A 12-week, weekend-based training plan for product managers who want to go beyond prompting AI coding tools and actually operate in a codebase.
Product managers who:
- Have some CS background (even if it was years ago)
- Are expected to fix contained bugs and ship POC features using AI coding tools
- Want to stop being "just a prompter" and start understanding what the AI is actually doing
- Need to credibly hand off code to engineers with real context
The full curriculum is in plan.md. It's structured as:
| Phase | Weeks | Focus |
|---|---|---|
| 1. Environment & Navigation | 1–3 | Terminal, git, reading code, how services work |
| 2. AI-Assisted Development | 4–7 | Claude Code mastery, Codex mastery, first bug fix, first feature |
| 3. Production Readiness | 8–10 | Code review literacy, testing & CI/CD, multi-language comfort |
| 4. Capstone & Reflection | 11–12 | Build something real, document your journey |
Time commitment: 4 hours/week (2 hrs Saturday, 2 hrs Sunday). 48 hours total.
Tools: Claude Code + OpenAI Codex. VS Code as IDE.
Primary language: Go (with Kotlin and JS/TS exposure).
After 12 weeks, you should be able to:
- Navigate and understand an unfamiliar codebase
- Fix a contained bug (no external dependencies) with AI assistance
- Ship a simple feature as an alpha/POC
- Write and run tests
- Respond to code review comments intelligently
- Hand off your work to an engineer with clear documentation
- Read
plan.md - Fill in start dates for each week
- Follow the Saturday/Sunday structure
- Keep a learning journal (3 bullets per session: learned, confused, explore next)
- Check off tasks as you go
This is a living document. PRs welcome.
- Completed the plan? Open a PR sharing what worked and what didn't.
- Adapted it for a different stack? (Python, TypeScript, Rust) Share your fork.
- Have a better exercise or resource? Submit a PR.
- Want to discuss? Use the Discussions tab.
MIT