|
Volmarias posted:I'm not 15 and work slack isn't ICQ. i feel like if my default way of existing is "you can expect a fairly quick response from me on slack unless my status says otherwise" then it will encourage people to slack me when they need a fairly quick response. if my default way of existing is "you can never know when to expect a fairly quick response from me on slack" it will encourage people to call me, which i very much want to avoid
|
# ? Mar 12, 2024 16:11 |
|
|
# ? May 23, 2024 16:26 |
|
prom candy posted:i feel like if my default way of existing is "you can expect a fairly quick response from me on slack unless my status says otherwise" then it will encourage people to slack me when they need a fairly quick response. if my default way of existing is "you can never know when to expect a fairly quick response from me on slack" it will encourage people to call me, which i very much want to avoid Your main issue here is having a phone number available to your coworkers, or even having a phone available. In this, I cannot help you.
|
# ? Mar 12, 2024 16:36 |
|
Think he just means call on slack (or whatever VoIP they use)
|
# ? Mar 12, 2024 16:51 |
|
But if they are not there to answer the message, who is there to receive the call?
|
# ? Mar 12, 2024 16:58 |
|
They are probably assuming you are there, just not responding. Slack is actually bad for this because it has no notification for a missed call. I'll hear ringing from the study and by the time I get back from the toilet who called me
|
# ? Mar 12, 2024 17:02 |
|
Mega Comrade posted:Sadly that's exactly why, we have a bunch of legacy and often there's one person who knows why X has broken in prod. prom candy posted:setting a slack status when you're going to be a away for a while is a courtesy, i don't really mind doing it To be clear: it's not setting a status that I find a problem. I agree, it's a nice courtesy, I don't mind doing it. It's useful to give folks an expectation that you won't reply for a while if you know that you won't. What I said "no thanks" to is the "telling off" for not doing so, and the implied expectations around responsiveness on Slack, and the further implied work culture/norms. I'm not saying even that they're the wrong choice for whatever your organizational circumstances are, just that that is a work environment I don't, personally, want to be in. I'd rather slack be treated as the asynchronous tool that it is; I'm generally on top of responding to things but at the same time, often I want to have an hour or two of focus time (so, yeah, I don't want to have it be a problem if I don't respond within 2 hours, whether it's because I'm heads down or walking my dog). If something is truly urgent, page me. If such things are so frequent that doing so feels unreasonable, then, again, that's a work environment I don't care to be in. And from my perspective, perhaps indicates a lack of structure or discipline.
|
# ? Mar 12, 2024 17:46 |
|
Mega Comrade posted:Slack is actually bad
|
# ? Mar 12, 2024 18:49 |
|
Slack is loving terrible and I can't believe I'm actually excited to move to Teams. It's not like Teams is much better, but it's a step up.
|
# ? Mar 12, 2024 18:57 |
|
I just block people if they suck on chat and make them email me if they want help so I have documentation. I used to dread having chat open because it would be entire days of "Hi" and questions, but it turns out that answering asinine questions over chat is not in my job description and is not connected to my salary in any tangible way.
|
# ? Mar 12, 2024 19:01 |
|
Canine Blues Arooo posted:Slack is loving terrible and I can't believe I'm actually excited to move to Teams. It's not like Teams is much better, but it's a step up. You have brain worms for real. Why on earth would this be an upgrade?
|
# ? Mar 12, 2024 19:19 |
|
Imagine you have to use both. And although 90% of your job is in teams, some outside contractors use slack and we are required to monitor it. End result: Nobody monitors slack except my boss who then angrily notifies us we're needed in some slack discussion.
|
# ? Mar 12, 2024 19:22 |
|
I don't think teams is better. The problem is that chat apps are used far too often in places where email would be the better choice
|
# ? Mar 12, 2024 19:55 |
|
I prefer slack for messages. Although it winds me up to no end how people won't respond to the actual thread and instead message the channel. But that's a user problem. Huddles are garbage though and teams is better at VoIP.
|
# ? Mar 12, 2024 20:06 |
|
Teams is the second best at it's three main features. So yes, slack is a better message platform, but Teams is a better calling/meeting/collaboration platform. Yes, Zoom is a better meeting platform, but Teams is a better etc.
|
# ? Mar 12, 2024 20:10 |
|
Teams kills conversation. I’ve been in two workplaces that switched from Slack to Teams and in both the small talk about code and software and the company just withered away. A lot of gems appear in small talk, it’s a real loss when it disappears.
|
# ? Mar 12, 2024 21:40 |
|
Teams is the worst chat program I've ever used, and friends, I've used some real stinkers. It's the only one that has repeatedly failed at the basic function of delivering messages to me.
|
# ? Mar 12, 2024 21:56 |
|
I for one welcome... Wait I'll restart: How have things changed when your team has been approved to use ChatGPT+similar for software questions? I realize that its sources are not dissimilar from stack overflow, but based on a past run through I did with someone, we found its solutions rather incomplete and non-functional. What I foresee is a notable change in effort: Those who use AI posting more code (LOC meh), while reviewers spend much more time explaining that the submitted algorithms are an utter mess. AI! Make lazy people lazier, and productive people overworked remedial software educators!
|
# ? Mar 12, 2024 22:18 |
|
PhantomOfTheCopier posted:I for one welcome... I was working with a junior a few years ago and in a PR I told him it's okay to get code from Stack Overflow, everyone does it, but it's not okay to copy-paste code from StackOverflow without understanding what it's doing. I think LLMs make it way easier to copy-paste code without understanding what it's doing, it's basically like a StackOverflow where all your variable names are plugged in already. LLMs are great, I use them a ton for my work, but blindly trusting the output is as bad or worse than blindly trusting StackOverflow. And at least with StackOverflow you have to go through the mental exercise of adapting the code to fit your context.
|
# ? Mar 12, 2024 22:27 |
|
My favorite feature of github co-pilot is adding ``` to the end of every suggestion
|
# ? Mar 12, 2024 22:31 |
|
Do Users Write More Insecure Code with AI Assistants?quote:We conduct the first large-scale user study examining how users interact with an AI Code assistant to solve a variety of security related tasks across different programming languages. Overall, we find that participants who had access to an AI assistant based on OpenAI's codex-davinci-002 model wrote significantly less secure code than those without access. Additionally, participants with access to an AI assistant were more likely to believe they wrote secure code than those without access to the AI assistant. Furthermore, we find that participants who trusted the AI less and engaged more with the language and format of their prompts (e.g. re-phrasing, adjusting temperature) provided code with fewer security vulnerabilities. Finally, in order to better inform the design of future AI-based Code assistants, we provide an in-depth analysis of participants' language and interaction behavior, as well as release our user interface as an instrument to conduct similar studies in the future.
|
# ? Mar 12, 2024 22:43 |
|
Gee that couldn't possibly be because the vast majority of searchable examples out there do poo poo like use wildcards for everything for the sake of brevity.
|
# ? Mar 12, 2024 22:51 |
|
Oooo thank you! I'd even go so far as to believe that languages more discussed online have fewer AI blunders; ie they can filter all the irrelevant SO answers more quickly than a human reading. Alas, I suspect accuracy drops off precipitously.
|
# ? Mar 13, 2024 06:31 |
|
Che Delilas posted:Gee that couldn't possibly be because the vast majority of searchable examples out there do poo poo like use wildcards for everything for the sake of brevity. we can simply train an ai only on good code and how do we identify that? well, we first have to train a separate ai to be able to identify good code,
|
# ? Mar 13, 2024 11:18 |
|
I could help by providing 20+ years of crappy code.
|
# ? Mar 13, 2024 11:46 |
|
PhantomOfTheCopier posted:Oooo thank you! This is definitely true. Basic chatgpt will also confuse recent major versions of libraries, IIRC this was an issue with Pydantic. E it also has some confusion for pandas dataframes vs polars dataframes
|
# ? Mar 13, 2024 12:51 |
|
redleader posted:we can simply train an ai only on good code there is no good code
|
# ? Mar 13, 2024 22:59 |
|
All code is error
|
# ? Mar 13, 2024 23:01 |
|
Blinkz0rz posted:there is no good code the good code is highlighted red in the pr diff
|
# ? Mar 14, 2024 01:08 |
|
The true pareto chart: 500 lines per hour (lph) of production ready code, tested and validated, performance and memory profiled 350 lph after explaining software best practices to paired peer 97 lph after manual style guide application 40 lph after team lead takes a look and decides "we don't need helper functions, separation of responsibility, testable code" 35lph after explaining to team lead: SOLID, basic encapsulation, and how code is not easy to read if names are global across a 1M line codebase 22lph after team lead asks two or three other leads to pile on with their opinions 14lph after deleting all of the above so the code can get "approved for release" sometime in the next couple weeks 11lph after two last attempts to hit the Yes/No black box that "owns" the code but can't describe what they want, only how things aren't what they'd write 8lph after the reviewer finally gets around to looking at things that matter in the code, and simple changes are now exceedingly difficult because all of the above made the solution unmaintainable and not clean code 5lph from having to redo all tests and profiling half a dozen times 4lph after recovering from the liquor+drugs necessary to destroy the last semblance of your creativity
|
# ? Mar 14, 2024 01:33 |
This remains the single bastion of good code on the internet
|
|
# ? Mar 14, 2024 21:27 |
Thread, today I honest-to-god read the documentation for a library update that I wanted to make sure I was using right And then I went back to just searching directly for the answers to what I wanted to do because the documentation still didn't tell me how to do it Reading docs properly is for suckers, search for snippets erryday
|
|
# ? Mar 14, 2024 21:30 |
|
ChickenWing posted:Thread, today I honest-to-god read the documentation for a library update that I wanted to make sure I was using right * Some limitations may apply.
|
# ? Mar 15, 2024 06:00 |
|
PhantomOfTheCopier posted:Wait I'll restart: How have things changed when your team has been approved to use ChatGPT+similar for software questions? I have a colleague in another part of the organization that is migrating a bunch of code over to Rust. The problem is that he has quite a bit of boilerplate code and had to jump straight into the deep end with macros. I think our internal-secret-safe AI thing has written half of that. A lot of the rest that I've seen--including me here--has been getting little shell script snippets. Shell scripting is a hellish combination of goofy syntax along with magic Linux knowledge that you won't get good at unless you do it every day, and I don't do it every day. It has been good there. I got my first screwup with it recently. I tried to get a C example for getting the arguments that launched the process on Linux if I didn't have access to argv. I did learn about /proc/self/cmdline, but the example it gave to read it was wrong. Given that C is a memory mismanagement language that sometimes is used to write software, it seeked to the end of the file first to get a buffer for building up the string. This doesn't work with that file--presumably because it's not a real file. When I called it out on it, it didn't have a better answer and mostly insisted I ensure I'm running on Linux. I tried to give it an announcement I was going to send to correct for spelling, grammar, style, and professionalism. It gave me something that, within half a sentence, started sounding like Martin from The Simpsons in my head. Other folks have received notes and announcements that they knew were run through that, and one is pretty sure they got one that was completely written by it. On the other hand, I've loved asking it my crazy questions because it doesn't drag me into an X-Y interrogation. It'll just gently caress up the answer.
|
# ? Mar 17, 2024 04:19 |
|
ChatGPT has been a godsend for generating CMake files and answering questions about that. Also C++ related questions Honestly because Google has such poo poo results now you can't really turn to it for specific advice. I basically ask ChatGPT and if I'm not sure how right it is I have more specific keywords to lookup on Google that hopefully lead me to the right place. I don't like Copilot as much because it uses live URLs and what it spits back is often based off answers in Stack Overflow. Like it'll repeat a lot of stuff like, "I don't know if there are better solutions" and when I click on the references there it's a link to SO where someone commented that exact phrase.
|
# ? Mar 17, 2024 08:36 |
|
I quite like Copilot, but I found it doesn’t work if there aren’t existing examples of code. Like when I was working with Rust, and using portions of a library that apparently no one had ever used in a public repo (and I did a deep search in Google and GitHub), Copilot just gave up. I wonder if ChatGPT would have been better.
|
# ? Mar 17, 2024 13:31 |
|
Thanks all. I think we'll know in a couple weeks who takes this opportunity to become even more lazy.Rocko Bonaparte posted:... A friend was working with her daughter on some translations and it was invaluable. They still had to correct it but it presumably did much better than standard translation tools.
|
# ? Mar 18, 2024 06:30 |
|
Just catching up on the last few pages of this thread, and wow, there's a lot of people confusing their burnout coping strategies for general approaches to well-being, and you might want to reflect on why the two aren't different to you
|
# ? Mar 18, 2024 18:18 |
|
I saw a horrifying view of the future in a Reddit question the other day: “Does anyone have any good resources on software version n-1? ChatGPT doesn’t know about version n yet.”
|
# ? Mar 25, 2024 13:34 |
|
smackfu posted:I saw a horrifying view of the future in a Reddit question the other day: i dunno seems like an opportunity to read up on version N of the software and charge large consulting fees for the company to be able to use that version.
|
# ? Mar 25, 2024 21:30 |
|
|
# ? May 23, 2024 16:26 |
|
Just paste the entire docs into ChatGPT before your question
|
# ? Mar 26, 2024 06:29 |