AI Coding 101 - Ultimate Prompt Guide (37 tips)

AI Coding 101: Ultimate Prompt Guide

This is a companion document to go along with my AI Coding 101 video. This document contains all the prompts discussed in the video - though I suggest you watch the video for all of the additional context and explanation that relates to these prompts and tips.

You can watch the full video here:

🎥 AI Coding 101: Ultimate Prompt Guide (37 tips)

Writing New Code

1. Specify technologies and frameworks

Prompt:

Please create [app/feature description] using [technology list]

Example:

Create a weather tracking app using React (vite), TypeScript, ShadCN, and Tailwind

2. Choose Popular Technologies

Recommendations:

Frontend: TypeScript, React, Tailwind, ShadCN  
Backend: TypeScript, Python, Express.js, FastAPI, Django  
Databases: Postgres, MongoDB, Redis  
Cloud: Cloud Storage (S3), Supabase Auth, Cloud/Edge Functions

3. Design the solution, then implement with AI

Prompt:

I am building a [feature]. It should work like this: [list of requirements]  
We should [implementation tips]. It should also handle these edge cases: [list]

Example:

We’re building a search feature. It should perform the search based on the product title and description.  
  
Create a new component, SearchBar, and add it to the existing NavBar component. Also, create a new search endpoint in the @server.ts file.  
  
When no results are found, we should display 'No Results' in place of the product list.

4. Break things down, give the AI a list of tasks

Prompt:

Please implement the following:  
1. [Requirement #1]  
2. [Requirement #2]  
3. [Requirement #3]

5. Find the right balance of task scope (not too big or small)

6. Use examples in your prompts

Prompt:

Please create a [feature] that follows this example: [example]”

Example:

Create a date formatting function. It should work like this:  
1.Input: '2024-02-07' → Output: 'Feb 7, 2024'  
2.Input: '2024-02-07T15:30:00' → Output: 'Feb 7, 2024 3:30 PM'

7. Provide sample code for the AI to follow

Prompt:

Please create a [feature]. Here is working sample code: [code]

8. Use @Docs and @Web context utilities

Prompt:

Please build a [feature]. Reference the [@Docs / link] docs

Please build a [feature]. Find the best practices and examples for building this on the [@Web]”

9. Make sure the code is secure and performs well

Prompt:

Please implement [feature]. Make sure to account for any potential security issues.

We just implemented [feature]. Can you check to make sure that it use the most efficient approach?

10. Proactively check for additional considerations

Prompt:

Are there any additional considerations for building [feature]?

11. Ask the AI to write tests & try Test-Driven Development

Prompt:

Please write the tests for the new feature we just implemented

Test-Driven Development:

We are building [feature]. Please write the tests to cover the following cases: [list]. Once complete, build the actual feature.

12. Have AI generate your documentation

Example:

Please add an explanation of this code to our README

13. Use precise naming

Example:

Please create a new csvParser method and add it to the dataParsers.ts file

Modifying Existing Code

14. Keep code organized by asking AI to refactor & simplify

Prompt:

Please refactor the code in [file / function] to split it up into [list of files / functions]

Example:

Please split up dataParser.ts into dedicated files including csvParser.ts, xmlParser.ts, and pdfParser.ts

15. Manage and balance the CONTEXT you give to the AI

16. Tag specific relevant files as context

Prompt:

Please implement [feature]. You’ll need to modify [list of files]. Here are additional files for context: [list of files]

17. Keep your files at 300 lines of code or less

18. Keep in mind that the full conversation is used as context

19. Start a new conversation for each new feature

20. Tell the AI what works well and what needs to be changed

Prompt:

This was a good start. [list of features] are working well but [list of problems] still needs to be fixed. Can you fix that?

Example:

The new search feature is properly calling the API and showing results but it is not properly searching by product description. Can you please fix that?

21. Ask AI to find edge cases or bugs

Prompt:

Are there any edge cases we should consider or handle?

22. Ask AI to review the code

Prompt:

Now that we have finished building [feature], can you do a full review of the implementation and the new code?

23. Ask AI about what the current code does

Prompt:

How does the code in [@file] work? Give a high level overview

How does the [function] work? Give a detailed description”

Example:

What does the code in @serverComm.ts do? How does it handle authentication?

Troubleshooting

24. Be specific about what is not working (and what is)

Prompt:

When I do [X], I see [Y], but when I do [Z], it seems to work.

Right now the [feature 1] is working correctly but [feature 2] is failing [failure description]. Can you fix that?

Example:

The buttons are visible now but when I click them nothing seems to happen. Can you fix that?

Example:

The layout is still incorrect. Please see [@screenshot] and note how the sidebar is not properly expanding.

Prompt:

It’s still failing - here are the errors: [full error logs]

27. The Beaver Method

The beaver method is a troubleshooting approach that works by asking AI to add logs throughout the code that you are troubleshooting, running the software, and feeding the logs back into the AI so that it can troubleshoot with all the relevant runtime context.

Step 1 Prompt:

Please add logs at every step of the process to make it easier to troubleshoot and figure out where the problem is.

Step 2 Prompt:

I ran the process and here is the result: [full logs generated from step 1]

28. Ask AI to explain the code

Similar to #23 but in the context of troubleshooting to help bridge the disconnect between developer expectations and the way the code actually works (which the developer currently perceives as incorrect).

Example:

This code doesn’t convert dates correctly. Can you explain how it works?

29. Ask AI to use a “radically different approach”

Prompt:

This still isn’t working. Let’s try a radically different approach.

30. Know when to stop asking AI and read the code yourself

Learning To Code

The following tips match what was covered in the AI Coding 101 episode, but the Learn To Code page contains more detailed tips for learning to code specifically including a technology glossary that can help guide your exploration.

31. Tell AI to keep it simple + that you are a new dev

Example:

I am learning how to code. Can you simply explain how the frontend connects to the backend?

32. Ask AI to explain code “line by line”

Prompt:

Please add comments explaining [function] line by line

Example:

I am learning how to code. Can you explain the getContactDetails function line by line?

33. Ask AI to explain specific technologies & concepts

Prompt:

Can you explain how [concept] works? What does [technology] do?

Example:

I'm a new dev. Can you explain what databases are for, when I should use them, and what Postgres is specifically?

34. Ask AI how to build something

Prompt:

I want to build a [app/feature]. How would you approach this problem from a technical perspective?

Example:

I want to build a chat app. How would you architect this and what considerations do we need to take into account?

35. Ask AI to show you examples

Prompt:

Can you show me a working example of a [concept] and explain it?

Example:

I am learning to code and want to understand how objects work. Can you create some examples and explain them?

36. Tell the AI what you already know vs what is unclear

Prompt:

I know how [concept] works but can you explain how it relates to [concept 2]?

Example:

I have an API and understand how requests come in but can you explain how I could add authentication to my endpoints?

37. Focus on learning and understanding core concepts

AI Summary

Summary of AI Coding Tips Video

Writing New Code

Be Specific with Technologies: Use a specific list of technologies or approaches for the AI to ensure compatibility and higher output quality. Popular languages and frameworks like JavaScript, Python, and PostgreSQL are recommended.

Design Solutions: Design the solution first, then ask AI to build it, providing specific details about the feature implementation.

Break Down Tasks: Organize prompts into a list of tasks for the AI to focus on.

Balance Scope: Find the right balance of scope to give to the AI; too little leads to slow progress, too much is unrealistic.

Use Examples: Provide examples of desired outputs or sample code to guide the AI.

Reference Documentation: Utilize AI coding tools that can reference official documentation or search the web to improve solution quality.

Ensure Security and Performance: Ask the AI about security vulnerabilities and efficiency satisfyafter implementation.

Test Driven Development: Consider having the AI write tests first, then implement features to those tests.

Document Code: Have the AI help with documentation for both developers and end-users.

Working with Existing Code

Update Naming and Organization: Continuously refactor and update naming to manage context for the AI.

Refactor Code: Use prompts to ask the AI to refactor and simplify code.

Manage Context: Balance the amount of context given to the AI to avoid confusion and ensure compatibility with existing code.

Start New Conversations: For new features, start fresh conversations with the AI, tagging relevant files.

Iterate with Feedback: Tell the AI what works and what needs to be changed during iterations.

Find Edge Cases: Ask the AI to identify potential bugs or edge cases in the code.

Code Review: Use the AI to review code changes and provide feedback.

Understand Code: Ask the AI to explain what the current code does to gain better understanding.

Troubleshooting

Be Specific with Issues: Clearly describe what is working and what is not when troubleshooting.

Use Screenshots and Errors: Share screenshots and error messages with the AI for clarity.

Logging Method: Implement logs to help the AI troubleshoot issues more effectively.

Radical Approaches: If stuck, ask the AI to try something radically different.

Know When to Code: Recognize when to pause AI assistance and review the code yourself.

Learning to Code with AI

Keep It Simple: Ask the AI to simplify explanations for beginners.

Explain Line by Line: Request line-by-line explanations or comments in code to learn syntax.

Understand Concepts: Have the AI explain coding concepts, technologies, and how software components fit together.

Ask for Examples: Request examples to see practical implementations of concepts.

Clarify Understanding: Communicate what you understand and what is unclear for targeted explanations.

Focus on the Big Picture: Concentrate on understanding the architecture and components of software rather than just syntax.

Additional Notes

A link to a prompt guide with all prompts and a glossary of technologies is mentioned in the video description.

The importance of naming conventions and context management is emphasized throughout the video.

The video is part of an AI coding 101 series, with a follow-up video on full stack applications.

(Note: No CLI commands, website URLs, or detailed instructions were provided in the transcript for extraction.)

ThirdBrAIn.tech