AI Coding Adventures: A Journey of Trials and Errors
Join our CTO, Aleksander, as he embarks on a weekend quest to test AI’s coding prowess. The challenge? Refactoring our dashboard from NgRx to vanilla services with signals.
Key Takeaways:
- Task Complexity: A simple migration with no new logic—ideal for AI.
- AI Performance:
- Microsoft Copilot: Offered minimal help; significant content deleted.
- Cursor.ai: Caught in an endless error loop.
- GPT-4o: Generated empty services with vague suggestions.
- Claude 3.7: The standout, delivering usable code but still leaving errors.
Despite the struggles, Claude managed to produce 200 usable lines over a full day of work.
Final Thoughts:
Aleksander’s experience highlights the potential—yet limitations—of AI in coding tasks.
👉 Curious about the future of AI in programming? Share your thoughts below! Let’s discuss how we can harness AI’s full potential in tech!
