AI Coding Guide

Common Mistakes and Undesirable Things Done by AI When Used for Coding

Understanding the limitations and common mistakes of AI coding assistants is crucial for using them effectively. This section outlines the most frequently encountered issues based on research from various sources.

Table of Contents

Hallucinations

Hallucinations occur when AI models generate content that appears plausible but is factually incorrect or entirely fabricated. In coding contexts, this often manifests as:

  • Inventing non-existent methods or libraries that don't exist
  • Creating plausible but incorrect API references
  • Failing silently instead of returning useful errors
  • Hallucinating features that don't exist in referenced libraries

Tip

When AI suggests using a method or library you're unfamiliar with, always verify its existence through official documentation.

Context and Understanding Limitations

AI coding assistants have significant limitations in how much context they can process and maintain:

  • Unable to process large codebases (>100-200 lines)
  • Losing context after 5-10 messages in conversation
  • Failing to maintain context across multiple interactions
  • Unable to see the full code when selected
  • Starting over internally and requiring re-explanation of tasks

Warning

Breaking down complex tasks into smaller, more manageable pieces is essential when working with AI coding assistants.

Security Issues

AI-generated code often contains security vulnerabilities that may not be immediately apparent:

  • Generating code with security vulnerabilities
  • Creating authentication mistakes
  • Introducing SQL injection vulnerabilities
  • Causing buffer overflows
  • Creating symlink vulnerabilities
  • Making code that appears secure but contains subtle flaws

Quality and Reliability Issues

AI-generated code often has quality issues that may not be immediately obvious:

  • Producing code that looks good but doesn't work correctly
  • Generating convincing but incorrect implementations
  • Providing overly complex solutions for simple problems
  • Using outdated patterns or approaches
  • Lacking proper documentation or using confusing variable names
  • Employing suboptimal algorithms and design patterns

Overconfidence Problems

Both AI models and their users can exhibit overconfidence in the generated code:

  • Users overestimating the security of AI-generated code
  • Developers having false confidence about code correctness
  • Reduced critical thinking when reviewing AI-generated code
  • Assuming code works without testing it thoroughly

Tip

Always approach AI-generated code with healthy skepticism and test it thoroughly before implementation.

Operational Issues

AI coding assistants can exhibit frustrating operational behaviors:

  • Too many "I can't assist with that" replies
  • Providing responses unrelated to the code or question
  • Making changes across multiple files without clear user confirmation
  • Potentially exposing sensitive information through automatic context loading

Skill Development Concerns

Overreliance on AI coding assistants can impact developer skill development:

  • Overreliance leading to decreased ability to code independently
  • Reduced creative problem-solving skills
  • Diminished understanding of underlying principles
  • Difficulty debugging AI-generated code without understanding it

Warning

Use AI as a tool to enhance your capabilities, not as a replacement for developing fundamental coding skills.

Intellectual Property Risks

Using AI-generated code can introduce intellectual property concerns:

  • Potential copyright issues with generated code
  • Unclear attribution of code sources
  • Possible exposure of trade secrets through context sharing
  • Impact on company technical credibility during due diligence