Our company provided us with a GitHub Copilot license, and upper management launched a survey asking if we're really using it.
No doubts here, it's awesome. (Or any of the free alternatives out there—not talking about Copilot specifically.)
Sure, the main reason people defend generation tools like Copilot is testing.
🤓 <snobby-nerd-voice>- but you still need to tune the code - </snobby-nerd-voice> - and I'm glad I still have to review things!
personally, I've gone from 'let's create this sub-task in the board' to: 'I can get them done in 15 minutes'.
But let me add another:
It's fantastic for learning. I've learned more CSS in six months using Copilot than i was learning in ten years; because it provides those general patterns you're looking for.
Patterns... there are patterns in algorithms that I don't even consider: mappings, indexes, and the whole '+1 because the element is to the right of X' thing.
And now, to the key point of this post: something I feel not talked about enough.
That 'first stone' 🪨
If you're not sure where to start, the 'first stone' is super super super valuable.
Even if not a single line of the original suggested code remains, it gets your brain to reply: 'Wait, no, this should be like this and that.'
My brain shifts from a 'blank page' to fixing mode. And see,
We're better at editing than writing from scratch.
Top comments (3)
I've been using CoPilot for about a month coding in R. It seems about in par with Wikipedia for analytic and statistical solutions. Waiting patiently for some intelligence here.
endless possibilities...
if it takes someone 10 years to learn CSS who uses it often, then AI is not gonna help where it matters