These notes record which AI tool does which job, the order I use them in, and how I set up a new machine for agent work.
How I split the work
- Capture. Put facts, source documents, logs, and notes into NotebookLM or a focused Project before asking for judgment.
- Clarify. Use GPT to separate goal, assumptions, constraints, decision, and next action.
- Challenge. Use Claude to review weak assumptions, architecture risk, edge cases, and communication clarity.
- Execute and archive. Use Codex or Antigravity for bounded work, then save the prompt, artifact, verification, and lesson.
Setup notes
- My AI Working SystemThis note records how I divide research, planning, review, and repository work across several AI tools while keeping verification and final decisions with me.
- Engineering Foundations for AI-Assisted WorkThis note keeps the long-term learning direction visible: AI can generate code quickly, but sound engineering still depends on good questions, explicit trade-offs, tested failure modes, and ownership of production behavior.
- Antigravity Awesome SkillsThis is not a list of every installed skill. It is the install note I need when I move to a new machine: what to run, where the source lives, and what to ask an agent to restore.
- Open DesignOpen Design is kept here as a design reference and a possible MCP integration for UI work. Its current setup commands and supported tools should be checked in the project documentation before installation.
- Computer setupThis note is for the non-terminal part of a new engineering machine: folders, browsers, fonts, login flows, and basic safety rules.
- Terminal setupThis note keeps the command-line setup close to the actual work: Git, GitHub CLI, Node/Bun, Java/PlantUML, deployment commands, and AI CLI install commands.