3 ms·

Finally?

refulgentis 3y ago
There's a persistent rather large minority that has a nuanced take: it can't write code they like (don't want to edit), but it's great for weekend projects (where they're trying new things without established personal preferences). Forest for the trees if you ask me, but, to each their own.
- skissane 3y ago
  Sometimes, I find writing pseudocode easier than code. And then I ask an AI to turn it into code. Sometimes the results aren’t too bad, and just need a few tweaks for me to use it-overall I’ve saved mental effort compared to translating the pseudocode into code by hand. And if the results aren’t useful, I’ve only wasted a few seconds, and then I just have to do it manually.
huytersd 3y ago
[flagged]
- plufz 3y ago
  I don’t know if I’m slow but I try to use ML for dev work but I have a hard time using it to be productive. Maybe it is because I don’t want to use GPT and copilot but I rather use local codellama, mistral, dolphin, etc. But I have found very limited use for it. I use ML for other things like transcriptions with whisper, translations with m2m100 and summerizing articles with custom python scripts and different models. Not even the simplest things like convert this Apache conf to nginx works in a satisfactory way. It can’t solve bugs. It can’t help me with like edge cases for weird libraries etc. It’s like it only can help me with very obvious things that is quicker to just write myself. How do you use it?
  - huytersd 3y ago
    I don’t personally use it but my team is atleast 5x faster at what they do (validated by sprint output, up to 130 points for a two week sprint! Previous completion rate was between 25-30 points) and seem to be able to relax more (also validated in 1-on-1s). I even use it to estimate story points before a backlog planning meeting with pretty good correlation to what the devs come up with. I’ve done spot code reviews and I see next to no drop in levels of code quality. All our human written tests (I hope, because that’s what we mandate) have seen no uptick in failure rate for new checkins.
    - plufz 3y ago
      Five times faster sounds almost too good to be true. Could you give some concrete examples of prompts? Is this senior or junior devs? I would love to hear more.
      - huytersd 3y ago
        Mix of junior and senior. Our stack is react, node, AWS. Telehealth domain. I would post jira screenshots if I was at work, it really is amazing. I’m not sure what prompts the engineers are using but from conversations I think they begin by feeding in the entire jira ticket (minus confidential information) as a prompt which has a goal, acceptance criteria etc. listed.
        plufz 3y ago
        Amazing that the seniors are more productive as well. Have your team read any specific articles or other sources to learn how to do useful things within LLM for code? I really want to learn haha. Which LLM do you use?

Finally?

There's a persistent rather large minority that has a nuanced take: it can't write code they like (don't want to edit), but it's great for weekend projects (where they're trying new things without established personal preferences). Forest for the trees if you ask me, but, to each their own.

skissane 3y ago
Sometimes, I find writing pseudocode easier than code. And then I ask an AI to turn it into code. Sometimes the results aren’t too bad, and just need a few tweaks for me to use it-overall I’ve saved mental effort compared to translating the pseudocode into code by hand. And if the results aren’t useful, I’ve only wasted a few seconds, and then I just have to do it manually.

huytersd 3y ago

[flagged]

plufz 3y ago
I don’t know if I’m slow but I try to use ML for dev work but I have a hard time using it to be productive. Maybe it is because I don’t want to use GPT and copilot but I rather use local codellama, mistral, dolphin, etc. But I have found very limited use for it. I use ML for other things like transcriptions with whisper, translations with m2m100 and summerizing articles with custom python scripts and different models. Not even the simplest things like convert this Apache conf to nginx works in a satisfactory way. It can’t solve bugs. It can’t help me with like edge cases for weird libraries etc. It’s like it only can help me with very obvious things that is quicker to just write myself. How do you use it?
- huytersd 3y ago
  I don’t personally use it but my team is atleast 5x faster at what they do (validated by sprint output, up to 130 points for a two week sprint! Previous completion rate was between 25-30 points) and seem to be able to relax more (also validated in 1-on-1s). I even use it to estimate story points before a backlog planning meeting with pretty good correlation to what the devs come up with. I’ve done spot code reviews and I see next to no drop in levels of code quality. All our human written tests (I hope, because that’s what we mandate) have seen no uptick in failure rate for new checkins.
  - plufz 3y ago
    Five times faster sounds almost too good to be true. Could you give some concrete examples of prompts? Is this senior or junior devs? I would love to hear more.
    - huytersd 3y ago
      Mix of junior and senior. Our stack is react, node, AWS. Telehealth domain. I would post jira screenshots if I was at work, it really is amazing. I’m not sure what prompts the engineers are using but from conversations I think they begin by feeding in the entire jira ticket (minus confidential information) as a prompt which has a goal, acceptance criteria etc. listed.
      - plufz 3y ago
        Amazing that the seniors are more productive as well. Have your team read any specific articles or other sources to learn how to do useful things within LLM for code? I really want to learn haha. Which LLM do you use?