Free practice · ordering · hard

Practice

Control a complete AI conversation

Order the stages of a deliberate multi-turn AI task.

Bounded iteration loop

Specific feedback should move a draft toward a visible stopping rule

  1. TargetSet outcome and stopping criteria
  2. ExampleShow an approved pattern and boundary
  3. DraftGenerate one checkable part
  4. FeedbackName the exact source-based repair
  5. DecideApprove, continue, refresh, or stop

Examples and small steps reduce ambiguity. Context refresh prevents drift, while the criteria determine when to approve, continue, or stop.

Arrange in the correct order
  1. 1. Approve, continue, or stop using the criteria
  2. 2. Refresh critical context when needed
  3. 3. Give specific feedback on each draft
  4. 4. Break the task into checkable parts
  5. 5. Provide an approved example and source packet
  6. 6. Set task, data, owner, and success boundaries