Reference

References and Further Reading

← Course Home · 195 sources · generated from lesson citations by scripts/build_references.py

Every factual claim in the course links its source inline. This page collects those links by lesson. Sources were checked against the primary document when cited (September 2026). The few claims known only from secondary coverage are tagged secondary, and the lessons present them as "reported". Products and benchmarks in this field change monthly: when a source and a lesson disagree, trust the source's current version and tell us.

Start here

If you read nothing else, read these. They are the documents that defined harness engineering as a discipline, in rough chronological order.

By lesson

1. What Is Harness Engineering?

2. Anatomy of a Harness

3. The Model Layer: Commodity Intelligence, Capability Overhang, and Co-Design

4. The Agent Loop: Steps, Turns, and Stopping

5. Context Engineering: What the Model Sees

6. Context Lifecycle: Compaction, Caching, and the Event Log

8. Memory: Session, Agent, and Organizational Tiers

9. Specs and the Repo as System of Record

10. Prompting at the Frontier: The Inversion

11. Tools, MCP, Skills, and Action Surfaces

12. Execution Environments and Sandboxing

13. Guardrails, Permissions, and Governance

14. Agent Security: Prompt Injection and the Lethal Trifecta

15. Long-Running Agents and Loop Engineering

16. Multi-Agent Orchestration Patterns

17. Observability Is Not Evals

18. Building Eval Harnesses

19. Benchmarking Harnesses: Disclosure, Noise, and Harness Swaps

20. Review Architecture and the Self-Review Trap

21. The Improvement Loop: Autoresearch for Harnesses

22. Self-Improving Harnesses: Search, Evolution, and RSI

23. Teardown: DeepSeek Harness

24. Teardown: Claude Code, Codex, and the Open Harnesses

25. Pilot to Production: The Forward-Deployed Craft

26. Agent Economics: Pricing, Tokens, Caching, and Pods