I do not even check what code Claude generates anymore
>>110118696the Great Brainrot of your time begins.
>>110118696nobody has for months now, it shows. draw.io fucking broke on firefox the snap grid scales with your zoom level
>>110118696Okay. If I were your tech lead, I'd have a conversation with the bosses. Just saying.
>>110119022>you're wasting our time and fired. Please send in limp bizkit so we can let him know about his promotion
>>110118696>I do not even check what code Claude generates anymoreThis will lead to massive fuckups without human review. Claude can process things faster and more intelligently than humans but ultimately it's not sapient and will make mistakes. Maybe it will make less mistakes than humans but it will still lead to mistakes from time to time. The main issue I see is that people will become retardedly reliant on it and it will degrade their ability to diagnose or see issues when they do pop up, but hey I guess that leaves a very niche market for low level tech goblinoids that are actually valuable to solve niche cases.
>>110118696Many such cases. Sad!
>>110120557Actual pseudonym: limpy.
>>110118696I wish I had an idea so original that even these chat butts couldn't copy it from somewhere.
>>110120588>Claude, fix all of the mistakes you made over the last year
>>110120557Wouldn't happen. Someone who uses AI and that's it is useless and is dragging everyone down.
>>110121901I fucked up. I meant someone who only uses AI is useless and is dragging everyone down. Since I fucked up the reason is AI does very verbose code that needs a lot of patience from other devs and will be rejected constantly if you know what you're doing. Someone like OP believes he's doing good enough until someone is fed up and asks to get out of the project.
just make claude also write tests it's fine i think
>>110122027Absolutely, tests are always catch all edge cases and that's even more true if an SI (formerly AI) designs the test.We are entering the bug free era of computers programs.
>>110122090I've never had a bug in anything I've vibecoded.
>>110122119Anything even remotely complex has bugs, unless you confidently say all edge cases where behavior changes drastically are features.
>>110122195>Anything even remotely complex has bugsThis is reflective of your own ability rather than something intrinsic to complex programs.
>>110122237No, it's not. It's rather reflective of your ignorance concerning data science.
>>110122250>every complex program is data scienceYeah, you're full of shit.
>>110122262If you think your program is programmed without any bugs or possible exploits then you are simply wrongEven if we pretend you programmed it perfectly you'd still need to trust the computer/interpreter and every other part of the chain assume it works exactly. The same all on systems and that you can predict how the compiler compiles your program.You can't be real.
Frick you i won't do what you tell me!!!!love linken the park man
>>110122307Just because you can’t write code that isn’t riddled with bugs doesn’t mean no one can, Rajesh.
>>110122307>If you think your program is programmed without any bugs or possible exploits then you are simply wrongCan you prove that? You know absolutely nothing about me. As stated previously, >>110122119>Even if we pretend you programmed it perfectly you'd still need to trust the computer/interpreter and every other part of the chain assume it works exactly.That's why you vibecode tests.>The same all on systems and that you can predict how the compiler compiles your program.Tests also cover this.
>>110122343How do you know your tests cover all use cases and all edge cases even if you rigorously make them yourself?How do you know your LLM of choice is as rigorous?
>>110122338Tell me what you've made, link me a project or tell me what corporate shitware you've made.
Who people check the code?
>>110122425Yes
>>110122375The LLM makes sure.
>>110121901Buddy, I was told to quit wasting company resources (i.e. my time) touching code or even doing human PR/documentation just so I could be familiar with the code.>If you don't trust the code, you didn't do your job on the acceptance ceiteria. When it fails, we do not touch code. We fix the prompt/output/test pipeline.You have zero idea of how much this AI psychosis has effected leadership. They envision a future with 0 devs, 0 creatives, 0 devops, just their pipeline and a bunch of VPs/product managers with checkists of features sourced directly from customers.
>>110122425Them.
>>110122605Have you considered the point of contention is that not all of us work for retard slave goycattle
>>110122605man that's going to produce a lot of unusable garbage no one wantsand by the time they figure this out it'll be too late for them
>>110122638I mean, it's possible. But given how much you've let this place influence your rhetoric, I doubt it.
>>110122650You mean for the teams they fire, and then maybe have to hire back later. Except probably not if my team is anything to go by. They're planning to rebuild from Canadian GTS/TFWP because h1b got too expensive stateside.
>>110122605>>110122650You guys are ignoring the unfortunate reality of diminishing results. Every ehnacement offered at the top past a certain point is prohibitively expensive. But moving in the opposite direction means dropping offering a prduct with slightly shittier features or poor quality or reliability (in markets that tolerate it) can net you significant savings to undercut your competitors. Enshitification works. It's the future, whether we like it or not.
>>110122250>data sciencethis is the single most jeet coded made up term in existence.
>>110118696I just have gpt check claude then have claude check what gpt says then put that back into gpt to check it then I have copilot rate what they wrote /10 then I have claude do a final pass. I really need to have them make me a script to do this for me...
>>110119022jokes on you, I am the tech lead!
this only works in jeet sweatshops.everywhere else they force you to do manual PR's, cuz coders are lazy as shit by default and would skip it.
me too
I ask gemini free which libraries i should use for my vibeslop projects, and then let qwen flash next assemble and debug the shit locally over ssh
what are you going to do when you forget how to code?
Learn to code pfffffft hahahahahahahahaha learn to code am I right hahahahahhahahahaha
>>110124866Jokes on you, i never learned how to code. i'm not an autist
>>110122650>going to produce a lot of unusable garbage no one wantsso nothing is going to change?
>>110120588Claude 6 will fix it
it's actually based, you will create so much techdebt slop further token usage on the project will grow exponentially till corpos start limiting shit more again
>>110124955>weekly token usage limit reached>summary of fixes: three typos and one line removed (0.0001% done)
Anthropic just lost 80bn in three months bozo. It's game over.
>>110124996Every piece of software will simply be rebuilt from scratch every year
Same. If I need another set of eyes I share the conversation with another chatbot.
>>110125006Except if you lost this much you’d be in jail. Anthropic will just keep borrowing and increasing prices. It’s too addictive for people and companies to jump off of it.
>>110125057ah yes, spend untold amounts of energy to reinvent the bicycle every time. Recreate new bugs, new security holes and so on just to spend even more energy to fix everything. Truly a good business model
>>110125083>ah yes, spend untold amounts of energyah yes, assuming that GPUs will never advance and LLMs won't be cheaper to run. Always a safe betAin't no way people of the future would be spending untold amounts of energy just to keep the lights on in their house at night, said some guy 120 years ago probably
>>110122425>Who peopleThe literal whoist of people
>>110125143i'm sure improvements in price and efficiency will forever be exponential!
>>110125169Based bollywood
>>110122605I've been a software developer for 20 years and I'm a tech lead. I think I know something about it. Your leaders might be saying that but they're going to realize very soon that your code will be unmaintainable and none of you understands what if does or the business.
>>110124774Not for long
>>110118696I don't even know how to code anymore
that's not good, but if it's working for you who am i to stop you? i at least hope you've got it to tell you what it spits out :P
>>110125391The fun part is when managers push to go all-in on AI and then they start getting emails and specs full of claude slop all the time. Enjoy it fuckers
>>110118696I don't review all the code anymore but I'm not completely hands off either.I ask it for a plan (written to a markdown file, not plan mode which I don't use anymore). I check the plan and ask for some changes. Then I clear context and ask for a review of the plan, with issues/suggestions presented as interactive Q&As in the terminal, and have it revise the plan with my answers.I then get it to break the plan down into phases, ending at functional and testable points. After each phase I test and I usually do look at at least the shape of the diff, plus some critical things like SQL migrations. If no issues then I commit and ask for the nest phase. I keep a copy of the plan file around for future reference.I tell myself that I do this because if I don't then the rest of the industry will and it'll be impossible to keep up, so I have no choice etc. There's some truth to that but if I'm honest, I do it because it works well enough and it's a lot easier.
>>110127490>it'll be impossible to keep upI'll be honest, I don't see how it can be that hard to prompt an AI. I slow rolled it while everyone was telling me I was being left behind 2 years ago, and I don't feel like I got left behind. You just tell it what you want and shit. The best way to understand how models and context actually work is to [spoiler]roleplay with them[/spoiler] but desu the world is not ready for that conversation.If anything, It seems riskier to avoid writing code since presumably that experience is the only reason you're better at prompting than a manager. There's no way it doesn't decay if you never write anything yourself.
