人工智能只适合编写样板文件吗？
I’m doing pretty basic stuff, and the moment Claude has to work across 3+ files, it progressively gets worse and becomes basically useless.
我正在做非常基本的事情，而当 Claude 必须处理 3 个以上的文件时，情况会逐渐变得更糟，并且基本上变得毫无用处。
Do you have a personal experience where you wrote something with AI and it almost oneshotted it?
你是否有过这样的经历：你用人工智能写了一些东西，但它几乎一炮而红？
>>109586382
Niggers on Twitter are oneshotting games with questionable gameplay and pretty crazy graphics for being just basic geometric shapes strapped together rather then vertices.
Twitter 上的《黑鬼》是单机游戏，游戏玩法值得怀疑，图形也相当疯狂，因为它们只是基本的几何形状而不是顶点。
>>109586467
Yeah but what's you personal experience?
是的，但是你的个人经历是什么？
>>109586382
>the moment Claude has to work across 3+ files, it progressively gets worse
>当 Claude 必须处理 3 个以上文件时，情况会逐渐变得更糟
Skill issue
技术问题（菜得抠脚）
You have to give Claude Code/Cursor/Codex more guidance than "fix my code"
你必须给 Claude Code/Cursor/Codex 更多指导，而不是“修复我的代码”
>>109586382
ai is very good at using command line tools and tireless trial and error loops
ai 非常擅长使用命令行工具和不知疲倦的试错循环
give it a task with well defined inputs and outputs and it will read manuals, tutorials, man pages, error outputs, log messages until it gets something that works
给它一个具有明确定义的输入和输出的任务，它会阅读手册、教程、手册页、错误输出、日志消息，直到它得到有效的东西
don't expect it to give you code that is safe, performant, easy to understand and maintain, flexible enough to handle edge cases or modify in the future, though. unless you have the ability to measure any of that and insert it in the loop
不过，不要指望它能为您提供安全、高性能、易于理解和维护、足够灵活以处理边缘情况或将来进行修改的代码。除非您有能力测量其中任何一个并将其插入循环中
>>109586528
It’s becoming very tiresome hearing people describe talking to Claude as a skill
听到人们将与克劳德交谈描述为一种技能，这变得非常令人厌烦
If this is all your job is you should be permanently unemployed
如果这就是你的全部工作，你应该永久失业
>>109586382
you can also feed it pseudocode, but whats the point of it then?
你也可以给它提供伪代码，但这有什么意义呢？
i use mine to ping-pong ideas against, making it compare my solutions to industry standards
我用我的想法来进行乒乓球对抗，使其将我的解决方案与行业标准进行比较
and as second pass to check for issues it can find
作为第二遍检查它可以找到的问题
i dont rely on it 100% but it is helpful.
我不是 100% 依赖它，但它很有帮助。
this can easily be done with the free tier of chat gpt once they see you need the nicer models (they have multiple models and settings and they try to converge on the cheapest for them that still satisfies your use and count on blueballing you to make you pay the 20usd subscription.)
一旦他们发现您需要更好的型号，这可以通过聊天 gpt 的免费层轻松完成（他们有多种型号和设置，他们会尝试为他们提供最便宜的产品，但仍能满足您的使用，并指望给您蓝球让您支付 20 美元的订阅费。）
for easier stuff like quick summaries or general fuzzy search i use perplexity. their model is source-oriented, its basically a summary machine with fuzzy search capabilities, but despite that its pretty good at "reasoning" for what it is. probably bc it builds said "reasoning" on the fly, based on the sources it found
对于更简单的东西，例如快速摘要或一般模糊搜索，我使用困惑度。他们的模型是面向源代码的，它基本上是一个具有模糊搜索功能的摘要机器，但尽管如此，它非常擅长“推理”它是什么。可能是因为它根据找到的来源即时构建了所说的“推理”
>>109586567
You sound like you are unemployed
听起来你失业了
>>109586382
Yes, context is a bitch. You're using skills, aren't you?
是的，上下文是个混蛋。你正在使用技能，不是吗？
>>109586382
Give this is a try. Make a CLI tool for uploading files on Exhentai. Choose the title, category and point to the directory with files.
尝试一下。制作一个用于在 Exhentai 上上传文件的 CLI 工具。选择标题、类别并指向包含文件的目录。
>>109586590
Other people not indulging your delusion is a touchy point for you, got it
其他人不沉迷于你的妄想对你来说是一个敏感点，明白了
>>109586588
>i use mine to ping-pong ideas against, making it compare my solutions to industry standards
>我用我的想法来对抗，使其将我的解决方案与行业标准进行比较
kek. lobotomizing yourself one session at the time
凯克。一次对自己进行脑白质切除术
>>109586382
in my experience, it's best suited to a scenario where:
根据我的经验，它最适合以下场景：
1. you initially set up a codebase with some sort of framework, consistent code style, good practices, etc.
1. 您最初建立了一个具有某种框架、一致的代码风格、良好实践等的代码库。
2. you create stubs to be implemented
2. 创建要实现的存根
3. you have claude implement one stub
3. 你让 claude 实现了一个存根
4. you fix up the implementation to ensure it meets your standards
4. 修复实施以确保其符合您的标准
5. repeat
5.重复
6. have it introspect and generate rules, skills, etc. based on what it learns about how you want code to look and function
6.让它根据了解的关于您希望代码的外观和功能的信息进行内省并生成规则、技能等
7. only at this point should you start to attempt tasking it with implementing entire features on its own
7.只有在这一点上你才应该开始尝试分配它自己实现整个功能的任务
8. review the code together with it
8. 一起回顾一下代码
9. update the code to meet your standards
9.更新代码以满足您的标准
10. repeat, and so on
10.重复，依此类推
>>109586612
No, Im not using skills. I maintain a log for the general structure of the code - what goes where step by step etc. it's a relatively simple code - dfs, fuzzy matching, scoring/thresholding, tree walks...
不，我没有使用技能。我维护代码的一般结构的日志 - 一步步走到哪里等。这是一个相对简单的代码 - dfs、模糊匹配、评分/阈值、树遍历......
Opus 5 absolutely lost the plot here.
Opus 5 在这里完全失去了情节。
The whole thing is a scam to get you to burn tokens.
整个事情都是一个骗局，目的是让你烧掉代币。
If it was actually smart it could quote you a price and deliver on that price or your money back if it doesn't work.
如果它真的很聪明，它可以向您报出价格并按照该价格交付，或者如果它不起作用，您可以退款。
>>109586692
Which come to think of it is using skills technically...
想到它是在技术上使用技能......
>>109586643
>delusion
The reality of what software engineering is in 2026?
2026 年软件工程的现实是什么？
You will never be a software engineer, you will always be a delusional NEET
你永远成不了软件工程师，你永远是一个妄想的NEET
>>109586660
That's how I started initially but then I handed him some tedious and boring things like regexes and scoring and now I find bugs everywhere
我一开始就是这样开始的，但后来我给了他一些乏味无聊的东西，比如正则表达式和评分，现在我发现到处都是错误
>>109586764
they're overeager interns (i.e., willing and able to do bitch work, but you need to either hold their hand, or ensure the tasks are extremely linear and structured)
他们是过于热心的实习生（也就是说，愿意并且能够做一些卑鄙的工作，但你需要要么握住他们的手，要么确保任务是非常线性和结构化的）
>>109586382
Not only that but there is a lot of 'dark algos/CS/implementations' that is not in the literature so no AI would be able to produce satisfactory results, let alone a state-of-the-art design.
不仅如此，还有很多文献中没有的“黑暗算法/CS/实现”，因此没有人工智能能够产生令人满意的结果，更不用说最先进的设计了。
TLDR People who use AI work in low skill software domains
TLDR 使用人工智能的人在低技能软件领域工作
>>109586692
Skills help you keep context at bay since only the description is ever kept inside the context. Only when they are invoked they get fully loaded.
技能可以帮助您控制上下文，因为上下文中只有描述。只有当它们被调用时，它们才会被完全加载。
You won't be able to complete even mid scale projects by just raw dogging it.
即使是中等规模的项目，仅靠原始的跟踪也无法完成。
>>109586382
AI can do it, the real question that niggercattle like >>109586528 don't think to ask is whether all the work and token spending that goes into the result is actually any faster or more effective than traditional programming.
人工智能可以做到这一点，像>>109586528这样的黑牛不会想问的真正问题是，结果中的所有工作和代币支出实际上是否比传统编程更快或更有效。
At what point are you spending more time fiddling with agent loops and markdown files than just modifying the damn code yourself? AI can be very useful, but AI psychosis is real.
在什么时候你会花更多的时间摆弄代理循环和 Markdown 文件而不是自己修改该死的代码？ 人工智能可能非常有用，但人工智能精神病是真实存在的。
>>109586382
yeah, it roughly one-shotted a small feature for me over many more files.
是的，它大致为我在更多文件中一次性实现了一个小功能。
I changed some details but claude's attempt wasn't bad.
我改变了一些细节，但克劳德的尝试还不错。
did you use plan mode with opus/opusplan?
您是否在 opus/opusplan 中使用了计划模式？
>>109586567
seems to be a skill if so many retards can't do it properly.
如果这么多弱智都做不好的话，这似乎是一项技能。
>>109587123
>how the pros do it properly
>专业人士如何正确地做到这一点
mayebe its just a bit shit, actually
也许实际上只是有点狗屎
>>109587126
werks on my machine
在我机器上能跑
>>109586730
The delusion that this is a “skill.”
错觉这是一种“技能”。
If you have to intentionally deflect twice in a row it’s clear it’s a sore spot for you.
如果你必须连续两次故意偏转，那么显然这对你来说是一个痛处。
>>109587139
yeah, but how, and at what price?
是的，但是如何以及以什么价格？
bc not well enough for anthropic, or not cheaply enough for msft
bc 对于人类来说不够好，或者对于微软来说不够便宜
(plus the absolute state of wintra$h, but i digress...)
plus plus this can be actually an euphemism for having to pay manhours to clean the mess on top of burning cash on tokens, bc msft also has ai products and wouldnt want to do anti advertisement for themseleves by saying
另外，这实际上可能是一种委婉说法，表示除了在代币上烧钱之外还必须支付工时来清理混乱，BC MSFT 也有人工智能产品，并且不想为自己做反广告，说
>our ai is shit so we tried antropic's
>我们的人工智能很糟糕，所以我们尝试了 antropic
>turns out theirs is also shit, so were back to doing things the old way
>结果他们的也很糟糕，所以我们又回到了旧的方式做事
imagine the absolute fucking shitstorm that would cause among the investors
想象一下这会在投资者中引起一场绝对他妈的狗屎风暴
fucking fecal fragmentation cluster ammunition grade lamao
他妈的粪便碎片集束弹药级lamao
>>109587166
I’m not wasting my time arguing with a NEET
我不会浪费时间和NEET争论
>>109587167 cont
>>109587167续
im still laughing from the mental image
我仍然在从脑海中的形象中笑出来
sata nutella capretbombing everyone because everyone has fingers in eachothers pockets
sata nutella capret轰炸每个人，因为每个人都把手指插在彼此的口袋里
the whole thing hinges on technology that doesnt even exist yet
整个事情取决于尚不存在的技术
and datacenters that havent yet been built
以及尚未建成的数据中心
but most of all
但最重要的是
the boomer, who like a thrird world kid picks up one of the misfires and it explodes covering him in shit because his 401k plummets into the shitter
婴儿潮一代，就像一个第三世界的孩子一样，捡起了其中一个失火，它爆炸了，他浑身都是屎，因为他的 401k 掉进了屎里。
fragments of shit flying everywhere
粪便碎片飞得到处都是
vaporized shit sprayed onto unsuspecting bystanders who just hodl the sp500
蒸发的粪便喷到毫无戒心的旁观者身上，他们只是持有 sp500
long term shit residue that causes birth defects and infertility through long term catastrophic erectile dysfunction from the trauma of just how hard the market got btfod
长期的粪便残留物，通过长期灾难性的勃起功能障碍，导致先天缺陷和不孕不育，而这些创伤是由于市场对 btfod 的猛烈攻击而造成的。
us had its dust bowl
我们有它的尘碗
but the indian festival thats coming is gonna eclipse whatever that was lamao
但即将到来的印度节日将会使拉茂的一切黯然失色
>>109586528
>Skill issue
>技术问题。
Nigger what? Anons on twitter are getting ChatGPT to do original research by giving it one sentence commands lol, get this pretentious shit out of here, you prompt, pray and let Claude rip like everyone else
黑鬼什么？ Twitter 上的匿名者正在让 ChatGPT 进行原创研究，方法是给它一句话命令，哈哈，把这个自命不凡的狗屎滚出去，你提示，祈祷，让克劳德像其他人一样撕裂
>the moment Claude has to work across 3+ files, it progressively gets worse and becomes basically useless
>当 Claude 必须处理 3 个以上的文件时，情况会逐渐变得更糟，并且基本上变得毫无用处
tell us more about what sort of problems you're running into so we can help you out anon
告诉我们更多关于您遇到的问题，以便我们可以立即帮助您解决
>>109587241
*would be coming
*将会来
because if the whole thing unravels slowly or the techbros actually deliver, the market will be just fine
因为如果整个事情慢慢解开或者技术兄弟真正交付，市场就会很好
it can absorb losing 7% of the s&p 500, given its distributed in time
考虑到其分布时间，它可以吸收标准普尔 500 指数 7% 的损失
>>109586710
I don't even understand how people burn tokens. My "record" is a bit over 20 % of daily limit and even that felt like cranking it on coke brains off.
我什至不明白人们如何燃烧代币。我的“记录”比每日限额的 20% 多一点，即使这样也感觉像是在喝可乐大脑。
The absolute slop did win us some hackathon money so there's that. Execs only care about the looks and the pitch.
绝对的失败确实为我们赢得了一些黑客马拉松的钱，所以就是这样。高管只关心外观和宣传。
>>109586807
luddite cope
卢德分子式自欺欺人
>>109586510
When I tried to make it create a game for me that wasn’t racing game or broken barely functioning fps it failed. It’s amazing in UI design and kind of crazy when it comes to 3D modelling things in three.js, at least for LLMs, but other then that it’s kind of shit. It can do 2 days worth of making functional game settings in a single hour, but the gameplay will be horrendous and polishing it yourself will be next to impossible because you don’t understand the code and it will waste millions of tokens misunderstanding your desires. It’s basically the same as diffusion images we’re back in 2022-2025 where it was giving very impressive outputs but you had very little control, so the main people who saw significant liking in it were uncreative degenerates with no vision who just had very vague ideas or no ideas, or few autists who tinkered with it to such extent only to get slight improvement over the degens.
当我试图让它为我创建一款不是赛车游戏或几乎无法运行的 fps 游戏时，它失败了。它在 UI 设计方面令人惊叹，而在 Three.js 中进行 3D 建模时则有点疯狂，至少对于法学硕士来说是这样，但除此之外，它就有点糟糕了。它可以在一个小时内完成 2 天的功能性游戏设置，但游戏玩法会很糟糕，你自己打磨它几乎是不可能的，因为你不理解代码，而且会因为误解你的愿望而浪费数百万代币。它与我们在 2022-2025 年的扩散图像基本相同，当时它提供了非常令人印象深刻的输出，但你几乎无法控制，所以对它有很大兴趣的主要人是没有创造力、没有视力、只有非常模糊的想法或没有想法的退化者，或者很少有自闭症患者，他们对它进行了如此程度的修改，只是为了比退化者稍微有所改善。
>>109587494
Then game development will be completely solved by 2030 by your pace.
那么游戏开发将在2030年按照你的节奏彻底解决。
>>109586382
>Do you have a personal experience where you wrote something with AI and it almost oneshotted it?
>您是否有过这样的个人经历：您用 AI 写了一些东西，并且几乎一次性完成了它？
yes, when a modern greenfield project is structured well and gets larger, it can oneshot stuff without much handholding. when the project is small the LLM has no reference and just makes bad assumptions and you need to be very precise and hand-holdy.
是的，当一个现代的绿地项目结构良好并且变得更大时，它可以在没有太多帮助的情况下一次性完成任务。当项目很小时，法学硕士没有参考资料，只会做出错误的假设，你需要非常精确和掌握。
using claude sonnet btw, not even the frontier stuff
顺便说一句，使用克劳德十四行诗，甚至不是前沿的东西
>>109586467
>questionable gameplay and pretty crazy graphics
>有问题的游戏玩法和相当疯狂的图形
graphics are pretty easy to achieve if you got the asset rendering performance is the question. Also if gameplay is crap then the game is crap. "Oneshotting" a game like pokemon and never testing it beyond 1min demo just shows its garbage.
如果你有资产渲染性能是问题的话，图形是很容易实现的。另外，如果游戏玩法很糟糕，那么游戏也很糟糕。 “Oneshotting”像《口袋妖怪》这样的游戏，却从不对其进行超过 1 分钟的演示，这只会显示出它的垃圾。
>>109586807
you can give it a research paper and it will implement it I did it for some rf simulation stuff it works however you cannot verify the results without more expansive software so you have a script that kinda works but is hard to extend thats produces results that kinda match your expectation.
你可以给它一篇研究论文，它会实现它，我为一些射频模拟的东西做了它，它可以工作，但是如果没有更广泛的软件，你就无法验证结果，所以你有一个可以工作但很难扩展的脚本，它产生的结果有点符合你的期望。
It's good at automating things and teaching you new ways to autoomate things. Boilerplate is a non-issue, it can blow right through that. So now that boilerplate is able to be written quickly, you don't need things like third party libraries as much anymore.
它擅长自动化事物并教您自动化事物的新方法。样板文件不是问题，它可以直接解决这个问题。因此，既然可以快速编写样板文件，您就不再需要第三方库之类的东西了。
The real hard truth: In corporate programming, in many enterprises, delusional people set lines of code as a metric to be strived for. This leads to things being made for the purpose of being made, work being made for the purpose of work. "Makework", i.e, work that serves no purpose. This is why software has become so bloated needlessly. Because of the gamification of the lines of code metric. That is the real absolute reason and the absolute truth that nobody wants to tell you. Thousands of hours of programming work has been done purely in the vain of satisfying the lines of code metric. You need less lines of code, not more.
真正残酷的事实是：在公司编程中，在许多企业中，妄想者将代码行数设置为要争取的指标。这导致事物被制造是为了被制造的目的，工作是为了工作的目的而被制造。 “Makework”，即没有目的的工作。这就是为什么软件变得如此不必要地臃肿。因为代码行度量的游戏化。这才是真正的绝对道理，也是没有人愿意告诉你的绝对真理。数千小时的编程工作纯粹是为了满足代码行数指标而白费。您需要更少的代码行，而不是更多。
>>109586382
Massive skill issue I've worked on personal projects with 40k lines super easy even with the dumbest LLM
巨大的技能问题我曾在个人项目上工作过 40k 行，即使是最愚蠢的法学硕士也非常容易
You need to have good documentation in optimized LLM format so they can read it quickly
您需要拥有优化的 LLM 格式的良好文档，以便他们可以快速阅读
>>109587734
This. You need to setup the appropriate CLAUDE.md, introduce a ROADMAP file, etc. I know managing state in flat files sounds like a total backwards move but trust me, in AI this is the pattern that works.
这。您需要设置适当的 CLAUDE.md，引入 ROADMAP 文件等。我知道在平面文件中管理状态听起来完全是倒退，但相信我，在 AI 中这是有效的模式。
Key skills to unlock your success:
开启您成功之路的关键技能：
>MCP fundamentals
>MCP基础知识
>Agents
>Guardrails
>Skills (skills.md format)
>技能（skills.md格式）
>RAG
>CoT Prompting
>CoT 提示
And yes, you need git.
是的，你需要 git。
>>109587139
Your job isn't to ship code, it's to advertise on 4chan. You haven't done shit.
您的工作不是发送代码，而是在 4chan 上做广告。你又没干过坏事
>>109586382
Lol I made that image
哈哈，那张图片是我做的
>>109587511
>will
>shall
>going to
> 去
>about to
>关于
>set to
>设置为
>due to
>由于
>bound to
>绑定到
>destined to
>注定要
>intend
>plan
>aim
>propose
>expect
>project
>prepare
>forecast
>predict
>anticipate
>target
>schedule
>>109587989
based listanon
基于列表
>>109586382
>Do you have a personal experience where you wrote something with AI and it almost oneshotted it?
>您是否有过这样的个人经历：您用 AI 写了一些东西，并且几乎一次性完成了它？
>>109586467
>>109586555
>>109586588
>>109586623
>Niggers on Twitter are oneshotting games
>推特上的黑鬼都是一击必杀的游戏
>ai is very good at using command line tools and tireless trial and error loops
>ai 非常擅长使用命令行工具和不知疲倦的试错循环
>you can also feed it pseudocode
>你也可以给它提供伪代码
>Make a CLI tool for uploading files on Exhentai
>制作一个CLI工具，用于在Exhentai上上传文件
Way to not answer OP's question at all.
根本不回答OP的问题。
>>109586623
There isn't any model out there that can do this.
没有任何模型可以做到这一点。
>>109588046
^expectations
<reality
>jeet from bangalore > openais frontier models
>来自班加罗尔的jeet > openais 前沿模型
¯\_(ツ)_/¯
>>109588046
>>109588089
what i mean by that is that one-shotting anything beyond 500 locs starts to get wonky
我的意思是，一次性射击任何超过 500 个位置的东西都会开始变得不稳定
which is a good benchmark for the evolution of llms proper, wo taking into account the support of harnessing
这是 LLMS 本身发展的一个很好的基准，考虑到利用的支持
but also revelatory of the underlying scaling/diminishing returns problems
但也揭示了潜在的规模/收益递减问题
>>109587719
>the lines of code metric
>代码行数度量
True, but should be generalized to any arbitrary metric of "work done" that is unrelated to any measurable value to the customer or even bottom line for the company
确实如此，但应该推广到任何“完成的工作”的任意指标，这些指标与客户的任何可衡量价值甚至公司的底线无关
(Jira tickets closed, changes shipped, PRs merged, reports generated, etc)
（Jira 票证关闭、更改已发送、PR 合并、生成报告等）
>>109586567
>If this is all your job is you should be permanently unemployed
>如果这就是你的全部工作，你就应该永久失业
The job is to be backup of claude when the budget for tokens runs out
工作是当代币预算用完时充当克劳德的后援
>>109587719
>The real hard truth: In corporate programming, in many enterprises, delusional people set lines of code as a metric to be strived for. This leads to things being made for the purpose of being made, work being made for the purpose of work. "Makework", i.e, work that serves no purpose. This is why software has become so bloated needlessly. Because of the gamification of the lines of code metric. That is the real absolute reason and the absolute truth that nobody wants to tell you. Thousands of hours of programming work has been done purely in the vain of satisfying the lines of code metric. You need less lines of code, not more.
>真正残酷的事实：在公司编程中，在许多企业中，妄想者将代码行数设置为要争取的指标。这导致事物被制造是为了被制造的目的，工作是为了工作的目的而被制造。 “Makework”，即没有目的的工作。这就是为什么软件变得如此不必要地臃肿。因为代码行度量的游戏化。这才是真正的绝对道理，也是没有人愿意告诉你的绝对真理。数千小时的编程工作纯粹是为了满足代码行数指标而白费。您需要更少的代码行，而不是更多。
Is this still a thing? Even Tokenmaxxing died down for the most part
这还是一件事吗？就连 Tokenmaxxing 也大部分消亡了
>>109588761
That type of corporate dysfunction predates LLMs by decades.
这种类型的企业功能障碍早于法学硕士出现数十年。
>>109586567
Wait until you get your first job and discover the managerial class.
等到你找到第一份工作并发现管理阶层。
>>109588949
I feel like one of the nice things about LLMs is that it will eventually make corporations drop lines of code as a metric because it can be gamed by pushing overengineered verbose slop now and setting it as a metric will actively destroy any codebase
我觉得法学硕士的好处之一是，它最终会让公司放弃将代码行数作为衡量标准，因为现在可以通过推动过度设计的冗长废话来进行游戏，并将其设置为衡量标准将主动破坏任何代码库
>>109589343
I am old enough to remember when programmers also agreed their job was not a skill, now there is an influx of morons who think supervising Claude is in any way the same as what the job used to be
我已经足够大了，还记得程序员们也承认他们的工作不是一项技能，现在有一群白痴涌入，他们认为监督克劳德在任何方面都与过去的工作相同
>>109586382
It fucks up even for boilerplate. I have to constantly fix shit up in Java even though I provided templates and strict but simple instructions. It's like twice as fast as me but since I have no idea what it wrote I have to thoroughly check basic stuff. The people saying it's a personal pajeet codemonkey are right
即使是样板文件也很糟糕。即使我提供了模板和严格但简单的说明，我也必须不断修复 Java 中的问题。它的速度是我的两倍，但由于我不知道它写了什么，所以我必须彻底检查基本的东西。人们说这是个人 pajeet codemonkey 是对的
>>109586382
I have to be honest I was pretty spooked by all the coders saying Claude can do their job
我必须说实话，我对所有说克劳德可以完成他们的工作的程序员感到非常震惊
I hooked it up to one of my problems and let it think for a while and it came back with stuff that sounded good and worked to my surprise
我将它与我的一个问题联系起来并让它思考了一段时间，它返回的内容听起来不错，并且令我惊讶。
But then I looked inside and saw it was full of small mistakes in design decisions made over and over again and the logic was just surface level, if I let it continue it was certainly going to absolutely destroy the codebase and make it an unmaintainable disaster of shitty hacks to hold the bad decisions together
但后来我往里面看，发现它在设计决策中充满了一遍又一遍的小错误，而且逻辑只是表面层面的，如果我让它继续下去，它肯定会彻底摧毁代码库，并使其成为一场无法维护的灾难，由糟糕的黑客将错误的决定结合在一起。
Idk what fields are getting fucked over by this, are these all webdev niggers?
我不知道哪些领域被这个搞砸了，这些都是网络开发黑鬼吗？
>>109590146
the only people who're getting replaced are people with horrible attitude, aka the tech "priests" who think being able to leetcode is what gives them an identity and a pass to act like insufferable cunts
唯一被取代的人是那些态度恶劣的人，又名科技“牧师”，他们认为能够学习 Leetcode 就可以给他们一种身份，让他们可以像令人难以忍受的混蛋一样行事。
replacement is ultimately dependent on whether management thinks they are willing to let someone who has a better attitude use AI to do the work versus having to listen to those insane luddites trying to gatekeep their profession
替换最终取决于管理层是否认为他们愿意让态度更好的人使用人工智能来完成工作，而不是不得不听那些试图把关他们的职业的疯狂勒德分子的声音
managers know there's all kinds of shit like this going on
经理们知道有各种各样这样的事情发生
truth be told, a lot of engineers i came across are just absolutely terrible human beings. they equate their skillfulness with their "understanding" of the world are some of the most insecure and immature people ive ever worked with
说实话，我遇到的很多工程师都是非常糟糕的人。他们将自己的技能等同于对世界的“理解”，是我共事过的最没有安全感和最不成熟的人
>>109590247
Oh yeah, don't get me started on those "agile practitioners"
哦，是的，别让我开始谈论那些“敏捷实践者”
>>109590253
if she's middle age, and she tries to peddle your team agile, your company already lost
如果她人到中年，却试图兜售你的团队敏捷，那么你的公司就已经输了
>>109586382
>Do you have a personal experience where you wrote something with AI and it almost oneshotted it?
>您是否有过这样的个人经历：您用 AI 写了一些东西，并且几乎一次性完成了它？
Yes. For about 99%, actually. But it requires that you aren't a complete retard and actually use planning mode tools to refine beforehand, and that you force the model into a higher reasoning mode than 'basic dumb blonde bitch' aka 'standard'-level. Treat it like what it is: 20 WIS; 0 INT.
是的。事实上，大约99%。但它要求你不是一个完全的弱智，并且实际上使用计划模式工具来预先改进，并且你迫使模型进入比“基本愚蠢的金发碧眼的婊子”又名“标准”级别更高的推理模式。对待它本来的样子：20 WIS； 0 INT。
>>109586467
Nigga, I don't think you're aware how easy it is to get lighting and basic bitch le ebin realistic shsders set up in UE5. It was made to look flashy out of the box.
老兄，我想你不知道在 UE5 中设置光照和基本的 le ebin 现实 shsders 有多么容易。它开箱即用，看起来很华丽。
>>109587511
Yeah, probably. But I also expect AGI by that time, or at least something that can cause enough problems to the world as AGI, forcing governments to do UBI or do some weird workaround regulations or model restriction access because LLMs become capable of making actual bioweapons with cheap toolkits and unskilled human assistance.
是的，可能是。但我也预计到那时 AGI 会出现，或者至少会像 AGI 一样给世界带来足够多的问题，迫使政府实施 UBI 或制定一些奇怪的变通法规或模型限制访问，因为法学硕士有能力利用廉价的工具包和不熟练的人类援助来制造实际的生物武器。
>>109591363
I am not talking about lightning but about AI making vector graphics that look decent.
我不是在谈论闪电，而是在谈论人工智能制作看起来不错的矢量图形。
>>109587139
>still twerking
>还在扭臀舞
get out
滚出去。
>>109587375
libtard arrest
自由逮捕
>>109586382
Nigga, i got literal 8 different 50k C++ files done all good in literally sillytavern because i only use claude opus api for rp, rarely for work. I never bothered with claude code
黑鬼，我在sillytavern中得到了8个不同的50k C++文件，一切都很好，因为我只使用claude opus api进行rp，很少用于工作。我从不关心克劳德代码
Sounds like skill issue to me
对我来说听起来像是技能问题
>>109587719
define 'boilerplate'
定义“样板”
I don't think I'd consider any code I've ever written to be boilerplate since whenever I find something too tedious or repetitive to type out, I generally find another design that avoids having to write out that code at all.
我不认为我写过的任何代码都是样板文件，因为每当我发现一些东西太乏味或重复而无法输入时，我通常会找到另一种设计来完全避免编写该代码。
>>109586382
skill issue/are you using sonnet or something?
技能问题/你使用十四行诗还是其他什么？
This was a big issue like a year or two ago but just have fable orchestrate everything and have well organized code and it shouldn't be an issue. You DO need to know how to code so you can actually organize it and set it up though, you DO know how to code, right?
这是一两年前的一个大问题，但只要有寓言编排一切并拥有组织良好的代码，它就不应该成为问题。您确实需要知道如何编码，以便您可以实际组织和设置它，但您确实知道如何编码，对吗？
>>109592507
So boilerplate is stuff like <!DOCTYPE html> or generally anything that you write by convention tediously over and over again but teh standard says you should have it. So like the web standards say that you need to have exactly <!DOCTYPE html> but in reality if you use <!doctype html> or similar you'd be breaking standard but this doesn't really hurt your program or your webpage or whatever in any way.
因此，样板文件就像 <!DOCTYPE html> 之类的东西，或者通常是您按照惯例一遍又一遍地乏味地编写的任何东西，但标准说您应该拥有它。因此，就像网络标准所说，您需要精确地 <!DOCTYPE html> 但实际上，如果您使用 <!doctype html> 或类似的内容，您就会违反标准，但这并不会真正损害您的程序或网页或任何方式。
>>109591483
>thinks he’s getting UBI when he has no skills to leverage
>当他没有技能可以利用时，他认为自己正在获得全民基本收入
We let homeless people starve right now kek, it’s not happening
我们现在让无家可归的人挨饿，这没有发生
>>109593297
>we let homeless people starve right now kek
>我们现在让无家可归的人挨饿kek
Who is "we"? Just wanna be clear on this one. Because that isn't a flex.
“我们”是谁？只是想弄清楚这一点。因为那不是弹性的。
>>109593175
I don't write html tho?
我不会写html吗？
in PHP, an example of boilerplate would be <?php however, in PHP there might be consequences for breaking the standard, such as your code not executing.
在 PHP 中，样板文件的示例是 <?php 但是，在 PHP 中，违反标准可能会产生后果，例如您的代码无法执行。
>>109593362
I also don't write php what do you only do websites?
我也不写php，你只做网站做什么？
besides idk writing five letters at the beginning and end of your script is kinda acceptable to me it's like a shebang
除了 idk 在脚本的开头和结尾写五个字母对我来说有点可以接受之外，它就像一个 shebang
and I don't need AI for that it takes like, maybe two seconds.
我不需要人工智能，因为这可能需要两秒钟。
>>109593426
okay, so here are some more examples of boilerplate, that also have consequences of things breaking and not working, but that are required for valid reasons
好的，这里还有一些样板文件的示例，这些示例也会造成损坏和无法工作的后果，但这是有正当理由的
In python, you'd do things like import libraries. This is shown at the top of most python programs and is also a trait shared with Java. In enterprise Java, it gets almost absurd. Thankfully, we have tools like IntelliSense that make that easier to deal with.
在 python 中，您可以执行诸如导入库之类的操作。这显示在大多数 Python 程序的顶部，也是与 Java 共享的特征。在企业级 Java 中，这几乎变得荒谬。值得庆幸的是，我们有像 IntelliSense 这样的工具，可以让处理变得更容易。
In C++, C, pretty much any programming language that has libraries, you will see this pattern. It is boilerplate, and AI is really good at doing it, and it's also really good at making fully functional programs and people have made a lot of cool things with it.
在 C++、C 以及几乎所有具有库的编程语言中，您都会看到这种模式。它是样板文件，人工智能非常擅长这样做，而且它也非常擅长制作功能齐全的程序，人们用它做了很多很酷的东西。
So, no. I don't just do websites. How about you tell me what you need?
所以，不。我不只是做网站。你告诉我你需要什么怎么样？
>>109586382
Using an AI for just boiler plate is like using a car for just air conditioning. It only makes sense if you're european.
仅将人工智能用于样板就像仅将汽车用于空调一样。只有当你是欧洲人时这才有意义。
>>109593461
That is a terrifying metric. It makes me wonder if the US is underreporting or if there are major differences between how the EU and the US classify cause of death. I know in the US at least it's probably highly fragmented state by state so getting reporting data in a consistent format is probably hard.
这是一个可怕的指标。这让我想知道美国是否漏报了，或者欧盟和美国对死因的分类是否存在重大差异。我知道至少在美国，各州可能高度分散，因此以一致的格式获取报告数据可能很困难。
>>109593446
idk my language server and auto completion in vim just shows me a popup of the namespaces and objects and when I select it it automatically adds the import statement. I could also write it by hand desu I don't really depend on much stuff maybe one or two libs a project.
idk 我的语言服务器和 vim 中的自动完成功能只是向我显示名称空间和对象的弹出窗口，当我选择它时，它会自动添加导入语句。我也可以手工编写它 desu 我并不真正依赖太多东西，也许一个项目只有一两个库。
I just don't need an AI for this and it takes me maybe thirty minutes combined across a project so I'm just trying to understand what you mean honestly?
我只是不需要人工智能来做这件事，而且整个项目可能需要我三十分钟的时间，所以我只是想诚实地理解你的意思？
>>109593538
I'm asking you, what is an issue that you face that you need help with?
我问你，你遇到什么问题需要帮助？
>>109593542
idk I need to do my dishes cause I skipped it last evening but programming wise I'm pretty gucci desu
我不知道我需要洗碗，因为我昨晚跳过了它，但就编程而言，我很漂亮 gucci desu
it's all fun so I don't want an ai to do it cause that takes the fun out of it
这很有趣，所以我不想让人工智能来做这件事，因为这会失去乐趣
>>109586382
>oneshotted
>gaymen lingo
>同性恋行话
negmi
>>109586567
Wait until you find out what management does
等到你知道管理层是做什么的
>>109586382
AI is not intelligent !!!
AI不智能！！！
Intelligence in name is just marketing slogan!
名义上的智能只是营销口号！
AI - stand for
AI——代表
A - algorithm
A-算法
I - interface
I-接口
!!! NOT !!!
Artificial Intelligence
人工智能
It is basically advanced autocomplete
它基本上是高级自动完成
It can work only with templates which was trained on!
它只能与经过训练的模板一起使用！
It can NOT work with code which was not exposed to!
它不能与未接触过的代码一起工作！
It definitely can not create completely new algorithm or unique new code!
它绝对不能创建全新的算法或独特的新代码！
You have to stop believe that it is Intelligent because it is NOT!!!
你必须停止相信它是智能的，因为它不是！
>>109595039
It being intelligent or not has nothing to do with what actually matters, it getting shit done.
聪明与否与真正重要的事情无关，重要的是把事情做好。
>>109586467
>oneshotting three.js slop
Wow
Epic
Ebin
>>109595328
you are replying to a bot, anon
你正在回复机器人，匿名
>>109586382
I seriously hope you guys don't actually use these as single instance continuous tasks in a single context window. You're quite literally using it wrong.
我真心希望你们不要在单个上下文窗口中将这些实际用作单个实例连续任务。你确实用错了它。
>but if I have to delegate that means I have to know how to program and design software architecturally!
>但如果我必须委派，这意味着我必须知道如何在架构上编程和设计软件！
>>109593297
Right now we live in meritocratic or pseudo-meritocratic system where usefulness is not just encouraged but forced. Or at least the idea of usefulness. Once we get AGI meritocracy breaks down because in face of true AGI/ASI all humans are useless. So we will split into either UBI communism because there will no longer be any meritocracy-based justification to keep homeless homeless, or we will get Nick Land apocalypse where machines kill all of humanity and probably life while we let them do it because they have more merit then us.
现在，我们生活在精英或伪精英体系中，在这种体系中，有用性不仅被鼓励，而且被强迫。或者至少是有用的想法。一旦我们实现了 AGI 精英统治，就会崩溃，因为面对真正的 AGI/ASI，所有人类都是无用的。因此，我们要么会分裂成全民基本收入共产主义，因为不再有任何基于精英的理由让无家可归者无家可归，要么我们会陷入尼克·兰德末日，机器杀死全人类，甚至可能杀死生命，而我们却让他们这样做，因为他们比我们有更多的优点。
>>109596136
FALC.
>>109596136
>Once we get AGI meritocracy breaks down because in face of true AGI/ASI all humans are useless.
>一旦我们实现了 AGI 精英统治，就会崩溃，因为面对真正的 AGI/ASI，所有人类都是无用的。
These kind of claims all ignore or gloss over what the actual cost of AI-powered robots may actually be.
这些说法都忽略或掩盖了人工智能机器人的实际成本。
>>109596226
Except for few instances of LLMs costing 10x of what engineers do this summer, they have always been cheaper, and the reason why companies are wasting millions on Fable tokens is combination of tokenmaxxing and creating convoluted agent harnesses that make the AI model engage in books worth of debate with sub agents.
除了少数 LLM 的成本是今年夏天工程师工作量的 10 倍之外，它们一直都比较便宜，而公司在 Fable 代币上浪费数百万美元的原因是 tokenmaxxing 和创建复杂的代理工具的结合，这使得人工智能模型与子代理进行了值得争论的书籍。
>>109595328
It all depends what you consider
这完全取决于你的考虑
"getting shit done"
“把屎做完”
Certainly it can in accounting
会计上当然可以
There is not much different ways how to
没有太多不同的方法
count 1+1 !
数1+1！
But try ask AI come with some really obscure fast code ...
但是尝试要求人工智能提供一些非常晦涩的快速代码......
In coding AI doing ONLY boiler plates!
在人工智能编码中只做样板！
It is not able do anything new
它无法做任何新的事情
You need skilled Coder for that!
为此，您需要熟练的编码员！
It is AUTOCOMPLETE if you like it or not!
不管你喜欢与否，它都是自动完成的！
Basically it is Scam so huge People are not even able grasp scale of it!!!
基本上这是一个巨大的骗局，人们甚至无法掌握它的规模！
It basically bottomless money pit with 0 results and even it eats jobs (eat economy itself) sort of Turbo SCAM!!!
它基本上是无底洞的钱坑，结果为零，甚至它吃掉了工作（吃掉了经济本身），有点像涡轮骗局！
>>109586382
Deepseek-v4-flash and deepseek-v4-pro do that, Xiaomi mimo-v2.5 does it nicely in opencode and i even made a really fun text editor work it using qt and c++ and it works great with very large text files.
Deepseek-v4-flash 和 deepseek-v4-pro 可以做到这一点，小米 mimo-v2.5 在开放代码中做得很好，我什至使用 qt 和 c++ 制作了一个非常有趣的文本编辑器，它可以很好地处理非常大的文本文件。
I do try to avoid using western ai models as they serve only the shareholder, not the customer and certainly not public good by any stretch.
我确实尽量避免使用西方人工智能模型，因为它们只为股东服务，而不是客户，当然也不是公共利益。
>>109586382
Writing a full fledged software with proper software engineering on a Codex $20 plan is taking me months
按照 Codex 20 美元的计划编写一个具有适当软件工程的成熟软件需要花费我几个月的时间
>>109596350
>Except for few instances of LLMs costing 10x of what engineers do this summer, they have always been cheaper,
>除了少数情况下 LLM 的费用是今年夏天工程师费用的 10 倍外，它们一直都比较便宜，
No they haven't, and its a big reason why various politicians have been supporting an open invasion of browns into western countries. A robot vacuum is never going to be faster than just sweeping the floor with a simple broom.
不，他们没有，这也是众多政客支持棕色人种公开入侵西方国家的一个重要原因。 扫地机器人永远不会比用简单的扫帚扫地更快。
It's one thing for LLMs to automate white collar type work. But as explosive as the growth in Big Tech has been, you can't ask an LLM to make you a sandwich out of thin air. There's ALWAYS going to be resource scarcity to consider even if you assume the cost of inference itself goes to zero (it won't).
对于法学硕士来说，自动化白领类型的工作是一回事。 尽管大型科技公司的发展呈爆炸性增长，但你不能要求法学硕士凭空为你做三明治。 即使您假设推理成本本身为零（它不会），也始终需要考虑资源稀缺问题。
（如果你觉得这篇文章有启发，可以点击这里付费）
