In my experience working with GPT4, if I give enough context on types, other functions definitions and the libraries I use, I get very accurate results. But it is a tedious task to copy paste from multiple places (type definitions, function definitions, packages, etc.).
In addition to the selected lines, does Continue support getting related definitions from the language server and inject them in the prompt? That would be huge.
This is very near on the roadmap, and we agree it will be awesome!
As of now, if there are collections of definitions that you frequently reference, you could save them in the system message, or write custom slash commands that let you prefix your prompt with these definitions.
After I copy every piece of code that is relevant, ask it to do something, then correct it's errors using human knowledge, then ask senior devs if it makes sense to revise my prompts and repeat, fix bugs from reviews from other humans, it's like this magical thing I get the mostly right answer!
For smaller pieces, you are right. But as a BE engineer, I was prototyping a concept and I didn't know much about Typescript+React+React-Router-Dom+React-Hooks-Form, etc etc. So I listed the libraries, a few lines of boilerplate and API definitions that RTK toolkit generated for me. Then asked GPT to generate the full page. It was much faster than I could "code it myself".
And that's why it is a "pain point". These all can be done automatically.
If you ever need to extend that full page to do something else or something additional you might find the time not so different after all. Because you still don't know much about Typescript+React+React-Router-Dom+React-Hooks-Form and now you have to read that stuff and try to make sense of it, even if just for passing it back into the bot.
Actually writing code is a minority of my time spent developing software, I don't need to trade that for need to spend time to know what and more importantly why something is where it is.
I guess it does work if you just want it to generate a webpage and you never want to then add any functionality. Or a glorified boilerplate generator for stuff not on the prod path.
> I don't need to trade that for need to spend time to know what and more importantly why something is where it is.
I had two options: 1. Learn everything first and then start prototyping. 2. Start prototyping and learn along the way. I chose the latter. But instead of searching stackoverflow and putting the pieces together, I used GPT and learned from the generated outputs in the context of my own problem.
It may not save a lot of time in larger projects because as you said, one should ultimately learn the "what and why", but it definitely provides a more pleasing experience. And I guess the time-saving part becomes more relevant with better tooling, like the suggestion I had.
Fair - I guess I'm just used to the old ways and hated things showing up without me specifically telling them to show up there first. But then, I still learn while prototyping, ChatGPT might be easier since it does give you a template to go off of.
But then, I don't think this really go beyond entry level prototyping, any complexity and chatGPT doesn't have enough tokens.
I'd say it's more like driving vs assisted driving with shouting inputs (e.g. TURN LEFT! WATCH FOR THE PEDESTRIAN! CRUISE AT 70 MPH! SLOW DOWN THERE'S ANOTHER CAR IN FRONT! USE THE SECOND RAMP TO EXIT!). The only benefit is that you don't have to hold the wheel any more, the negative being that it still only works like 90% of the time, and you end up with laryngitis.
Couldn't agree more—it's worth the extra effort to know exactly what enters the prompt. But control isn't mutually exclusive with removing the need to copy/paste. Continue lets you highlight code with perfect precision, and this is much lower effort.
In addition to the selected lines, does Continue support getting related definitions from the language server and inject them in the prompt? That would be huge.