|
smackfu posted:I saw a horrifying view of the future in a Reddit question the other day: I used to have a difficult time believing that people could be too dumb to do google searches but watching people try to use chatgpt as a resource and failing at it somehow reduces my already extremely shaky faith in humanity even more.
|
# ? Mar 26, 2024 13:07 |
|
|
# ? May 4, 2024 14:25 |
|
Bruegels Fuckbooks posted:I used to have a difficult time believing that people could be too dumb to do google searches but watching people try to use chatgpt as a resource and failing at it somehow reduces my already extremely shaky faith in humanity even more. ? Maybe I'm the idiot, but the free chat gpt's knowledge was frozen a while ago, and it routinely confuses less known stuff with well known stuff, like confusing Polars dataframes with Pandas. The person deserves to be laughed at for being too reliant on chat gpt, but it's a legit problem with the tool
|
# ? Mar 26, 2024 13:53 |
|
I guess the straight from OpenAI stuff might be better. I wouldn't know; I closed my account once and they don't let you reopen it ever so I'd have to get a new phone number (Google Phone does not work) to open another one. However, the Bing GPT is absolute dogshit; it just repeats the summary of the top three search results your query would get. I can do that. Asking it to come up with something original so easily runs afoul of one of their many filters. I'm not even asking it to write porn. I told it to give me a witty summary of Das Kapital and it told me that jokes about political topics could offend someone. Sometimes it will write the answer to something completely innocuous, come across a naughty word in its own answer, and then delete the entire thing in front of your eyes.
thotsky fucked around with this message at 14:09 on Mar 26, 2024 |
# ? Mar 26, 2024 14:06 |
|
the most notable thing AI has done for me is that one of my coworkers uses it to generate the most loving worthless pull request descriptions that just itemize every change he made in a file with zero context on why the hell he was changing anything love to read the equivalent of SEO spam drivel in PRs and JIRA tickets
|
# ? Mar 26, 2024 15:41 |
|
StumblyWumbly posted:? Yes, the tool has many problems. So does ChatGPT even
|
# ? Mar 26, 2024 15:54 |
|
My job just released automatic AI analysis on PRs. Sometimes it can give context to something I don't totally understand but it mostly just says in five sentences what could be said in one. Its full of padding, like when a college student is trying to hit a page minimum. I can see its usefulness somewhere in the future. I can also see myself having to explain to juniors why something the AI says is wrong a million times.
|
# ? Mar 26, 2024 16:04 |
|
abraham linksys posted:the most notable thing AI has done for me is that one of my coworkers uses it to generate the most loving worthless pull request descriptions that just itemize every change he made in a file with zero context on why the hell he was changing anything Murder is permissable in this scenario
|
# ? Mar 26, 2024 17:19 |
|
The current state of AI is having a junior dev you have to watch and review. It can be useful. But you can’t just accept what it puts out without editing.
|
# ? Mar 26, 2024 18:37 |
|
ChickenWing posted:This remains the single bastion of good code on the internet How do we monetize this?
|
# ? Mar 26, 2024 20:43 |
|
Love Stole the Day posted:How do we monetize this? Like so: code:
|
# ? Mar 26, 2024 20:46 |
|
So I was an early employee at a startup with an abusive CEO, I built the entire first version of their product, and then I was fired. Now, two years later, I’m watching them slowly launch. And it hurts knowing that they might succeed with a product that I poured so much sweat into. I don’t want to be bitter, but it’s hard sometimes.
|
# ? Mar 26, 2024 22:40 |
|
lifg posted:The current state of AI is having a junior dev you have to watch and review. It can be useful. But you can’t just accept what it puts out without editing. this will be the state of ai forever, barring some radical new approach. get used to it
|
# ? Mar 27, 2024 00:44 |
|
I haven't used AI at all because I get paid by the hour.
|
# ? Mar 27, 2024 03:07 |
|
redleader posted:this will be the state of ai forever, barring some radical new approach. get used to it Basically this. Also, it likes to hallucinate just as much as everything else in AI. i've gotten some mileage out of Copilot, but it's not ever going to solve 'big problems'. It's going to write out some LINQ that I don't have to construct myself (which is nice).
|
# ? Mar 27, 2024 03:14 |
|
lifg posted:So I was an early employee at a startup with an abusive CEO, I built the entire first version of their product, and then I was fired. Now, two years later, I’m watching them slowly launch. And it hurts knowing that they might succeed with a product that I poured so much sweat into. I don’t want to be bitter, but it’s hard sometimes. Worst case scenario, I hope you can at least put it on your resume and say in interviews that you built it from the ground up.
|
# ? Mar 27, 2024 15:35 |
|
What's the term for not-actually-gaslighting but engineers being repeatedly blocked from sending working and validated code to production, having release mechanisms where individuals don't use approved PRs but decide they want to reinvestigate all the code and make new complaints, being unable to follow industry best practices and standards for engineering such as proposals/designs/documentation/prototype sharing/bugfix cycles and instead needing to revert working solutions and regress projects, engineers needing to go to leads to explain how to improve the planning process and then (being told it's fine as is but there is still an unmeasured expectation of "completion") needing to demonstrate basic task ownership/grooming/planning...? The most recent drop-dead-laughable was an engineer that shipped the code and, the next steps being privileged-only on-host script execution, just for staging verification mind you, assigned to the lead with the steps, was reassigned the ticket by the lead, eng returned after a one day absence to find the work not yet done, reassigned to another lead with more familiarity, who was on vacation but logged in to reassign it to the first lead, who reassigned it to the engineer again, who wrote up instructions again (the first being buried in two pages of needless noise) and reassigned it to a privileged manager. Then the first lead jumped back on the ticket to dissect the second set of instructions, with a paragraph per step (either reiterating what the engineer said the first time, or complaining about where the steps were recorded in jira, not sure I ran out of time to decipher their mess). All this for what amounts to less than five minutes of actual work. Unfortunately I'm not in a position to fix all these situations, but the calendar is going to soon be full of meetings getting these confused people into a room every time they need to actually behave like leads and help their engineers release code.
|
# ? Mar 28, 2024 01:56 |
|
The term is "being in hell" and I'm hoping you escape
|
# ? Mar 28, 2024 02:06 |
|
that's called "our process"
|
# ? Mar 28, 2024 02:13 |
|
That all sounds completely insane, where does everyone find these places?
|
# ? Mar 28, 2024 02:21 |
|
PhantomOfTheCopier posted:What's the term for not-actually-gaslighting but “We’re agile!” in the job posting.
|
# ? Mar 28, 2024 02:29 |
|
ElehemEare posted:“We’re agile!” in the job posting. (* The spelling, not the process.)
|
# ? Mar 28, 2024 02:41 |
|
https://x.com/tekbog/status/1773079314453426365?s=20
|
# ? Mar 28, 2024 04:50 |
|
https://archive.org/details/SimpleSabotageFieldManual/mode/1up Modern day business practice and sabotage is just 2 circles on top of eachother.
|
# ? Mar 28, 2024 06:44 |
|
Are you familiar with how dogs will piss on the same tree to show they're on top? Pull Requests work the same way
|
# ? Mar 28, 2024 17:38 |
|
StumblyWumbly posted:Are you familiar with how dogs will piss on the same tree to show they're on top? Imagine four developers pissing off the edge of a cliff, development works the same way
|
# ? Mar 28, 2024 18:27 |
|
Ah the tyranny of the short sighted mba program. Raise your hand if you've ever used documentation for a programming language? (100% minus the liars). And if you use any source that's written text during software work at least once a week? (Yeap SO is centralized code comments and cookbooks). And if you've ever onboarded anywhere and learned anything reading a document? Whelp... Then ensuring everyone must come to you for knowledge must just be job security.
|
# ? Mar 28, 2024 23:34 |
|
I've never read documentation, just stackoverflow posts
|
# ? Mar 29, 2024 01:03 |
|
I only read chat-gpt answers trained on stack overflow posts.
|
# ? Mar 29, 2024 02:23 |
|
Cup Runneth Over posted:I've never read documentation, just stackoverflow posts I do read the documentation, after everything else failed.
|
# ? Mar 29, 2024 02:24 |
Xarn posted:That all sounds completely insane, where does everyone find these places? The probably of this being your place of work hits an inflection point the moment you hire a project/product manager and increases exponentially with every new hire from there. Each addition to the PMO increases the exponent
|
|
# ? Mar 29, 2024 04:34 |
|
PhantomOfTheCopier posted:Ah the tyranny of the short sighted mba program. Raise your hand if you've ever used documentation for a programming language? (100% minus the liars). And if you use any source that's written text during software work at least once a week? (Yeap SO is centralized code comments and cookbooks). And if you've ever onboarded anywhere and learned anything reading a document? Writing poo poo down isn't inherently good or bad. Writing useful poo poo down that other people will read is very useful. Writing a summary of your notes from the meeting you had to discuss the potential future possibility of changing the button colour from green to dark green is not very useful. Also I wrote a shitload of documentation on my last team and I'm pretty sure I'm the only person that reads it, which I do to refresh myself when someone comes to me with questions they wouldn't have if they read any of the documentation I wrote.
|
# ? Mar 29, 2024 04:46 |
|
ChickenWing posted:The probably of this being your place of work hits an inflection point the moment you hire a project/product manager and increases exponentially with every new hire from there. Each addition to the PMO increases the exponent Possibly, but good PMs/PgMs will act as bullshit umbrellas in the same way that good managers do.
|
# ? Mar 29, 2024 05:44 |
|
Docs are only good if they're well written and mostly up to date. And lots of people are poo poo at writing, especially technical writing, or just don't care. If your team has a process for something that's done manually with humans, it should be well documented so that it's repeatable and understandable by someone new after the 1st or 2nd walk through by someone experienced. Even better if a complete newbie can follow it. Making docs this way requires a team that has the expectation of good docs. IMO this should be the norm in a professional setting. You can keep docs updated by following them while you do the task. Just update bits that are unclear or outdated as you work through them. You don't have to open them up every time, but once in a while is probably enough.
|
# ? Mar 29, 2024 12:51 |
|
a dingus posted:You can keep docs updated by following them while you do the task. Just update bits that are unclear or outdated as you work through them. You don't have to open them up every time, but once in a while is probably enough. Yeah, in our team we luckily don't have much manual bullshit to do, but the manual bullshit duty rotates each week and the guidance is to follow the instructions on the wiki, and if there's anything unclear to ask the senior people for more info and then amend the docs accordingly. This has kept the docs reasonably coherent. If even the newer/more junior people can figure it out then the docs are good, if not the docs get improved.
|
# ? Mar 29, 2024 15:26 |
|
I got a "we just went through them while onboarding (Coworker), so they should be pretty up-to-date" in regard to some setup docs. Then I tried to follow them and they were poo poo. Then I didn't fix them
|
# ? Mar 29, 2024 15:42 |
|
CPColin posted:I got a "we just went through them while onboarding (Coworker), so they should be pretty up-to-date" in regard to some setup docs. Then I tried to follow them and they were poo poo. Then I didn't fix them Having a new person write down all the places they got stuck and where the docs deviated from what they got told to actually do has been a pretty good way to audit onboarding instructions.
|
# ? Mar 29, 2024 21:09 |
|
Volmarias posted:Having a new person write down all the places they got stuck and where the docs deviated from what they got told to actually do has been a pretty good way to audit onboarding instructions. Yeah, I thought that was what they meant when they said they'd just gone over everything, but nope!
|
# ? Mar 29, 2024 22:29 |
Volmarias posted:Possibly, but good PMs/PgMs will act as bullshit umbrellas in the same way that good managers do. True but that's not as funny to post
|
|
# ? Mar 30, 2024 00:19 |
|
Volmarias posted:Having a new person write down all the places they got stuck and where the docs deviated from what they got told to actually do has been a pretty good way to audit onboarding instructions. I've also seen this as a good canary for identifying high potential newbies. The ones who cannot communicate about when/where they are stuck, or end up with tons of feedback on their documentation update diffs are the ones that have empirically worse outcomes. Even when the actual work is solid, not being able to communicate results can be limiting (to their individual career, but also project outcomes).
|
# ? Mar 31, 2024 23:47 |
|
|
# ? May 4, 2024 14:25 |
|
Most documentation is only useful if people read it, and companies like Atlassian make it very hard to understand how people are engaging with content on their platforms. Think about why, even though tracking views is one of the easiest and cheapest features they could possibly implement, it's in their best interest not to surface this information
|
# ? Apr 1, 2024 13:53 |