Does GitHub Copilot Improve Code Quality? Here's How We Lie With Statistics
Does GitHub Copilot Improve Code Quality? Here's How We Lie With Statistics
A critique of GitHub's research blog and marketing-driven science.
TLDR: no.
There, I saved you 15 minutes and 6 advertisements.
I make it no secret that I am against the AI Hype train.
I’m not a scientist either, and my opinions aren’t facts
I think I'm just going to stop right there. I give them credit for honesty, But if you have to come right out and say everything you're about to say is complete unverifiable bullshit, I have no reason to read it unless I'm in that echo chamber.
If you get code from copilot, theres an above average chance it's commented and commonly used.
Does/did stack exchange improve code quality? Does anybody really care?
stats published elsewhere show experienced programmers loose productivity using these new tools because they have to debug and fix things.
I think such tools work better for inexperienced programmers who are making small code stuff, like school assignments and self teaching, or need a skeleton made to fill in later.
The problem is that this is not where the big money is at, so a bubble and questionable actions have occurred trying to capitalize on it.
Our team isn't using it to generate huge swaths of code, just boilerplates for tedious stuff.
When I use it, I use it to crap out boilerplate stuff. Give me a flask app with these endoints, setup debug logging and give me jinga templates with these variables.
I don't need to write that code more times in my life :)