Grok for Advanced Reasoning and Complex Problem Solving
Lesson 2: Structured Self-Critique and Multi-Perspective Review Loops
Lesson Objectives
By the end of this lesson, students should be able to:
- Design and run structured self-critique prompts that evaluate output against explicit success criteria
- Request the model to adopt multiple perspectives (optimist, skeptic, risk manager, end user) and synthesize the feedback
- Use critique output to drive targeted, specific revisions instead of vague improvement requests
- Know when a review loop has reached diminishing returns and human judgment should take over
Lesson Content
One of the most powerful advanced habits is treating the first good draft as the starting point for deliberate improvement rather than the finish line.
A structured self-critique loop usually follows this pattern:
- Generate an initial version using a strong goal + context prompt.
- Ask the model to critique its own output against the original success criteria, looking for specific weaknesses, missing elements, or risks.
- Request targeted revisions based on the critique.
- Optionally introduce additional perspectives (for example: "Now critique this from the viewpoint of a skeptical reader" or "from a risk-management perspective").
- Synthesize the feedback and produce a revised version.
- Stop when further changes produce only marginal improvement.
This approach is far more effective than repeatedly saying "make it better." It forces the model (and the user) to be specific about what needs to change and why.
Enroll to continue this lesson.
The preview above shows the lesson objectives and opening lesson content. Enroll to view the full lesson, complete the practice work, and take the lesson quiz.