🧠 Prompt Like a Pro: The Vibe Coder's Field Guide

Improving your prompts is a key skill when working with AI, especially for tasks like coding, learning, research, and project development. The foundation of better prompts lies in clearly defining the outcome you desire, rather than requesting a generic answer. This approach is akin to writing a function in programming where you provide complete inputs, constraints, and expected outputs.

For instance, instead of asking "Teach me Ruby," specify the desired outcome: "Act as an experienced Ruby engineer and mentor. I have basic Ruby knowledge and want to enhance my debugging skills for software engineering interviews. Guide me through realistic examples, let me attempt solutions before providing solutions, and give feedback on my reasoning."

The six essential components of a strong prompt include:

1. Goal: Articulate what you aim to achieve. Example: Enhance the onboarding experience for new users to complete their first workout log with less friction in an existing Ruby on Rails application.

2. Context: Provide relevant background information. Example: This application has authentication, workout models, and a dashboard, and you want to improve the onboarding flow without altering existing functionality.

3. Constraints: Define limitations and rules for the AI. Example: Keep the current routes and database behavior intact; avoid adding unnecessary dependencies; refrain from refactoring unrelated files.

4. Process: Outline the steps the AI should follow. Example: Analyze the current implementation, pinpoint the root cause of the issue, propose a minimal change, and implement it after explaining the approach.

5. Output Format: Specify the desired response structure. Example: Present the diagnosis, suggested fix, modified files, pertinent tests, and any lingering risks.

6. Success Criteria: Describe how you will evaluate the quality of the response. Example: The feature must function correctly for authenticated users, handle invalid input, maintain existing behavior, and pass relevant tests.

When crafting prompts, you don't need to include all six elements every time. Focus on those that significantly impact the desired outcome.

A practical framework for everyday prompting is the G-C-C-F model:

- Goal: Specify the desired outcome.

- Context: Provide necessary background information.

- Constraints: State any limitations or rules.

- Format: Outline how the answer should be structured.

For example, to prepare for software engineering interviews:

Goal: Prepare for entry-level software engineering interviews.

Context: Basic Ruby knowledge and Rails experience, aiming to improve problem-solving skills.

Constraints: Start with easy problems; do not reveal solutions prematurely; use idiomatic Ruby.

Format: Provide one problem at a time, offer hints when needed, and evaluate correctness, efficiency, edge cases, and readability after each attempt.

The ten most useful iterative prompting techniques include:

1. Change One Thing at a Time: Make incremental adjustments to improve a response, making it easier to identify what works.

2. Diagnose Before Retrying: Identify weaknesses in the previous answer and formulate a revised response that addresses those issues.

3. Provide Specific Feedback: Offer clear, actionable feedback instead of vague dissatisfaction.

4. Define What "Better" Means: Clearly articulate the criteria for improvement, such as reducing answer length, enhancing readability, or including specific examples.

5. Break Complex Tasks Into Stages: Divide complicated tasks into sequential steps, allowing for thorough research, evaluation, implementation, testing, and review.

6. Request Alternatives: When unsure about the best approach, ask the AI to compare multiple options, weighing their advantages, disadvantages, costs, and risks.

7. Ask Clarifying Questions: When information is missing, seek clarification from the AI to avoid misunderstandings and ensure accurate responses.