AI For Programming

I wanted to note down my thoughts here about using AI for programming. I’ve had plenty of experience in the past month using AI for programming, the tools I have used:

• Github Copilot
• Cursor
• Windsurf
• Cline

They all feel pretty similar in what they do, but what stood out most to me is how much better they get when you give them real examples from your codebase. If they can see how your app is actually structured, they start making much more useful suggestions.

This has been especially true when it comes to writing tests. If I give the tool a couple of sample tests, it gets a pretty good sense of the setup and can fill in the rest. That said, you should definitely never trust the output. I’ve had tests with wild assertions that don’t make any sense, or setups that delete key parts of the test when it gets too complex.

One thing I still have doubts over is whether or not this is saving me time… I do feel more productive but I’m not sure if that’s being offset by the additional time I am taking to check the output from the LLM 🤔