Feedback learning loop: change one variable at a time

Use this 3-round loop

1. run one bounded task
2. record one main issue
3. change one variable on the next run

Round 1: baseline run

Action:

• run once without extra adjustments
• write down input, output, and one main problem

Completion criteria:

• the main problem is described in one sentence

Round 2: one-variable adjustment

Change only one variable:

• output length
• output format
• tone
• step constraints

Completion criteria:

• you can clearly compare before and after

Round 3: retest and lock the pattern

Action:

• rerun on similar input
• compare output quality
• save the better version as a reusable template

Completion criteria:

• one reusable pattern exists
• one next improvement item is recorded

Next step: run this pattern in one real channel from Channel integration basics.