4 ms·

I use chatgpt as my thinking partner writing code. I chat with it all day everday to finish work. My company has approved copilot but Copilot autocomplete has

I use chatgpt as my thinking partner writing code. I chat with it all day everday to finish work. My company has approved copilot but Copilot autocomplete has been an awful experience. company hasn't approved copilot chat ( which is what i need) . But I would love something similar that can run on my laptop for my code to generate unit tests, code comments ect ( ofcourse with my input and guidance).

deergomoo 3y ago
> My company has approved copilot but Copilot autocomplete has been an awful experience I had the same experience, I feel like I must be crazy because so many of my colleagues have been singing its praises. I found it immensely distracting and disabled it again after a couple of days. It was like having someone trying to finish my sentence while I was still speaking; even when they were right, it was still annoying and knocked me out of my flow (and very often, it wasn’t right).
- fipar 3y ago
  I actually find copilot quite useful but I use it from emacs and it only provides suggestions when I intentionally hit my defined shortcut for it, so it never gets in the way. It may be worth for you to try setting it up in a similar way in the tool you’re using it from, as I agree I’d find the experience awful if it was always trying to autocomplete my sentences.
moffkalast 3y ago
Fwiw, there are now some local models that rival 3.5-turbo in code chat, like the Codeninja I tried out the other day. Not nearly as good as 4 which iirc runs the copilot backend, but for sensitive data that can't leave the premises it's the only real option. Or getting a dedicated instance from OpenAI I guess.
cassianoleal 3y ago
If you use VS Code or a JetBrains IDE, Continue works well with Ollama and it’s really easy to get going. [0] https://continue.dev/ https://continue.dev/ [1] https://ollama.ai/ https://ollama.ai/
- unshavedyak 3y ago
  Any opinion on what the best experience is, currently?
  - cassianoleal 3y ago
    What do you mean by experience? If you mean the model, I’ve been happy with Deepseek Coder. Mistral is a popular alternative as well.
    - espe 3y ago
      yes i'd second deepseek-coder and check out wingman. it's a relatively new extension, snappy and with good ux.
      - cassianoleal 3y ago
        Is this the one? https://marketplace.visualstudio.com/items?itemName=nvms.ai-wingman https://marketplace.visualstudio.com/items?itemName=nvms.ai-... There's another one called Wingman Copilot as well: https://marketplace.visualstudio.com/items?itemName=WingmanCopilot.wingman-copilot https://marketplace.visualstudio.com/items?itemName=WingmanC...
        espe 3y ago
        yes, it's the nvms one.
    - unshavedyak 3y ago
      By experience, i mean the whole UX - which is as much the interface to the LLM as it is the LLM itself, imo. Ie i'm not terribly interested in constantly telling an LLM what i want, prodding it to do the right thing, etc. I'm more interested in ways it can be easy and correct - even if that's limited in functionality. Ie perhaps automatically suggesting docs, knowing what style to use, etc.
      - 3y ago

deergomoo 3y ago

> My company has approved copilot but Copilot autocomplete has been an awful experience I had the same experience, I feel like I must be crazy because so many of my colleagues have been singing its praises. I found it immensely distracting and disabled it again after a couple of days. It was like having someone trying to finish my sentence while I was still speaking; even when they were right, it was still annoying and knocked me out of my flow (and very often, it wasn’t right).

fipar 3y ago
I actually find copilot quite useful but I use it from emacs and it only provides suggestions when I intentionally hit my defined shortcut for it, so it never gets in the way. It may be worth for you to try setting it up in a similar way in the tool you’re using it from, as I agree I’d find the experience awful if it was always trying to autocomplete my sentences.

moffkalast 3y ago

Fwiw, there are now some local models that rival 3.5-turbo in code chat, like the Codeninja I tried out the other day. Not nearly as good as 4 which iirc runs the copilot backend, but for sensitive data that can't leave the premises it's the only real option. Or getting a dedicated instance from OpenAI I guess.

cassianoleal 3y ago

If you use VS Code or a JetBrains IDE, Continue works well with Ollama and it’s really easy to get going. [0] https://continue.dev/ https://continue.dev/ [1] https://ollama.ai/ https://ollama.ai/

unshavedyak 3y ago
Any opinion on what the best experience is, currently?
- cassianoleal 3y ago
  What do you mean by experience? If you mean the model, I’ve been happy with Deepseek Coder. Mistral is a popular alternative as well.
  - espe 3y ago
    yes i'd second deepseek-coder and check out wingman. it's a relatively new extension, snappy and with good ux.
    - cassianoleal 3y ago
      Is this the one? https://marketplace.visualstudio.com/items?itemName=nvms.ai-wingman https://marketplace.visualstudio.com/items?itemName=nvms.ai-... There's another one called Wingman Copilot as well: https://marketplace.visualstudio.com/items?itemName=WingmanCopilot.wingman-copilot https://marketplace.visualstudio.com/items?itemName=WingmanC...
      - espe 3y ago
        yes, it's the nvms one.
  - unshavedyak 3y ago
    By experience, i mean the whole UX - which is as much the interface to the LLM as it is the LLM itself, imo. Ie i'm not terribly interested in constantly telling an LLM what i want, prodding it to do the right thing, etc. I'm more interested in ways it can be easy and correct - even if that's limited in functionality. Ie perhaps automatically suggesting docs, knowing what style to use, etc.
    - 3y ago