I still don’t understand how AI helps people code better. Better auto completion and debugging? Snippet generation? The advantages over your average Neovim config always seemed minimal to me.
I use it to fill in things I generally know how to do, but I can’t be bollocksed to look up the exact syntax. For example, I want to convert a shell git config command to an ansible task. I generally know how to look it up and approximately how it’s supposed to be, but not the specifics, so asking an LLM to quickly write the task for me and then I review it in-line, is faster.
It can really speed up a senior dev working in an unfamiliar ecosystem. Concepts often apply across tech but syntax and implementations are different. Especially for the “declarative” things where, in my experience, the syntax is fiddly and it’s not always easy to see where you went wrong. Speeds me up with Terraform, or a recent example was configuring some OpenSearch indices and queries. I’ve never worked with that before but it was able to help me get some boilerplate going that let me start iterating, much more quickly. I get more done when I’m outside my wheelhouse.
But one of the downsides of course is it occasionally sends me down a hallucinatory rabbit hole for something that doesn’t exist, lmao.
I now just code for fun but I use it to speed up writing, not replace it. For example I needed to help my wife with an Excel VBScript macro. I could have spent several days learning the syntax of VBScript and the custom Excel methods but instead I had AI write it. I then only needed to verify and debug it.
I still don’t understand how AI helps people code better. Better auto completion and debugging? Snippet generation? The advantages over your average Neovim config always seemed minimal to me.
I use it to fill in things I generally know how to do, but I can’t be bollocksed to look up the exact syntax. For example, I want to convert a shell
git configcommand to an ansible task. I generally know how to look it up and approximately how it’s supposed to be, but not the specifics, so asking an LLM to quickly write the task for me and then I review it in-line, is faster.It can really speed up a senior dev working in an unfamiliar ecosystem. Concepts often apply across tech but syntax and implementations are different. Especially for the “declarative” things where, in my experience, the syntax is fiddly and it’s not always easy to see where you went wrong. Speeds me up with Terraform, or a recent example was configuring some OpenSearch indices and queries. I’ve never worked with that before but it was able to help me get some boilerplate going that let me start iterating, much more quickly. I get more done when I’m outside my wheelhouse.
But one of the downsides of course is it occasionally sends me down a hallucinatory rabbit hole for something that doesn’t exist, lmao.
I now just code for fun but I use it to speed up writing, not replace it. For example I needed to help my wife with an Excel VBScript macro. I could have spent several days learning the syntax of VBScript and the custom Excel methods but instead I had AI write it. I then only needed to verify and debug it.
It’s like having an intern working for you.