which language do you work with? also, what kind of code do you write?
i do web apps in typescript, fullstack, and so many things are just layers upon layers on the same thing. you make a model, then you make a migration for that model, then a controller for a dead simple crud with just enough custom validation and shit that it’s hard to autogenerate, then swagger docs, then factories and unit tests. most of the time, i just write the model, paste it to the top of the file in a comment, then start writing the thing and the ai immediately wants to take it and do it by itself, so i just let it. there are usually a few mistakes, exactly in those hard to autogenerate places, so i fix a few and then the ai can help fix the rest with little intervention. (i do need to delete the mistake and palce the cursor, but it tends to know how to fix it). it’s been a huge time-saver for me.
but i noticed that it’s also damn good at just generic typescript/javascript. i have it enabled pretty much all the time now and i have yet to find a task where it doesn’t speed me up, even with the weird shit i do for fun.
however, i heard similar complaints quite a few times working with other languages. the amount of open source code that’s out there for your specific language seems to have a large effect on copilot’s effectiveness
yeah, makes sense. when i’m not doing a test-driven workflow, i usually only allow copilot to generate simple things. with unit tests it’s a lot easier because the tests are easy to read and verify, and they expose the errors in other parts of the code
Who has actually gained productivity from using Copilot? Maybe it’s just me, but it slowed me down.
which language do you work with? also, what kind of code do you write?
i do web apps in typescript, fullstack, and so many things are just layers upon layers on the same thing. you make a model, then you make a migration for that model, then a controller for a dead simple crud with just enough custom validation and shit that it’s hard to autogenerate, then swagger docs, then factories and unit tests. most of the time, i just write the model, paste it to the top of the file in a comment, then start writing the thing and the ai immediately wants to take it and do it by itself, so i just let it. there are usually a few mistakes, exactly in those hard to autogenerate places, so i fix a few and then the ai can help fix the rest with little intervention. (i do need to delete the mistake and palce the cursor, but it tends to know how to fix it). it’s been a huge time-saver for me.
but i noticed that it’s also damn good at just generic typescript/javascript. i have it enabled pretty much all the time now and i have yet to find a task where it doesn’t speed me up, even with the weird shit i do for fun.
however, i heard similar complaints quite a few times working with other languages. the amount of open source code that’s out there for your specific language seems to have a large effect on copilot’s effectiveness
I tried it with php which also has a huge amount of code online. The constant checking of the generated code and fixing it was what slowed me down.
yeah, makes sense. when i’m not doing a test-driven workflow, i usually only allow copilot to generate simple things. with unit tests it’s a lot easier because the tests are easy to read and verify, and they expose the errors in other parts of the code