Give it to me straight, /g/.
直说吧 /g/。
编程完蛋了吗？
>>109798315
The rate at which thing are progressing I doubt if there are gonna be any human coders after 3 years. If computer use also gains enough progress in h5 years its en of all white collar jobs
按现在的进展速度，我怀疑 3 年后还会不会有人类程序员。如果 computer use 在未来 5 年也取得足够进展，那就是所有白领岗位的终结
AI shits out unmaintainable slop but it works
AI 拉出无法维护的垃圾，但它能用
I guess it really doesn't matter if it's unmaintainable if humans don't need to touch it in the first place
我想如果人类本来就不用碰它，那能不能维护确实无所谓
>>109798333
cargo cultist
货物崇拜者
>>109798315
yup. but not for ai-related reasons
对。但不是因为 AI 相关的原因
unless you live in bengaluru
除非你住在班加罗尔
>>109798315
We answered this a million times, yes it's dead. It was already dead even before AI.
这问题我们答了一百万次了：是的，死了。AI 出现之前就已经死了。
Get into pluming or healthcare (if you can handle dealing with a bunch of foids and their drama). Whatever you do, just don't pick CS
去干水管工或医护（如果你受得了应付一堆女碧池和她们的狗血剧）。干什么都行，就是别选 CS
>>109798350
>I guess it really doesn't matter if it's unmaintainable if humans don't need to touch it in the first place
>我想如果人类本来就不用碰它，那能不能维护确实无所谓
A fraction of the existing manpower is enough to maintain it using AI
现有人力的一小部分加 AI 就够维护它
>>109798360
Lol even in bengaluru it's dying buddy. Have you not read news about the mass firings in India and reduced hiring there too? Though not as bad as in US.
哈哈连班加罗尔都在完蛋，兄弟。你没看印度大裁员、招聘缩减的新闻吗？虽然没美国那么惨。
Traditional coding will be dead as a career, yes, but it’ll take quite a few years while businesses integrate AI architecture to their existing systems. Once it happens, people will still do it as a hobby like woodworking or knitting. And the new tech jobs will be configuring and managing systems of agents.
传统编程作为职业会死，是的，但还得好几年，企业要把 AI 架构整合进现有系统。等那发生后，人们还会像木工或编织一样把它当爱好。新的技术岗位是配置和管理 agent 系统。
>>109798360
>yup. but not for ai-related reasons
>对。但不是因为 AI 相关的原因
>unless you live in bengaluru
>除非你住在班加罗尔
I'll be honest I'm a jeet and things aint rosy here either. Companies can make do with less number of employees each quarter and hiring has slowed down a lot
老实说我就是印度佬，这边日子也不滋润。公司每季度都能靠更少的人凑合，招聘也慢了很多
>>109798429
yeah nah
是啊才怪
"
Ongoing demand uncertainty, slower client decision-making, and increasing use of AI-driven efficiencies have all contributed to tighter headcount strategies.
持续的需求不确定性、客户决策变慢，加上越来越多使用 AI 提效，都导致了更紧缩的人员编制策略。
Scale-driven growth to a skills-first model
从规模驱动转向技能优先
>Hiring has not fallen back to FY24 levels, but the overall trend still points to limited clarity on growth. Back then, the sector had seen a sharp contraction of over 69,000 jobs.
>招聘虽没跌回 FY24 水平，但整体趋势仍显示增长不明朗。当时该行业曾锐减超过 6.9 万个岗位。
"
agi in 2mw and so on
两周内 AGI，诸如此类
>>109798315
No
>t. 11 year career programmer
>t. 11 年工龄程序员
No, but it is shrinking. There is no need for a low/mid IQ code monkey or grunt programmer anymore. It will be all higher IQ seniors who operate on more of a jack of all trades basis. They will do the analysis work, planning, implementation, testing, and deployments/infrastructure. It's already been heading this way. AI makes the meat and potatoes of coding quicker but it honestly was never the bottleneck.
不是死，是在缩。低/中 IQ 码猴或苦力程序员没需要了。剩下的是更高 IQ 的 Senior，以万事通模式运作：分析、规划、实现、测试、部署/基础设施全包。方向早就这样了。AI 让编程的肉土豆部分更快，但说实话它从来不是瓶颈。
>>109798441
yeah, thats what happens when you hit market saturation
对，市场饱和就是这结果
it stands to reason llms help with efficiency too
按理说 LLM 也确实在提效
but to say things are over thats retarded
但说一切都完了那是弱智
youre just used to a boom phase cuz its all you saw
你只是习惯了繁荣期，因为你只见过繁荣期
its a dynamic that already existed post covid
这个动态疫情后就已经存在
>>109798360
Sorry to burst your bubble but India is in a massive youth unemployment crisis right now, especially in the engineering sector. That's why you're hearing so much news related to anti government protests coming from that country right now.
不好意思泼冷水，印度正处在大规模青年失业危机中，工程行业尤甚。这就是为什么你现在老听到那个国家的反政府抗议新闻。
>>109798315
Yes. There's a couple things happening.
是的。有几件事在发生。
1. People really slack off in these companies, I'm talking real slackers. Musk was unironically right about one thing, it doesn't take 10000+ people to run a company. You can reduce a company's headcount by 70% and after a couple weeks of lag, you'll be back to routine operation.
1. 这些公司里的人真在摸鱼，我说的是真摸鱼。马斯克有一件事没说错：运营一家公司不需要 1 万+人。把员工砍掉 70%，过渡几周，日常运转照旧。
2. This was obvious but they over hired during covid.
2. 这很显然但疫情期他们超招了。
3. AI is giving some gains it has definitely made coding cheap and increased some amount of productivity, which again reduce the number of people required to keep systems up.
3. AI 确实带来了一些收益，它确实让编程变便宜、提升了一些生产力，再次减少了维持系统所需的人数。
Unless we see new companies, and new jobs being created. We already have a oversupply of developers, and it's going to get worse because of "everyone should learn to code"
除非我们看到新公司、新岗位被创造出来。开发者已经过剩，而「人人都该学编程」会让它更糟
>>109798350
It will matter once errors start compounding. We're already seeing this with microslop products and the general degradation of software quality across the industry, once corpos start trying actual no-human-in-the-loop workflows it will REALLY go to shit. There's a reason they didn't manage to outsource all dev work to India years ago, and it's the same reason AI will fail to eradicate coding
等错误开始复利叠加就有所谓了。我们在微软产品和全行业软件质量退化上已经看到了。一旦公司真搞全无人回路工作流，就会真正烂掉。他们多年前没能把全部开发外包给印度是有原因的，AI 灭不掉编程也是同一个原因
>>109798466
yeah but it sectors are usually isolated because theres an actual barrier to entry
对，但 IT 板块通常是隔离的，因为真有准入门槛
your it sectors slowdown is 100% a matter of intl market dynamics, not a domestic problem with unemployement
你们 IT 板块的放缓 100% 是国际市场动态的问题，不是国内失业的国内问题
unless its because of incompetent intl market policies
除非是因为无能的国际市场政策
but then id be expecting this being a larger crisis than the 2023 one
但那样的话我预期这会是比 2023 年那波更大的危机
which coinicides with a contraction of the it market in the isa iirc
而那波我记得恰好对应美国 IT 市场的收缩
thats the relation, im saying
我说的就是这层关系
rn your it sector is less impacted by widespread unemployement bc thats how the it sector works
现在你的 IT 板块受普遍失业影响较小，因为 IT 板块就是这么运作的
>>109798466
>>109798491
btw, is this a problem of industrial modernization/automation
顺便问，这是产业现代化/自动化的问题吗
or trade wars with idk, china?
还是跟比如说中国的贸易战？
>>109798434
>It will take a few years while businesses integrate AI
>企业要把 AI 整合进系统还得好几年
OpenAI and Anthropic don't have a few years, if they don't start producing trillions ASAP they're dead in the water
OpenAI 和 Anthropic 没有好几年，他们不能尽快开始产出万亿美元就是死水一潭
>>109798445
These companies used to hire like 50000-100000 per year back during covid, now it's down to 20000 per year. I don't know why /g/ is memeing that Indian IT sector is alive and booming lmao. Anyone living in that country knows the absolute shitshow right now. In fact quite opposite if you talk to people actually living there, not cherrypicked screenshots of job posting in India.
疫情期这些公司一年招 5-10 万人，现在降到一年 2 万。我不懂 /g/ 梗什么印度 IT 蓬勃发展 哈哈。住在那个国家的人都知道现在有多烂。事实上跟真住在那的人聊聊（而不是精挑细选的印度招聘截图），情况恰恰相反。
It's more like these companies firing 10 people in US, then posting those 10 job openings in India while firing 100s of the workers already working there.
更像是这些公司在美国裁 10 个人，然后在印度挂出那 10 个岗位，同时把已经在那工作的几百人裁掉。
>>109798511
I don't know man it looks like these VCs have infinite money. The bubble doesn't look like its popping anytime soon.
不知道，感觉这些 VC 有无限的 钱。泡沫短期内不太像要破。
if it is really done for. What can you pivot to easily and make enough money to live?
如果真完了。有什么是能轻松转向、又能赚到够活的钱的？
>>109798532
Any role where you have to plug physical things is still not dead in the water yet.
任何要插拔实体东西的岗位目前都还没死透。
I guess a network engineer or something relating to hardware?
我想大概是网络工程师或跟硬件沾边的？
>>109798462
>but to say things are over thats retarded
>但说一切都完了那是弱智
idk anon, the performance gains from chatgpt 3.5 to now has been immense. LLMs went went from shitting the bed after 20 lines of code to now handling vast codebases. Its not perfect but the corpos are happy if it all just werks
不知道 Anon，从 ChatGPT 3.5 到现在性能提升巨大。LLM 从 20 行代码就拉胯进化到能处理庞大代码库。不完美，但公司只要「能用就 happy」
Even if progress on LLMs stalls now it still seems enough for companies to continue lowering headcount with what is currently available. However LLMs continue to progress and they get better harnesses and optimizations that make the writing on the wall pretty clear at least to me
就算 LLM 进步现在停滞，现有能力似乎也够公司继续减员了。何况 LLM 还在进步、得到更好的工具链和优化，墙上写的字对我来说已经够清楚了
>>109798513
>source, reddit
>出处：Reddit
youre gonna have to do better than that
你得更拿得出手点
apparently its a competency churn
据说这是「能力大清洗」
>indians
competency churn meaning they eliminate everyone who wouldnt be there in the first place
能力大清洗意思是把本来就不该在这的人全清掉
*oracle says its because of ai
*甲骨文说是因为 AI
'sounds better than "getting scammed by thousands of people for years" and noticing it only now because yorue a bloated corpo and you strt cutting costs because of your ai-lmaoo investments
总比「多年来骗了几千人的钱、现在才发现，因为你是个臃肿大公司、因为 AI 投资，才开始砍成本」听起来好，哈哈
sure, llms may play a role in this, but this really doesnt sound like a wagiepocalypse
行吧，LLM 也许有点作用，但这听起来真不像打工人末日
>>109798509
The reason for high youth unemployment in India isn't due to a particular event. India couldn't become much of a manufacturing based economy and found its growth in the service sector in which IT/Tech jobs was the biggest domain for growth. So India over indexed on it and with AI will be severely impacted since other sectors havent really grown as much
印度青年高失业率的原因不是某个具体事件。印度没能在制造业立足，增长在服务业，其中 IT/科技岗位是最大的增长领域。于是印度在这上面过度加仓，而 AI 会严重冲击它，因为其他行业没真正长起来
>>109798580
youre just not pushing llms hard enough and thats why you dont see the fissures
你就是 LLM 推得不够狠，所以看不到裂缝
for instance, i use mine to research maths
举例来说，我拿它研究数学
i could also do a good job at generating code using even a local model im sure
我确定就算用本地模型我也能把代码生成干得不赖
but its a technology, with its applications and limitations
但它是一门技术，有它的应用场景和局限
maybe its good enough to apply paterns to things like databases and such
也许它对付数据库这类模式化东西够用了
i dont know, i always avoided webshit like the plague
我不知道，我向来像躲瘟疫一样躲网页开发
but whenever actual code of actual quality is needed
但每当需要真正的、有真质量的代码时
the things just under-perform
这些玩意就表现拉胯
and in either case theyre not good enough to operate autonomously, anyways
而且无论如何它们都不够格自主运行
and you need such code in more places than you'd suspect. like your fridge, or your car. or in vidyaslop. or os'es
而你需要的这种代码比你想的多得多。比如冰箱里、车里、游戏里、操作系统里
and i say that without a paradigm shift thats how its gonna be, incremental improvements here and there, but everything will tend towards performing at 66% lets say
我敢说没有范式转变就会一直如此：零零散散的渐进改进，但一切都会趋于 66% 的表现水平，这么说吧
good for some but not for others, like pretty much everywhere where actual code and actual quality of code is needed
对一些场景够用、对另一些不行，就像几乎所有需要真代码、真代码质量的地方
>>109798315
Fixing code that fails test suites is the only coding job left anon
修跑不过测试套件的代码，是仅存的编程岗位 Anon
>>109798641
The real blackpill is corpos have never cared about code quality.
真正的黑药丸是：大公司从来不在乎代码质量。
They only care about profits.
他们只在乎利润。
>>109798530
Doesn't seem like it to me, they're going for IPO because the private market dried up. VCs started demanding their ROI
我看着不像，他们冲 IPO 是因为私募市场枯了。VC 开始要回报了
>>109798601
i dont think thats whats happening.
我不觉得是这么回事。
ai is involved, yes, but in the measure where western companies, THE clients of india's it sector, are now tightening the belt because theyre all in on ai infrastructure and everyone is starting to run out of money
AI 确实有 involvement，但体现在：印度 IT 板块的客户——西方公司——正在勒紧裤腰带，因为他们 all in AI 基础设施，大家开始缺钱
i think thats whats happening
我觉得就是这么回事
although maybe the agentic focus of astra will have a large impact on a large segment of the market now that i think of it
不过想想，Astra 的 agentic 定位现在可能对市场很大一块造成大冲击
all it lacks is a voice to text module and i could really see it become a force multiplier
它就差个语音转文字模块，我真觉得它能成为力量倍增器
its too soon to call the actual impact of that advancement, i dont think theres enough adopters yet
这进展的实际影响现在下结论太早，我觉得采用者还不够多
>>109798666
Satan is right
撒旦说得对
>>109798675
>ai is involved, yes, but in the measure where western companies, THE clients of india's it sector, are now tightening the belt because theyre all in on ai infrastructure and everyone is starting to run out of money
>印度 IT 板块的客户——西方公司——正在勒紧裤腰带，因为他们 all in AI 基础设施，大家开始缺钱
>i think thats whats happening
>我觉得就是这么回事
This is accurate for the current situation. Also these clients are trying to open GCCs in India to see if they can save even further than low cost Indian IT vendors by hiring cheap Indian IT labor directly. However these GCCs are hiring only a fraction of the the people getting laid off in India
这符合现状。另外这些客户正试图在印度开 GCC（全球能力中心），看能不能比低价印度 IT 外包商省得更多、直接雇便宜的印度 IT 劳力。但这些 GCC 只招了印度被裁人员的一小部分
>>109798666
nah, yeah, i knew about that
嗯对，这个我知道
but better code = smaller chip
但更好的代码 = 更小的芯片
and smaller chip = better margins
更小的芯片 = 更好的利润
quality of code is not "off" by default
代码质量不是默认「关」的
its the result of the consideration of "how big of a chip we need to run it, its cost" vs "how much will it cost us to improve the code to run on a smaller chip"
它是「我们需要多大芯片来跑它、成本多少」vs「改进代码跑在小芯片上要花多少」权衡的结果
this translates into servert code etc, as in: better programs can be deployed on cheaper compute
换算到服务器代码上就是：更好的程序可以部署在更便宜的算力上
and agent might admin a database but its a human thats gonna be at the core of writing it
agent 可能管理数据库，但写它的核心还是人
dw, im talking about this in marketing mode now
别慌，我现在是用营销话术在聊这个
>>109798315
Coding is cooked and transgender
编程完蛋了，而且变性了
>>109798641
American car companies would not recall cars with severe defects that could kill people if handling it in court was cheaper. You really think corpos give a shit about code quality? Have you not noticed the quality deteriorating each decade
如果走法律程序更便宜，美国车企连能杀人的严重缺陷车都拒不召回。你真以为大公司 care 代码质量？你没注意到质量每十年都在退化吗
>>109798490
Most of that's intentional because they know Linux still sucks ass and they can get away with it. Macbook Neo is the only thing worrying them right now.
大部分是故意的，因为他们知道 Linux 还是很烂、可以蒙混过关。MacBook Neo 才是眼下唯一让他们担心的东西。
it's cooked bruhsir :cry: :rose: only cracked jeets from IIT will earn zilllions in IT now ser :broken_heart: just put the fries in the bag SON :cry: :cry: :water_gun:
完了老兄 :cry: :rose: 现在只有 IIT 出来的开挂印度佬能挣 IT 的亿点点钱了 ser :broken_heart: 把薯条装袋里 SON :cry: :cry: :water_gun:
>>109798709
its really tough to call just how much ai is a factor in all this
AI 在这一切里到底占多大因素，真的很难说
a huge portion of the people who werent rehired could have been fat to cut off in a way or another
很大一部分没被回聘的人，本来就是可以以某种方式被割掉的赘肉
cf. twitter and elon's cuts
参见 Twitter 和马斯克的裁员
9000->1.500
in the other article i posted microsoft were clear that its a competency shakedown
在我发的另一篇文章里，微软明确说这是能力大清洗
they kinda cant mention ai cuz theyve been seen forbidding its use to its workers
他们有点没法提 AI，因为他们被发现禁止员工用
and given my experience with shatbots, and what i know about the technology
鉴于我用屎机器人的经验，和我对这技术的了解
all in all, i think ai has had a very limited impact until now (if we ignore retard early adopters who ended up rehiring their staff after a couple months)
总而言之，我认为 AI 至今影响非常有限（如果不计那些几个月后又把员工全雇回来的弱智尝鲜者）
and for the astra model's impact were gonna have to wait several months, so that people have the time to adopt it, integrate into workflows, and the job market reacting
至于 Astra 模型的影响，得等好几个月，让人们有时间采用、整合进工作流、然后看就业市场反应
>>109798724
Maybe for microslop but every piece of software is hot garbage right now, not just Windows. Every single app is bugged to fuck
也许微软如此，但现在每款软件都是热垃圾，不只 Windows。每个 App 都 bug 到爆
>>109798723
calm down and read carefully:
冷静下来仔细读：
if 50k worth of manhours of code quality can reduce the production costs of a 1.000.000 run items by 5 bucks
如果 5 万美元人时的代码质量投入，能把一百万件产品的成本每件降 5 块
then these manhours will be invested into it
那这些人时就会被投进去
its how things work. this is how money gets made
事情就是这么运作的。钱就是这么挣的
if you are studying to work in tech, you are a fucking RETARD and will end up with crippling debt, and you will have successfully ruled yourself out of any skilled work because you wasted time on that.
如果你正在学科技类专业，你是个他妈的弱智（RETARD），最后会背上压垮人的债务，而且因为你把时间浪费在这上面，成功把自己排除出任何技能型工作。
do not study CS, do not waste time in tech, do not enter tech unless you know people who will hire you.
别学 CS，别在科技行业浪费时间，除非有人会雇你，否则别进科技行业。
It's simple arithmetic. There are only jobs for 2-4% of CS graduates today. If you're not top 1%, why bother? There are many other pursuits with a far better reward to effort ratio.
就是道算术题。今天 CS 毕业生只有 2-4% 有岗位。你不是前 1%，折腾什么？有太多投入产出比好得多的路。
>>109798441
>I'll be honest I'm a jeet and things aint rosy here either.
>老实说我就是印度佬，这边日子也不滋润。
That's implying your literal shithole had some better time instead of being the extinction level event of humanity.
言下之意你那字面粪坑曾有过更好的时候，而不是人类灭绝级事件。
>>109798315
No, I dont think it is, LLM generated coding is very expensive. In my small company the developers all have $20 codex plans, and in this month alone the tokens they used at API pricing would cost $3200, in not really willing to pay to oAI $10000 extra as the performance is pretty much the same, despite the code being sloppier and the developers losing the grasp on the codebase (I can see it already on code reviews with frequent code duplication, and so on).
不，我不觉得。LLM 生成代码非常贵。在我小公司，开发者都有 $20 的 Codex 套餐，仅这个月他们用掉的 token 按 API 价格算是 3200 美元。我不太愿意额外付给 OpenAI 一万刀，反正性能基本一样，代码还更草，开发者对代码库正在失去掌控（代码评审里频繁出现重复代码，我已经能看到了）。
Also, I tried it myself, if you tediously guide the expensive model (Astra) you can get some ok results, but if you want to be "cheap" and use something like Lua or Terra you quickly see that they are useless.
另外我自己试过：如果你耐心地一步步喂贵模型（Astra），能得到还行的结果；但想「省钱」用 Lua 或 Terra 之类，你很快发现它们没用。
>>109798906
skill issue + plus it's gonna be 10x-100x cheaper for current level of performance this time next year
技术问题 + 明年这时候同水平性能会便宜 10-100 倍
>>109798935
>10x-100x cheaper
>便宜 10-100 倍
>in 1 year
>1 年内
you sound mentally retarded, cargo cultist
你听着就像智力障碍，货物崇拜者
>>109799209
10x is already guaranteed by Rubin, Cerebras and Jalapeno
10 倍已经由 Rubin、Cerebras 和 Jalapeno 锁定
100x is plausible with algorithmic gains and distillation as shown by smaller models catching up
100 倍靠算法进步和蒸馏也说得通，小模型正在追上来就是证明
If we're being positive here one of my favorite improvements with AI is finding/fixing bugs. It seems really good at it and that's a pain point that it clearly solves without any regression. Feed it an error log and any other context and it's mint at triangulating the issue and offering a fix.
往好处说，AI 我最喜欢的改进之一是找 bug/修 bug。它似乎真的很擅长，而且这是个没有退步地被明确解决的痛点。喂它一份错误日志和任意上下文，它定位问题、给出修复的表现是尖的。
>>109798364
>>>>>>>>>>>"we"
Every response starting like this can be safely ignored.
所有这样开头的回复都可以安全忽略。
>>109799221
>10x improvement in energy efficiency
>能效提升 10 倍
imagine believing that
想象一下信这个
it has to be because youre mentally handicapped
那只能是因为你智力障碍
>>109799221
10x cheaper? That would be hard to achieve. How do you explain Astra and Fable 5.1 being more expensive than their predecessors? Check the "cost per task" on Artificial Analysys.
便宜 10 倍？那很难。那你怎么解释 Astra 和 Fable 5.1 比前代更贵？去看看 Artificial Analysis 的「每任务成本」。
>>109799322
hes a victim of a debilitating cognitive affliction
他是一种 debilitating 认知疾病的受害者
he probably was sniffing petrol as a kid
他小时候八成在闻汽油
and took the habit with him into adulthood
然后把这习惯带进了成年
>>109799348
I try to be reasonable but all the AI guys really seem to be cargo cultists that believe in infinite prosperity somehow? produced by a chatbot glued to a for loop.
我尽量讲道理，但那帮 AI 党真的看起来全是货物崇拜者，不知怎么就相信 prosperity 无限，由一个粘在 for 循环上的聊天机器人产出。
>>109799403
a portion of them are trolls from an altchan
其中一部分是某个替代匿名版来的钓鱼
but they cant account for the reddit brigades, which in turn cannot account for the wider media response
但他们解释不了 Reddit 爆吧团，而 Reddit 爆吧团又解释不了更广泛的媒体反应
theres a wider martketing push for chatbots
背后有更大规模的聊天机器人营销推进
which is kinda fucking retarded, the technology is uncooked yet
这挺他妈弱智的，技术还没煮熟
so theyre gonna blow their load now and the research will staganate again because money will dry up
所以他们现在要把弹药一次打光，然后研究会再次停滞，因为钱会干
now we might be already at the desperation stage, idk how much more money they can throw into the pit
现在我们可能已经在绝望阶段了，不知道他们还能往坑里扔多少钱
>>109798315
Yes, it's dead.
是的，死了。
It's awful code that only barely functions enough to pass a test. If you want to rely on that, go ahead, none of you will ever finish more than 15% of a project so why does it matter?
那是勉强能过测试的烂代码。你要依赖它请便，反正你们谁也做不完项目的 15%，有什么所谓？
Just stop posting your slop everywhere.
别再把你的垃圾到处发了。
>>109799403
>>109799444
buuuuut...
...that said...
……话虽如此……
the couple demos i saw with astra do look promising
我看到的几个 Astra 演示确实看着有戏
this does look like a product one could sell-
这看起来像是能卖的产品——
a sort of smart macro for office work
某种办公室工作的智能宏
maybe the whole 3d modeling thing too
也许整个 3D 建模那块也算
but thats in the capacity of an ai agent, the llm as a force multiplier but certainly not autonomous
但那是在 AI agent 的能力范畴：LLM 作为力量倍增器，但绝不是自主的
>>109799444
>>109799492
cont
and to really make it work they should package the thing into local instances
要真让它成事，他们应该把它打包成本地实例
because with the recent math paper fiasco information leaks are much more present in the normies heads
因为最近的数学论文翻车事件后，信息泄露在 Normal 人脑子里存在感强多了
>>109798315
>give it to me straight
>直说吧
low-to-mid tier code, yes. State-of-the-art algorithms that rely heavily on first principles construction are still untouchable afaik.
中低档代码，是的。高度依赖第一性原理构建的最先进算法，据我所知依然碰不得。
>>109799550
i think the bar is even lower
我觉得门槛还得更低
i would be surprized if chatbots can produce coherent, project-wide optimizations
聊天机器人要是能产出连贯的全项目级优化，我会很惊讶
i wouldnt call it state of the art, though. just competent, quality code
不过我不会称之为最先进。就是称职、有质量的代码
>>109799492
The demos are nice indeed, but they're just that, demos, mvps. If you want to turn them into "ok shit you can sell" that will be a few more tokens, and naturally problems will appear (as they do even with human handcrafted things) how are we supposed to solve these problems if we have no deep knowledge of the system innerworkings? This could be something like "this 3d asset has too much poligons" or "this specific websocket event is arriving at the wrong order" Could we just prompt the solutions away? Don't seem to be the case, at least for now, as we have barelly any robust "vibe-coded" software.
演示确实好看，但它们只是演示、MVP。想变成「能卖的货」还得再烧一堆 token，而且自然会出现问题（人手做的也一样会出）——如果我们对系统内部运作没有深度了解，该怎么解决这些问题？可能是「这个 3D 资产多边形太多」或「这个特定 websocket 事件到达顺序错了」。我们能用 prompt 把解决方案念出来吗？至少现在看不太行，因为我们几乎没有健壮的「氛围编程」软件。
I agree that it's a force multiplier in some cases, especially as a search engine, but that isn't revolutionary or lucrative as the LLM companies now need it to be.
我同意它某些场景下是力量倍增器，尤其当搜索引擎，但这既不革命也不暴利，而 LLM 公司现在需要它是。
>>109799582
>if chatbots can produce
>聊天机器人要是能产出
*cant'T
>>109799582
Just yesterday I did some work using Astra. It was really good at producing example implementations of the sort of algorithms I need which is great because without AI my manual workflow would've taken probably weeks. It started to stumble when I gave it things that don't really exist in public code, and bear in mind I have very stringent performance requirements. Now of course a regular dev would also fail if required to do the same thing with no context. So it seems knowledge for which there is barely any training data is still a formidable moat for both AI and humans.
就在昨天我用 Astra 干了点活。它生成我需要那类算法的示例实现很在行，这很棒，因为没有 AI 我的手工流程大概要几周。当给它不存在于公开代码里的东西时它开始磕绊——记住我对性能要求极其苛刻。当然，普通开发者没有上下文做同样的事也会失败。所以几乎没有训练数据的知识，对 AI 和人类而言都仍是可怕的护城河。
>>109799587
>but that isn't revolutionary or lucrative as the LLM companies now need it to be.
>但这既不革命也不暴利，而 LLM 公司现在需要它是。
thats gonna be a hard filter for most of them
对它们大多数来说这是道很难的过滤
i expect companies that have either other sources of revenue, like google meta, msft;
我预期活下来的是要么有其他收入来源的公司，比如谷歌、Meta、微软；
or companies that had a much narrower focus like perplexity, to survive
要么专注面窄得多的公司，比如 Perplexity
danger zone- anthropic. i wouldnt be surprised if they survive, nor if they die
危险区——Anthropic。他们活下来我不意外，死了也不意外
theyre amongst the big players but theyre the ones the closest to break-even afaik.
他们属于大玩家，但据我所知是离盈亏平衡最近的。
and then openai. i woudl expect them to die before astra
然后是 OpenAI。我预期他们比 Astra 先死
now with a focus on agentic workflows, this might change things for them
现在聚焦 agentic 工作流，这可能会改变他们的处境
its new, i dont have access to that model, i think the reasonable thing is to wait 2-6 mos before emitting a definitive opinion
它很新，我没拿到那个模型的访问权，我觉得合理的做法是等 2-6 个月再下结论
2mos needed to evaluate the product by the public id say
公众评估产品需要 2 个月
and 6mos to undeniable effects of the technology on employement dynamics, to definitively close the subject
以及 6 个月来看这项技术对就业动态的无可辩驳的影响，从而彻底关闭这个话题
>>109799667
ah, so basically same old chud gpt i guess
啊，所以基本还是那个老款红脖子 GPT
i heard a couple things that basically boil down to:
我听到几件事，基本归结为：
astra hes been worked on with a focus on agentic work
Astra 研发时聚焦 agentic 工作
one of the key highlights is that its token usage is 70% lower compared to previous models
主要亮点之一是 token 用量比之前的模型低 70%
but in what concerns programming it ranks pretty close to the competition and is not a major departure from past performance
但编程方面它排名跟竞品很接近，跟以往表现相比没有重大突破
in a couple rankings it ranks even lower than previous models
在某些榜单上它甚至比之前的模型排名更低
but myeah, again, 70% less tokens which is the most impressive gain in this alongside the agentic part imo
不过嘛，再说一次，token 少 70% 是这里最亮眼的收益，加上 agentic 部分，在我看来
so theyre getting somewhere product-shaped it seems
所以他们似乎做出了点产品形状的东西
and this may change things
而这可能会改变局面
so id be careful about forcasts
所以对预测要小心
its still a bubble but things may turn out to be less catastrophical than the worst scenarios
它还是泡沫，但结局可能没有最坏情景那么灾难
>>109798315
There will be a human in the loop. I can't quite put what the role fulfilled will be - tester, developer, architect, pm or something else. The amount of jobs lost will depend of how capable the AI gets and how much it costs. Both are big unknowns atm.
肯定还是会有个人留在流程里。这个角色到底是测试、开发、架构师、项目经理还是别的，我现在也说不准。到底丢多少工作，取决于 AI 能变多强、成本多少钱。这俩眼下都是大大的未知数。
>>109799794
The only roles left would be SMEs automating their tribal workflows that can't RLHFed. No need for human involvement in technical work.
剩下的角色就只有领域专家(SME)了，把那些没法用 RLHF 训出来的口传经验流程自动化掉。技术工作根本不需要人类参与。
>>109798333
>3
>more
>years
2 years into 6 months until all code is AI written by the way
顺便说一句，「还有 6 个月所有代码都是 AI 写的」这话已经说了 2 年了。
>>109798350
The problem is it only looks like it works. Like it will work with 1 input and fail everywhere else. And it's always massive security issues.
问题是它只是看起来能用。比如一种输入能跑通，换别的全挂。而且永远伴随海量安全问题。
>>109799883
>looks at our meaningless numbers going up!!
>看着我们那些毫无意义的数字往上涨！！
>oh cool.. who made these benchmarks btw?
>哦不错嘛..顺便问一句，这些跑分是谁做的？
>...we did
>……我们自己做的
>>109799883
wook at dis chawt! it say my wobot fwend smart
你看介个表表！它说我的机器人朋友好聪明哦。
Stop posting your slop dogshit, stop hosting it on github, stop wasting electricity on shit.
别再发你的垃圾 AI 屎了，别再往 GitHub 上挂了，别再浪费电了。
>>109799904
its worse than that
实际情况比这还糟。
>even in the numbers we provide ourselves
>连我们自己提供的数字都是这样
>we cant declare more than single digit improvements with a straight face
>提升连个位数都不到，我们都没法脸不红心不跳地宣传
>>109798674
It's not because the private market dried up, investors do their thing for returns. Returns need a liquidity event. That would be ipo. The VCs are definitely breathing down their neck about it. Endless VCs are looking for ways into anthropic's cap table. They're playing games like funding companies like legaltech and ai x bio in hopes anthropic buys out these companies to get them the equity this way.
不是因为一级市场没钱了，投资人做投资图的就是回报。回报需要一个流动性事件，也就是 IPO。VC 肯定在后面死死盯着催他们。无数 VC 想挤进 Anthropic 的股权表，各种玩花样，比如投 legaltech 和 AI+生物这类公司，盼着 Anthropic 收购这些公司，好通过这条路拿到股权。
>>109798756
This. Even strong AI advocate talk about it.
就是这个。连坚定的 AI 派都在谈这事。
>>109798773
This is only true in theory. In practice businesses are super inefficient and never do this math, they literally only operate on vibes.
这只在理论上成立。现实里公司效率极低，从来不算这种账，做事全凭感觉。
>>109799923
it's actually worse than that
实际情况比这还糟。
those incremental improvements require an exponential increase in computing power
那些渐进式的提升，得靠算力指数级增长才能换来。
>>109799550
>>109799582
Correct. It's particularly bad in large microservice setups where it can't even understand interface contracts even if it has full codebase access and knows the config. But even for small things like writing a simple parser for a clear but tedious file format, it still can't get it right and will make shit up and the most hilariously wrong code because it 'looks right' in the output format it gets from it so it decides the tests are passing etc.
没错。在大型微服务架构里尤其糟糕，哪怕它有整个代码库的访问权限、也知道配置，还是看不懂接口契约。就连写个简单解析器这种小事——格式清楚但繁琐——它也搞不定，会瞎编，写出错得离谱的代码，就因为输出格式上「看起来对」，它就判定测试通过了之类的。
>>109799727
>one of the key highlights is that its token usage is 70% lower compared to previous models
>一大亮点是它的 token 消耗比之前的模型低了 70%
Pure marketing, not even remotely true
纯营销话术，一点边都不沾。
>>109799950
yeah thats the big one
对，这才是最大的问题。
staring everyone involved in the face
明晃晃摆在所有相关人眼前。
and everyone involved pretending it doesnt exist
而所有相关人都在装看不见。
the second aibros started thinking about buying nuclear reactors they should have understood they have a scaling problem
AI 兄弟们一开始琢磨买核反应堆的时候，就该明白他们碰上扩展性瓶颈了。
>>109799943
and this is why theres room for competition for competent people
所以这才是留给有真本事的人的竞争空间。
i dont mind the old fortunes being lazy and bloated
我不介意那些吃老本的懒虫又肥又肿。
but thats an example of a failure mode.
但那只是失败模式的一个例子。
it doesnt support your argument in the slightest, its not a normal operating regime
这完全支撑不了你的论点，那不是正常运转状态。
>>109799883
If you actually used the models you'd know it's bullshit. If you actually worked on products using them like I do you would even have benchmarks internally to prove it.
你要是真用过这些模型，就知道这就是扯淡。你要是像我一样真在拿它们做产品，公司内部早就有跑分数据能证明这一点了。
But alas, you are merely an ai psychosed 65 IQ retard.
可惜啊，你只是个被 AI 洗脑的 65 IQ 智障。
>>109799998
>and this is why theres room for competition for competent people
>所以这才是留给有真本事的人的竞争空间
Absolutely, but there is one key problem: except in software with limited hardware requirements for hosting, you need starting capital. Nobody is providing this capital to competent people right now (total draught in the funding landscape across grants, angels and VCs).
完全同意，但有个关键问题：除了托管硬件要求不高的软件之外，其他行当都需要启动资金。现在没人给有本事的人提供这种资金（补助、天使、VC 整个融资环境全干了）。
>>109799988
>self published numbers
>自己发布的数字
i dont vouch for them, im just repeating what i heard
我不给这些数字背书，我只是复述听来的话。
>>109800002
you sound angry
你听起来很生气
>>109800012
You sound like a 65 IQ retard with ai psychosis.
你听起来就是个被 AI 洗脑的 65 IQ 智障。
>>109800007
you generate capital
你可以自己滚出资本来。
thats more or less the idea behind SME's
中小企业(SME)背后的逻辑差不多就是这个。
you dont need to start with a bilion usd budget and your crown-achievement project
你不需要一上来就拿着十亿美元的预算去做你的毕生杰作。
in fact, thats risky. its your first venture, build small, scale up, solve problems as you go
事实上那才叫冒险。这是你的第一次创业，先做小，再滚大，遇到问题边走边解决。
>>109800038
Pretty much everything outside software requires huge upfront capital to be able to get a first customer. After that you can generate capital all you want the normal way, but until then you're screwed. It's also not the kind of capital you can get if you're not born into wealth or 70 years old with a house to remortgage at full equity at least.
软件以外的几乎所有行业，都需要巨额前期资金才能拿到第一个客户。之后你就可以按正常方式随便滚资本了，但在那之前你就死定了。而且这种钱，你要么生在富贵人家，要么至少 70 岁有套房子能以全额净值做二次抵押，否则根本拿不到。
"Small loan of a million dollars" and all that.
「小小的一百万美元贷款」那一套呗。
That money is sometimes needed as soon as the prototype phase, following discovery, business planning (nobody cares about that nowadays either) and LOIs.
这笔钱有时原型阶段就得花，之前还要做需求探索、商业规划（现在也没人在乎这个）和意向书(LOI)。
In other businesses, you can do a prototype but can't make an MVP without that money and nobody will pay upfront without the prototype.
有的行业你能做原型，但没钱就做不出最小可行产品(MVP)，而没有原型也没人会预先付钱。
In other businesses still you can do the MVP but can't distribute without mass production capacity, and can't generate enough capital to afford mass production in a lifetime. As you can't produce fast enough, the audience just vanishes before you even get there.
还有些行业你能做出 MVP，但没有量产能力就没法铺货，而量产的钱你这辈子都攒不齐。产量跟不上，等你缓过来受众早就跑光了。
Again, not everything is a SaaS.
再说一遍，不是什么行业都是 SaaS。
>>109800078
nah, youre just ossified in your thinking
不，你就是思想僵化了。
consider this:
你想想这个：
vidyaslop.
minimum upfront capital
前期资金需求最小。
youre quite literally elling your ideas directly
你基本上就是在直接卖自己的点子。
this ia a counterexample to what youre saying thats why ill tell you that classical business approach still works
这就是你那套说法的反例，所以我才告诉你传统做生意那一套依然管用。
and note- its all you can have access to to it better do
注意了——你只能靠这个，所以最好靠谱点。
>>109800102
... vidya is software, anon...
……游戏也是软件啊，老哥……
>>109800112
yeah but im talking about it as counter example to your doom and gloom business model
是，但我拿它举例，就是为了反驳你那套悲观绝望的生意经。
neh, now is still the time to start in robotics and be amongst the early birds
不不，现在入场机器人领域还来得及，还能当个早起的鸟。
theres tons one could be automating. basically, everything, lol
能自动化的东西多得很。基本上啥都能自动化，哈哈。
you need some capital but not insane amounts of it
是需要点资金，但不用多到离谱。
>>109800138
It's not a counter example, I specifically said that software is the only exception. Robotics is a great example where you can't make it without early cash infusion after prototype and before mvp at best.
这不算反例，我明确说过软件是唯一的例外。机器人就是最好的例子：原型做完之后、顶多到 MVP 之前，必须有一笔早期资金注入，否则根本做不出来。
>>109800156
>early cash infusion
>早期资金注入
depends on the amount of cash were talking about
得看咱们说的是多少钱。
i bet youre thinking in volumes where you wouldnt consider my initial planned budget even as a cash injection
我打赌你脑子里想的量级，大到连我的初始预算都不配当一笔注入资金。
the secret sauce is you dogfood your technology, which you build up incrementally, to outcompete competitors in markets other than robotics
秘诀就是先自家用自己的技术，一点点积累，然后在机器人以外的市场把对手干趴下。
once you get your prototypes down, you go the rest of the way in finding suppliers, certifying your shit etc
等原型搞定，剩下的就是找供应商、给你的玩意儿做认证之类的。
by which point you probably should have the liquidity available to do so
到那时候你手里应该已经有足够的流动性来干这些事了。
>>109800156
>my initial planned budget even as a cash injection
>我的初始预算都不配当一笔注入资金
my *minimal initial planned budget even as a cash injection
我*最低限度的初始预算都不配当一笔注入资金
>>109800217
>>109800239
You've clearly never even so much as talked to anyone working at a non-tech startup, let alone founded a company. I have extensive experience in the former and have founded 2 companies so far. The point is you're so far off the mark it's hard to think about how you managed to get there.
你显然连一个非科技创业公司的人都没聊过，更别说创过业了。前者我有大量经验，而且到目前为止创过 2 家公司。重点是你偏得离谱，我都想不出来你是怎么偏成这样的。
>>109800247
i dont think you have done what youre pretending to have
我觉得你根本没做过你假装做过的事。
i think you were a wagie somewhere and you completed the dots with fanfic, actually
我觉得你就是 somewhere 当过打工人，剩下的全靠同人脑补的。
at best
往顶了说。
>>109800262
I don't understand why you bother to post since you clearly have no idea about the topic and instead decide to live in delusion and deny reality when it is presented to you by people with actual experience.
我搞不懂你发帖图什么，明明对这话题一窍不通，却宁愿活在幻觉里，有实际经验的人把现实摆你面前你还否认。
>>109798315
>coding
yes
>software engineering
>软件工程
no
>>109800288
you have zero experience though.
可你的经验是零啊。
what youre saying is a cookie cutter regurgitation of angsty barely-adult-anons, basically
你说的话，基本就是把那群焦虑到不行的半大孩子老哥的陈词滥调原样吐了一遍。
at the very best maybe its a lived experience from the eprspective of a wagie at a big corpo, but not even in marketing
往顶了说，顶多算大公司打工人的切身体会，而且连市场岗都算不上。
and youre raging at the basics of economics applied to the 21st cen.
而你在对着应用于 21 世纪的经济学基本常识暴怒。
in a case where said same basics stare you in the face whenever you order a burger in your local waffle house
这些常识你在家楼下华夫馆点个汉堡时都明晃晃摆在你脸上。
>t. im an elite navy salesman with 300 confirmed contracts with fortune 500 companies...
>署名：我是精英海军销售，手握 300 份世界 500 强企业的确认合同……
sure buddy
当然，伙计
>>109800320
You're having a melty because someone with experience explained to you how business development works at a very basic level...
有人凭经验跟你讲了点生意拓展的基本常识，你就破防了……
>>109800337
>n-no, ur angry
>不、不是我急，是你急
sure buddy, kek
行行行，你说是就是，哈哈。
>>109798315
>have enterprise accounts for every AI model provided by my main client
>我主要客户给我开了所有 AI 模型的企业账号
>vibe code all day since its just front end work
>整天挂机氛围编程，反正只是前端活
My job is mostly just talking to clients and managers at this point. Work on your social skills
到这个份上，我的工作基本就是跟客户和经理聊天。练练你的社交能力吧。
>>109798360
>cargo cultist
>货物崇拜党
Well ya. You know which BRICS country they are from.
那可不。你知道他们来自哪个金砖(BRICS)国家。
>>109798315
Cooked for juniors. t.senior
初级岗完蛋了。署名：一个高级工程师。
>>109798315
Yeah. Coding is fucked. Sorry m8.
是啊。写代码这行完蛋了。节哀，兄弟。
>>109798530
Infinite money but no return???
无限烧钱但是没有回报？？？
>>109798591
Oracle blowing up would be amazing. Lary is already selling personal shares to afford the penalty to walk away from his media mergers lel
甲骨文(Oracle)爆雷就精彩了。拉里已经开始卖个人持股，好凑钱付违约金退出他的媒体并购了，哈哈。
>>109798490
As long as software gets suckier across the board, then people will quicly get used to it. It won't be a problem.
只要软件全线一起变烂，人们很快就会习惯的。不成问题。
>>109799492
>demos i saw look promising
>我看的那些 demo 看起来挺有戏
Never ever confuse a demo with a product.
千万别把 demo 和产品混为一谈。
>>109798773
You're missing the creation side of things btw.
顺便说一句，你漏了「创造」这一侧。
>>109799943
yep, mostly its because businesses only look at short term profitability. Thats why I have companies that have ancient code that is a pain to maintain because they dont want to give us a budget to rebuild everything from scratch but become infinitely more profitable and efficient.
是的，主要就是因为公司只看短期盈利。所以我手上有一堆公司留着祖传老代码，维护起来痛不欲生，因为它们不肯给我们预算从头重写，哪怕重写之后能无限更赚钱更高效。
>>109799924
The public market is there liquidity. This whole thing is fucked up hahahah
公开市场就是他们的流动性。这整件事烂透了，哈哈哈哈。
>>109800288
The other anon is right. Nothing of value is conveyed via your words.
那位老哥说得对。你的话里没有任何有价值的信息。
>>109800714
That's the definition of IPO. IPO is specifically a liquidity event for investors. It has always been like this. The alternative is an acquisition.
这就是 IPO 的定义。IPO 专门就是给投资人办的流动性事件。一直如此。另一个选项是被收购。
>>109800738
Whatever helps you sleep at night, "other anon". I'm sure you know better than people who actually do these things for a living.
你怎么舒服怎么来吧，「另一位老哥」。我相信你比那些真靠这个吃饭的人懂行。
>>109800824
Well you imply that you are one of these people. Yet are completely incapable of conveying meaning in your words. Which I guess aligns...
呵，你暗示自己是那种人。可你说的话完全传达不出任何意思。倒也算自洽……
>>109800808
Oh absolutely. But once a company does go public. It usually remains public and then goes to shit. Because it has to go to shit. Because infinite growth is not real. So the people who actually made it something worthwhile cash out. And leave! And then what remains are the saars in an attempt to "grow" the company.
哦太对了。但公司一旦上市，通常就会一直公开交易然后烂掉。因为它不得不烂。因为无限增长是假的。于是真正把它做出价值的人套现离场！剩下的就是一群「萨萨尔」努力「做大」公司。
>>109798315
What do you pivot to? Does IT have more of a future? Do you just cut your losses and attept to enter the trades? I watched some plumber/electrician videos last night and realized how little I'd be interested in ether. Do you just suck it up and do something you have no interest in for 40 years?
转行转去哪？IT 还有前途吗？还是割肉认赔去干技工？我昨晚看了些水管工/电工的视频，发现自己对哪个都没兴趣。难道就这么忍下来，干一件自己毫无兴趣的事干 40 年？
>>109798906
>developers losing the grasp on the codebase
>开发者对代码库失去掌控
it happens really fast once you have over the reins
一旦把缰绳交出去，失控来得飞快。
I can see some benefit in allowing people to use LLMs to generate shit they would just be typing out, but it's not the silver bullet that I once thought it was
用 LLM 生成那些本来也只是机械敲出来的东西，我觉得有点用处，但它不是我曾经以为的银弹。
it seems like the killer feature of generative ai is that it allows people who to do things on a computer that they haven't have done before more easily
生成式 AI 的杀手锏似乎是：让那些以前在电脑上干不成事的人，现在能更容易地干成点事。
>>109800247
Most techfags are terminally online incels who think their hyper fixation on trivial minutiae makes them superior to everyone else. They're no different from NPCs that believe everything they see online, but they have an ego about it which makes them extremely susceptible to online propaganda/bullshit. Most of them have no understanding of the real world.
大部分技术宅就是重度网瘾的老光棍，觉得自己对鸡毛蒜皮细节的超级执念让他们高人一等。跟网上看到啥信啥的 NPC 没区别，还自我感觉良好，所以特别容易被网上宣传/垃圾信息洗脑。他们大多数根本不懂现实世界。
>>109801694
I think you have to get lucky to even pivot to a different industry. I've been doing prod support for the last couple years hoping I could move into a more serious site reliability engineer role.
我觉得就算转去别的行业也得靠运气。过去几年我一直在做生产支持(prod support)，指望能转成正经一点的站点可靠性工程师(SRE)。
but seriously I was ecstatically referred for a job to be an AI champion in an org that's big on power BI. I've got over 10 years experience doing backend enterprise software development prior to this prod support and sent them a link to my personal vibecoding agent.
说真的，有个重度使用 Power BI 的机构，我被人兴奋地内推了一个「AI 冠军」岗位。做生产支持之前我有 10 多年后端企业软件开发经验，我把我自己搭的氛围编程 agent 的链接发给了他们。
nothing lol. didn't give a shit. it really is totally over. I didn't believe it when you guys told me that. you guys have been telling me that for years but never was a problem.
啥回音都没有，哈哈。人家根本不当回事。这下是真的彻底完了。你们当初跟我说的时候我不信。你们这话说好几年了，可一直没出过问题。
>>109800515
They don't need a return. They can get loans forever from the money printer.
他们不需要回报。有印钞机在，贷款能无限续。
>>109801694
>>109801994
What about something like industrial/automation controls? Is it possible to get into as a CS fag if you go through the electrician route? Maybe it's not even worth it.
那工业/自动化控制呢？CS 佬走电工路线能不能进这行？也许根本不值当。
>>109801974
>Most of them have no understanding of the real world.
>他们大多数根本不懂现实世界。
Most people are like that. Not working in tech doesn't make you have more "street smarts".
大多数人都这样。不在科技行业干活也不代表你就更有「街头智慧」。
>>109802370
I never said otherwise incel. Thanks for proving my point kek
我从没说过相反的话，老光棍。谢谢你亲自证明我的观点，哈哈。
>>109798360
>no software Dev jobs
>软件开发岗位没了
>all India
>全是印度
gr8 b8 m8
>>109802407
You implied otherwise. You're a just a projecting loser. You obviously have zero experience in tech. The fact that you use kek in your response says it all.
你就是在暗示相反的意思。你就是个只会投射的失败者。你显然一点技术经验都没有。你回复里用「kek」这件事本身就说明一切了。
>>109802428
Wrong again incel. Try again kek
又错了，老光棍。再来一次啊，哈哈。
>>109802452
You're the incel loser. I won't try again. Keep on seething.
你才是那个失败光棍。我不奉陪了。继续破防吧。
>>109798315
I have a dev friend with years of exp that is about to start doing uber eats lol
我有个干了好几年开发的朋友，马上要去送 Uber Eats 外卖了，哈哈。
Shit is over
彻底完了。
>>109802459
Wrong again incel. Run away. Im glad you're running away cause this is just getting sad man.
又错了，老光棍。跑吧。你跑了我还挺高兴，因为这场面已经惨得没眼看，兄弟。
>>109798315
The stupid thing still won't vibecode a compiler. It just makes a shitty academia tier algo and asks me to program the real code.
这蠢货还是不会氛围编程写编译器。只会给我弄个学术垃圾级别的算法，然后让我来写真正的代码。
>>109798315
I slopcoded a thing and it works and I know 0 code. However I don’t know shit about optimization do I’m kind of lost there.
我用 AI 屎码了个东西，能跑，而我代码量为 0。不过优化我一窍不通，所以在那块有点懵。
>>109798315
It’s over gang coding is cooked fr fr no cap. it's straight up NOT bussin anymore
完了家人们，编程彻底凉了，真的真的不骗你。现在是彻底不香了。
Coding is cooked only for those who don't know how to code.
编程只会对那些不会写代码的人完蛋。
>>109802725
so it still bussin?
所以其实还是挺香的？
"Coding" was always jannie-type work which resulted in 100x more profit for the company owners.
「写代码」说到底一直是版务杂工型的活，给公司老板们赚了 100 倍的利润。
>>109802262
>They can get loans forever from the money printer.
>有印钞机在，贷款能无限续。
Is this actually what you believe?
你是真信这个？
>>109803768
How do you think fiat currencies work?
你以为法定货币是怎么运作的？
>>109803779
What happens if you print a whole bunch of money for things that don't end up actually having any value?
印一大堆钱砸到最后毫无价值的东西上，会发生什么？
>>109803826
The general price level in the economy goes up. Oh, would you look at that.
全社会物价整体上涨。哟，瞧瞧现在这不就来了嘛。
>>109803841
HAHAHAHAHAHA
>>109803841
>>109798315
Bro the AI companies are so successful they’re asking to be shut down.
兄弟，AI 公司成功到主动请求被关停了。
>>109801974
>competence is bad and nepotism is a skill hmmkay
>有能力是坏事，任人唯亲是门手艺，好的呢
>meritocracy is getting incompetent retards in high positions because they bought me a beer
>所谓精英政治，就是让买了我一杯啤酒的无能智障爬上高位
The absolute state
真是绝了
>>109803826
How do you think countries are at >>>100% debt to gdp permanently and only looking to borrow even more?
你以为那些国家债务占 GDP 常年超 100%、还只想着借更多，是怎么做到的？
>>109799403
Many are too focused on software and do not understand physical constraints
很多人只盯着软件，不懂物理世界的硬约束。
>>109804961
That's all the tech bros though. Also why the aifags think they can get into biology with 0 biology knowledge and make a difference, then fail at the first contact with the wetlab and go bankrupt overnight.
技术兄弟全是这德行。所以那帮 AI 党才会以为零生物知识也能搞生物并搞出名堂，结果一进湿实验室(wetlab)就翻车，一夜破产。
>>109798315
Yes it is cooked. Pissed on, even. Do not learn.
对，凉透了。甚至被浇过尿了。别学。
>>109805010
Let me be more clear: Do not learn for money**
说得更明白点：别为了钱去学
>>109805015
Come to sweden and study for free.
来瑞典，免费上学。
>>109804871
you forgot your meds incel
你忘吃药了吧，老光棍。
>>109798315
Yes. Even the biggest copers are starting to admit it and are moving full time to entertainment.
是。连最强自我安慰选手都开始承认了，正在全职转行娱乐区。
There are still copers around, but if you look closely, they ALWAYS sell coding courses (which are now completely useless and obsolete, aside for historical interest).
自我安慰选手还有，但仔细看，他们全都在卖编程课（现在除了当历史资料外已经完全没用、彻底过时）。
tldr, coding is completely useless in 2026.
一句话总结：2026 年，编程彻底没用了。
>But muh special big boy job with muh embedded devices or other copes
>可是我的大人物工作是搞嵌入式设备啊之类的自我安慰
Your job is not special and it's probably easier to automate than webshit.
你的工作毫无特殊，而且大概比网页垃圾活还容易自动化。
>Le stocastic parrot
>所谓随机鹦鹉
A stocastic parrot that can solve Navier Stokes lmao
一只会解纳维-斯托克斯方程(Navier-Stokes)的随机鹦鹉，笑死。
>B-b-but muh taste
>可、可是我的品味啊
This is just sad at this point.
到这份上只剩悲哀了。
>>109805161
What AI psychosis does to a sub-65 IQ retard lmao
AI 妄想症把一个 IQ 不到 65 的智障祸害成这样，笑死。
>>109805199
Tell me where I'm wrong exactly then, 280IQ-sama
那请 280 IQ 大人明确告诉我我到底错哪了？
>>109805214
You couldn't possibly understand, since I can't explain below the level a typical 7 year old could understand.
你不可能懂，因为我没法把话说得比一般 7 岁小孩的理解水平更低了。
>>109805224
I accept your concession.
我接受你的认输。
>>109805231
Good. I wouldn't want my lack of ability to explain simple concepts to retards to overshadow my genius. Thanks for accepting my concession.
好极了。我可不想让「没本事给智障讲明白简单概念」这点盖过我的天才光芒。谢谢你的认输。
>>109805237
>N-n-noo you!
>不、不是的，是你！
Pathetic. This is you btw.
真丢人。顺便，这说的就是你。
>>109805289
How did you find this picture of me?
你怎么搞到我这张照片的？
>>109805289
he is me
（如果你觉得这篇文章有启发，可以点击这里付费）
