[poll] The level you use AI in editor (Python ver)
Posted by Acpear@reddit | Python | View on Reddit | 21 comments
AI is everywhere now, ChatGPT, Gemini, Copilot, Grok and more. What level you use AI in your favorite code editor(s)?
- High: I use almost AI features provided by my editor, and I enjoy it a lot and it helps me a lot.
- Middle: I just use a few features, some are good, others are not.
- Low: I rarely use AI features provided by the editor, but at least I accept AI generated code (from my browser).
- None: I reject ALL AI features in my editor despite they are provided, or my editor has no AI features. All code is written by myself / humans.
Hint This post is a Python version of original post in C++, link, I thought AI would be much more acceptable among Python programmer community, comparing to C++, which requires to satisfy more(?) rigid(?) rules, sorry in advance to whom feels unsatisfied.
Terrible-Penalty-291@reddit
I will very occasionally ask Copilot a simple question and it will regurgitate some variation of a solution from Stack Overflow it was trained on. Beyond that, I don't use AI in my editor.
Helpful-Guidance-799@reddit
NONE.
Currently using Python and C++ programming for learning purposes so I choose to not use any AI tools with my code editor. Struggling is how growth happens. Using AI can take that important element away.
However, I do use AI when I want to learn about topics. Half the time I’ll read info from blogs, forums, YouTube videos, etc., and the other half I prompt an LLM to tell me about a subject and I ask follow up queries.
sloggo@reddit
Smart mate. Over time though, it may be wise to use ai tools a little and in certain areas, but I have no idea the optimal transition roadmap from “pure growth and learning” mindset to that.
insanjay@reddit
Just keep iterating over it, and better if you take notes of what is working and what's not... And you will found out the most personalized learning path/methods for yourself...
sloggo@reddit
I’m at a different point, something like 20 years in to my career. I’m pretty comfortable reaching straight for agentic AI and directing it to code, and to approach problems, the way I would without it. And fixing/correcting it where it falls short.
There’s things it just makes so convenient I’m constantly saying to myself I’m glad it didn’t exist while I was learning! I genuinely believe there’s something you get from “doing the hours”. Simultaneously though I wonder if I’m right, maybe someone “doing the hours” with all the AI support and the right diligence will become all the more proficient with those workflows.
insanjay@reddit
Yep, that's the point... I went straight to AI in past to write me the code for the project I got benefit and drawback...
Benefit: is that I gained the experience of the workflow or you can say a simulation of production workflow so now I'm not more someone who's confused where to start what to do, and so on...
Drawback: My foundation remained hollow which I'm trying to filling now, I'm in my final year clg so can't say if I'm in a better situation or not... But I'm trying to learn by doing and by researching also I'm using AI to get personalized answers to fill the gaps faster
What say???
insanjay@reddit
I'm shifting to what you already doing, I started with experiments and found out that - If you want to gain the experience faster and gain overview of a process then using AI in almost everything is good, but if you want to learn it for real you have to use research based learning method.... Cuz sometimes even youtube video doesn't fits to your learning goals
backfire10z@reddit
Slightly above Low. My company provided us with AI extension in VSCode that has proven useful sometimes and useless other times.
-LeopardShark-@reddit
None
No_Lingonberry1201@reddit
I usually ask deepseek for a summary and a few examples when I'm doing something new, but otherwise it's more hassle than it's worth.
JestemStefan@reddit
I use Github Copilot for code completion and that's it.
It's good at generating boilerplate code and tests.
For new code it's wrong 50% of the time.
We also fired a developer with 4 yoe that started using Cursor, because quality of his work became unacceptable
Dry_Term_7998@reddit
Depends on tasks: if I need generate fast script to scrap some data or fix some bullshit, Claude the way. One minute for nice print and generation and you are done, here you have validation speed vs writing from scratch. So in this sector medium - high
Production code - app code, low, only for generation docstrings, prototyping and linking/formating formatting for PR.
For deep research for example, in language field or ideas, I use Gemini pro.
I would say not use AI nowadays = not use linters 10 years ago, AI give nice speed, for some specific lightweight tasks, ofc if you want use AI to generate app on medium level + it will be piece of sh$&, you cannot trust to it, quality will be mixed and syntax will be shit, only if you spent a lot time with promts. But tbf what I love in same way in LLMs nowadays, use it like part of products, but not for generations but for predictions and analyzing, I build myself few of SLM and heavy use in home projects SLM like tiny llama, bitnet, mistral etc with fine graining, axolotl, qlora and other tools (PyTorch, Tensorflow, Transformers). Generative LLM overhyped and dumb people eat it, same as “humanitarian” engineers in IT world 🤨
Hermasetas@reddit
I work with sensitive information, so I will never trust a third party AI into my editor.
I use AI in the browser for research a lot though.
Balance-@reddit
GitHub CoPilot for code completion (never use the chat) and Claude for some larger things.
GrainTamale@reddit
None in my editor (VSCode, Python 1.13).
I do use Claude for helping me think about (compare and contrast) alternatives, and I've used it to write tests (don't tell me you enjoy writing tests)
Ringbailwanton@reddit
None. If I do use any sort of AI I’ll use it separately from a browser, but it’s generally not for the Python I’m writing, more for structural things.
BarRepresentative653@reddit
Anthropic looking for unpaid user surveys. FYI, I will never pay for LLM if LLMs like deepseek can exist for free. And TBH, can’t wait for hardware to catchup so we can have self contained LLM in our phones.
Yall will never be profitable
Acpear@reddit (OP)
I am not from any company, I ask this for my interests, because I installed new VSCode and I saw they EVEN added MCP server?! that's another AI feature I won't use in VSCode. Then I searched "does vscodium contain AI features" and it is true, I ask here and ask you, for like or dislike, use or not use AI features in editors.
17modakadeep@reddit
0 use
Cloned_501@reddit
None
There isn't any reason for me to use it.
ThereIsOnlyStardust@reddit
Same here. When my job originally started pushing it on us I figured it might at least be good for generating quick test scripts but I wouldn’t trust it with production code. After trying it out I no longer trust it with production code or test code.