Ralph Loop Practice

Practice the Ralph loop concept manually. This is a demonstration, not automation.

## The Loop (Manual Practice)

1. **Create `AGENTS.md`** - the context file:
   - Brief description of this project
   - How to build/run/test
   - Current focus or next task
   - Any constraints or patterns to follow
   - Keep it under 70 lines

2. **Run the agent** (2-3 iterations max):
   - Give the agent `AGENTS.md` as context
   - Let it make small changes
   - Stop after 2-3 iterations

3. **Review what changed**:
   - Check the diff
   - Did it move toward the goal?
   - What did it miss?

4. **Update `AGENTS.md`**:
   - Add what you learned
   - Clarify what was wrong
   - Update the focus/task

5. **Repeat**:
   - Run agent again with updated context
   - Review
   - Update
   - Continue

## The Point

This is slow mode. You're the checkpoint. The agent gets fresh context each time. You steer by updating `AGENTS.md`.

Once you understand the rhythm, you can automate it. But start here: manual loop, manual review, manual context updates.