) — code review as git notes; the CI-status mirror target
- Jujutsu (jj) — the VCS the fleet's workspace fan-out is built on