I spent years getting great with powershell so I can now confidently copy code out of chatgpt. Chatgpt’s ability to spit out close to correct code faster than I can type it is amazing, but useless if you don’t understand what the hell it’s trying to do.
Pretty much this, it’s the one use case for copilot, I know what I want to type anyway and copilot is usually close enough that 2 edits is faster than typing the whole thing and better for rsi.
Basically the work flow has changed from:
Find a framework that I need to integrate for whatever reason. Go to GumboChumbo.io read the docs.
Write some code based off of what’s in the doc, test the thing, read error message, read docs, ad new thing, but wall for obscure reason, spend thirty minutes looking through similar issues via Google-fu and then find an obscure comment from 6 years ago, That some how fixes this current issue. Implement it, get it working and then customize it.
Now it just streamlines finding these solutions.
But sometimes Copilot just uses too much words to present the answer, so I use ChatGPT which can be personalized.
(Maybe it is possible with Copilot, too, maybe I have to ask how to do it)
I find the opposite, chatgpt (free version at least) gives all the explanation and stuff then a code block, copilot (not Microsoft the GitHub one) just prints the boilerplate directly in the editor then you press tab to accept.
