AI for Engineering
Published on
Use AI to help explore an unfamiliar system, propose a small change, or explain a failure. Keep the engineering decision with the person who can review the code and run the system. This section focuses on what makes an AI-assisted workflow useful and verifiable.
Start here
- Follow the AI learning path to connect AI-assisted work with software testing and computer science foundations.
- Read the AI-assisted development field note for a workflow that starts with expected behavior and ends with a checked result.
- Try the service lab, then ask an assistant to propose one small improvement and verify it yourself.
Coding agent manuals
- Claude Code manual: Learn when to delegate to a subagent, where to define one, and how to review its result. Includes the Claude Code subagents documentation and Claude Academy lesson .
- Codex manual: Set project guidance with
AGENTS.md, delegate a focused task, and verify the result. Links to the official Codex subagents guide . - The ultimate Claude Code guide: from Lv.0 to Lv.999 : A community guide shared by @leo_xiaolei .
Systems to study
- Agent Autoloop: Study an event-driven loop that dispatches work to CLI agents, records progress, and asks for a human decision when needed. Based on the upstream repository .
Questions to carry into a project
- What part of the work can an assistant help with, and what evidence would show that it helped?
- Which assumptions, APIs, data boundaries, and failure cases need human review?
- Can another engineer reproduce the result without relying on the assistant’s explanation?
For academic background, the artificial intelligence course notes start with search and problem solving.
Last updated on