Aider vs Cline Using DeepSeek 3 - Codebase 20k Lines

---

---

AI Summary

Video Summary: AI Coding Tools Comparison - AER vs. Klein

Overview

• The video compares two AI coding tools, AER and Klein, using a medium-sized code repository for a sports betting platform.
• The repository contains over 240,000 tokens and more than 20,000 lines of code.
• The platform has a React frontend and an ASP.NET backend.
• The goal is to fix bugs and add features, not to create simple projects like a snake game or a landing page.

Tasks and Observations

• Bug Fixing:
  • The BET slip button issue: It initially shows a white panel instead of the BET slip.
  • Both AER and Klein struggle with the bug due to the size of the repository and limitations in handling diffs.
  • AER has tight Git integration, making commits as it codes.
  • Klein feels more in charge of the IDE, despite being an extension.
  • Both tools required manual intervention to add files and understand the context better.
  • AER uses Tree Sitter for the repository map, which has its pros and cons.
• Feature Addition:
  • Creating a YouTube API endpoint: The task involves backend and frontend work to search for football match highlights and embed them in the app.
  • AER completes the task quickly but with build errors that needed to be addressed.
  • Klein struggles with returning valid YouTube URLs and requires fixes to return the correct content.

Progress and Results

• Both tools show progress after iterations and fixes.
• AER and Klein eventually fix the BET slip button issue, with AER leaving a “fog of War” behind.
• The YouTube API feature works, with both tools embedding the video into the app after several iterations.
• The presenter suggests a workflow where Deep Seek does the heavy lifting, and another tool like Claud comes in for blockers or decreased velocity.

Additional Notes

• The presenter tests the code in a production MySQL database hosted on Microsoft Azure, which is not sponsored.
• The database contains over 450,000 football matches, including very recent data.
• The presenter hints at a future video testing a Gemini model with a larger context window.

Conclusion

• The video concludes with both AER and Klein versions working as expected, and the presenter invites viewers to comment on which version they prefer.

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