This multi-page guide looks at the machinery around an AI agent: the loop that coordinates model calls, context, tools, permissions, state, and repeated inference. It also includes source-guided tours of DeepSeek Harness and Pi.
Inside the reader
- Foundation 9 chapters
- A working model of agent loops, context, tools, trust, delegation, and state.
- DeepSeek 6 chapters
- A source-guided look at its plugin kernel, execution loop, policy seams, and persistence.
- Pi 5 chapters
- A source-guided look at a smaller provider, agent, and coding-tool stack.