ad
ad

Beware of AI Coding Assistants | GitHub Copilot vs JetBrains AI

Science & Technology

Introduction

As promised, I compared the most popular AI-powered coding assistants, intending to create a comprehensive test and reveal the results at the end. However, I immediately noticed some critical lessons worth sharing beyond my preference for GitHub Copilot over the competition. There are features you should adopt and others you should avoid. Let me start by briefly explaining why I chose GitHub Copilot.

Why GitHub Copilot?

An AI coding assistant is a tool designed to assist software developers in various ways. Its primary function includes auto-completion and code prediction. It suggests lines or even entire functions as you type a few letters, which can be quite fascinating, especially for first-time users.

How Do They Work?

These tools are built on large language models trained on vast datasets, including open-source and some private codebases. No human could process the same amount of data in multiple lifetimes. Consequently, these models have learned coding patterns and can predictively analyze the context of a project, suggesting accurate code predictions.

The Comparison: GitHub Copilot vs. JetBrains AI

GitHub Copilot

  • Auto-completion excellence: Highly responsive and accurate
  • Multiple predictions: Offers several suggestions at once, accessible via a dedicated tab or shortcut keys

JetBrains AI Assistant

  • Less responsive: Predicts code only on empty lines
  • Explicit prompts needed: Requires detailed instructions, disrupting the workflow

Lesson 1: Avoid Overreliance on Prompts

Relying too much on AI-generated prompts while coding can be detrimental, especially for junior developers. Here are the drawbacks:

  1. Imperfections in code: AI-generated code often contains syntactic errors, requiring significant refactoring.
  2. Logical flaws: The generated code may not be bug-free, risking logical errors.
  3. Skill degradation: Spending more time reviewing rather than writing code hampers the use of problem-solving skills.
  4. Bug detection difficulty: Especially challenging for junior developers to spot errors in code they don't fully understand.

Lesson 2: Avoid Test Autogeneration

Automatic test generation has its pitfalls:

  • Incompatibility with TDD/BDD: Doesn't align with test-driven or behavior-driven development practices.
  • Refactoring and reviewing: Similar issues as with code prompts.
  • False sense of security: Generated tests might mirror the logic of flawed implementations, making errors harder to identify and resolve.

Features to Love

Despite these concerns, AI coding assistants can significantly improve efficiency and productivity, making us better software engineers. Here are two highly recommended features:

  1. Explain Code: Excellent for flattening the learning curve for both junior and senior developers, especially when learning new languages, frameworks, or unfamiliar codebases.
  2. Simplify Code: Helps in discovering more elegant and efficient solutions. This feature analyzes code and offers alternatives, providing fresh perspectives on problem-solving.

Conclusion

AI coding assistants hold immense potential, but it's crucial to remain in control of your coding process. You are the pilot; the AI is your co-pilot or pair programmer. Keep enjoying coding, using your creativity, and honing your problem-solving skills. While AI may eventually replace software developers, that time has not arrived yet. Stay in control of your work.

If you found this article helpful, don't forget to support by liking, subscribing, and sharing it so others can learn something new.

Keywords

  • AI Coding Assistant
  • GitHub Copilot
  • JetBrains AI Assistant
  • Auto-completion
  • Code Prediction
  • Prompting
  • Test Autogeneration
  • Explain Code
  • Simplify Code
  • Junior Developers
  • Bug Detection

FAQ

Q1: Why did you prefer GitHub Copilot over JetBrains AI Assistant?
A1: GitHub Copilot is highly responsive and accurate with its auto-completion and offers multiple predictions simultaneously, making the coding process smoother and less disruptive.

Q2: What are the major drawbacks of relying on AI-prompted code?
A2: AI-prompted code often contains syntactic and logical errors, leading to significant refactoring. It can also hamper the development of problem-solving skills and make bug detection challenging, especially for junior developers.

Q3: Why should test autogeneration be avoided?
A3: Test autogeneration is incompatible with test-driven and behavior-driven development practices. It also shares the drawbacks of AI-prompted code, adding a false sense of security by generating tests that mirror the logic of potentially flawed implementations.

Q4: What are the effective features of AI coding assistants?
A4: The "Explain Code" feature helps flatten the learning curve for developers, and the "Simplify Code" feature offers alternative, often more efficient solutions for problem-solving.

Q5: How should developers approach using AI coding assistants?
A5: Developers should view AI as a co-pilot or pair programmer, maintaining control over the coding process and using AI to enhance creativity and problem-solving skills.