AI Pair Programming: A Month's Test, What Really Changes?
·2 min read·Intermediate
“
AI writes code instantly, but does it do it well? One developer tested three tools for a month, using them in daily work.
In 30 seconds
01A developer tested GitHub Copilot, Cursor, and ChatGPT for a month in pair programming.
02AI excels at simple, repetitive functions but struggles with complex logic and broad contexts.
→
💡
What this means for you
For non-programmers, it means the tools we use daily will become smarter and save us time on repetitive tasks. For developers, it's a concrete aid that frees up creative resources.
Building a desktop app feels like wizardry, right? This new open-source kit lets you make one, fast, just knowing web basics.
·2 min·2·Beginner
03It doesn't replace developers, but can speed up tedious tasks, freeing up valuable creative time.
0101
Can AI Really Code by Itself?
No, not yet. AI tools like GitHub Copilot or Cursor are great for small tasks, but they can't replace a human brain. One developer found AI to be an excellent assistant, not a boss making all the decisions.
Elsie Rainee, a developer, spent a month "pair programming" with three different AIs: GitHub Copilot, Cursor, and ChatGPT. Her experience, published in May 2024, showed both the limits and strengths of these tools. It's not a robot taking over, but a slightly overzealous study buddy.
GitHub Copilot, launched by Microsoft and OpenAI, was one of the tools Rainee used to generate suggested code directly in her IDE. It works well for completing lines or writing basic functions that don't require much imagination. So, is AI just a glorified calculator for code?
0202
When Is AI a Genius (and When Is It a Disaster)?
AI shines in repetitive tasks and writing standard functions, like those for manipulating strings or arrays. It becomes a disaster, however, when the logic gets complex or requires a deep understanding of the project's context.
📬 Enjoying this article?
Get the best AI news every week, straight to your inbox.
For instance, Rainee noted that ChatGPT produced nearly perfect code for isolated small problems, but failed miserably with logic involving multiple files. Cursor, for its part, struggled with poorly documented APIs, which are interfaces that let different programs talk to each other.
AI's real strength lies in freeing up time. If you need to write ten similar functions, AI can do eight of them in a flash. This means less time spent on tedious things and more room to think about overall structure or bigger problems.
Elsie Rainee concluded that, as of May 2024, AI is more of a "super-autocomplete" than a true autonomous programmer, useful for automating up to 30-40% of repetitive tasks.
0303
Will It Replace Us All, or Not?
No, it won't replace us, at least not in the short term. AI is a powerful assistant that can improve productivity, but the human touch remains essential for creativity, complex problem-solving, and overall vision.
Think of it this way: AI is like having a super-fast intern who can make copies and fill out forms, but can't decide company strategy. The human programmer is still the orchestra conductor; AI is an excellent musician, but one who only reads the score.
Rainee's experience suggests that integrating AI tools like Copilot and ChatGPT into a programmer's workflow can increase efficiency, not reduce the need for human expertise.
If you thought voice assistants were the peak of digital companionship, prepare yourself. Now there's an app that gives you a virtual character to cheer you on, everywhere you go.
