14 ms·
This! I wonder if it's a language difference thing (I use C# mostly), but my experience with Copilot has been nothing short of mind-blowing. It's not writing
by redhale 3y ago
This!
I wonder if it's a language difference thing (I use C# mostly), but my experience with Copilot has been nothing short of mind-blowing.
It's not writing every line, and when writing truly new feature code it's less useful. But here are two patterns that I've noticed it is especially and consistently good at as a potential place to look for initial value if you are skeptical:
- If you have any kind of repeated pattern, even very complex ones, like performing a set of operations on a set of objects, or initializing a set of things, or whatever, it will guess everything else after the first line 90% of the time. This stuff is almost always plumbing/wiring/boilerplate-type code, so pure time gained. Think about Excel filling incrementing numbers down a column, but for pattern-matched lines or blocks of code.
- For any reasonably testable class, if I write the name of the unit test, Copilot will write the entire unit test perfectly 90% of the time, down to variable names and //Arrange//Act//Assert comments that I stylistically prefer. Seriously, it's sort of scary how good it is at this.
- toyg 3y agoLanguages like C# and Java have always required a lot of boilerplate and repetitive (sorry, "patterned") code. They are perfectly suited to GPTs.
- justinlloyd 3y agoI've used ChatGPT and Copilot with Javascript & Typescript and a bunch of frameworks. Python, C, C++, C#, Go, 6502, Z80, ARM and a few other languages so far. They've all worked pretty good. I wish it had a wider breadth of APIs and documentation, but it is pretty good so far.