I see the same patterns being repeated across all the major subs almost verbatim these past few weeks.
过去几周,我在所有主要子版块中看到同样的模式几乎逐字重复出现。

Tons of bots using playbook scammer tactics in order to get you to buy a Claude subscription.
大量使用诈骗脚本的机器人,旨在让你购买 Claude 订阅。

Impersonation - I've been a Software dev for 15 years.. - When you look at their account they have either 0 posts made about software or it's a bought account that stopped posting for years and then suddendly started only posting about AI in a lot of subs they didn't ever post before
冒充 - 我已经是软件开发人员15年了... - 当你查看他们的账号时,要么他们根本没有关于软件的帖子,要么这是一条在多年未发帖后突然只在很多他们以前从未发帖的子版块中发帖关于 AI 的账号

Urgency - Anyone not using AI to write code has fallen behind.. - You MUST use AI to remain being a developer or artist in 2026. Very self explanatory, buy now or your gradma explodes type of subtext. This statement is NEVER backed up by any actual data. In fact little data we do have has pointed towards a slightly negative productivity output for companies who mass adopted AI tools as shown in this Harvard study .
紧迫感 - 任何不使用 AI 写代码的人都落后了… - 你必须使用 AI 才能在 2026 年继续成为开发者或艺术家。 这非常直白,买即可,否则你奶奶会爆炸之类的暗示。此说法从未得到任何实际数据支持。事实上,我们现有的少量数据表明,大规模采用 AI 工具的公司生产力略有下降,如本 哈佛研究 所示。

Emotion - I spent X years doing something and AI did it in a day.. There are no more games that are not coded with AI.. Every coder uses AI.. - Doomposting meant to mainly discourage newcomers to learn by themselves and make everything seem hopeless if you don't join in. You see this a lot in the AI industry leaders.
情绪 - 我花了 X 年做某事,AI 用一天就完成了… 现在没有不使用 AI 编码的游戏了… 每个程序员都使用 AI… 这是一种旨在劝阻新人自学、让他们觉得如果不加入就无望的恐慌帖。在 AI 行业领袖中常见。

Some other common patterns to look for:
一些其他常见的模式供参考:

  • AI has done X in Y time that took me Z time to do before
  • AI 在 Y 时间内完成了 X,而我之前需要 Z 时间

  • It's not perfect but imagine how good it will be in a month

  • 它不完美,但想象一个月后会有多好

  • EVERYONE uses AI now and people who say they don't are lying

  • 现在人人都在用 AI,说他们不使用的人是在撒谎

  • Post itself or one of the most upvoted comments in the post specifically mentions Claude

  • 帖子本身或帖子中最受赞同的评论明确提到 Claude

  • AI has ended software development. AI has solved game development. AI has solved Art. etc...

  • AI 已经结束了软件开发,AI 解决了游戏开发,AI 解决了艺术等…

  • Just like you don't understand assembly, you no longer need to understand code, Claude acts as the compiler

  • 就像你不懂汇编,你也不再需要懂代码,Claude 充当编译器

  • Developers moved to Agentic Engineers now, there's no more developers.

  • 开发者现在转为「Agentic 工程师」了,已经没有纯粹的开发者了。

  • If it makes a mistake is due to your harness or prompting skills. The model can do everything with the right harness.

  • 如果它出错是因为你的框架或提示技巧。有了正确的框架,模型可以做任何事。

There are many more, but hopefully you get how the grift works by now. They all boil down to "If you don't buy a Claude subscription, you are worthless".
还有更多,但希望你现在已经明白这类骗局的运作方式。它们都归结为 “如果不买 Claude 订阅,你就毫无价值”。

Some signs to look for:
一些需要警惕的迹象:

  • Post is never accompanied by any links that allow you to try the thing that supposedly was done in a day
  • 帖子从未附带任何链接,让你验证所谓一天内完成的内容

  • Post does have a link but the thing is barely functioning or flat out broken, and the thread is filled with "imagine how good it will be in a month if this took a day" (this is not how software works, anyone that has ever done a technical demo knows how easy it is to make it look like something works but it's light years away from the real thing)

  • 帖子如果有链接,内容可能 仅仅运行 或 完全失效,而评论区充斥着 "一个月后会很棒,如果它只用了一天"(这不是软件运作方式,任何做过技术演示的人都知道,容易让人误以为运行,但实际相差甚远)

  • Post is written by AI and OP first comment is saying "I use AI to help me with language barrier". Post history contains multiple perfectly written comments and posts prior to LLM's.

  • 帖子 是 AI 生成 的,OP 第一条评论说 "我用 AI 帮助克服语言障碍"。帖子历史中在 LLM 之前就有多条 perfectly 写作的评论和帖子。

Wild claims such as Any developer who does not use AI is lying with no data to back it up and no clarification of what that statement means. Using a local model to give you a potential bug report on a complex function is not the same as subbing to Claude 200$ plan and vibe coding everything.
如 任何不使用 AI 的开发者都在撒谎 这类夸张的说法,没有任何数据支撑,也缺乏对该说法的 解释。使用本地模型为复杂函数提供潜在的 bug 报告 并不等同 于订阅 Claude 200$ 计划并进行 vibe coding(即随意编写代码)。

  • You don't know assembly, so why do you need to know C++ if Claude is a compiler? Bad faith questions that jump of false assumptions*. Very nefarious especially for newcomers to the field who won't be able to counterpoint them.
  • 你不懂汇编,那为什么还要懂 C++,如果 Claude 是编译器? 这是一种 恶意 的提问,基于 错误* 的假设。对新人而言尤其有害,因为他们无法反驳。

Some FAQ that I know for sure the bots will flood the thread with if I don't adress it now.
如果我不现在处理,确定机器人会在帖子中灌水的常见 FAQ:

  • I am not Anti-AI. I worked in the AI field back when it was called machine learning and after I left it I worked in SaaS always with products adjacent to ML. I do think LLM's have some useful cases. I also think it's beyond overhyped and oversold by nefarious people. I am happy to go with any technical discussion in the comments if someone wants to actually have a conversation beyond hype.
  • 我 并非 反对 AI。我曾在机器学习刚兴起时在 AI 领域工作,之后转行做 SaaS,始终与 ML 相关的产品打交道。我认为 LLM 有一些有用的场景。我也认为它 被过度炒作 并 被恶劣势力售卖。如果有人想在评论中进行真正的技术讨论,我乐意参与,而不是停留在炒作上。

  • My current company leadership had a full on AI psychosis moment back in February/March and it had terrible effects on the business. AI has been pulled back massively and we have not allowed any AI written PR's since September. We do use AI has read-only (aiding with finding bugs pre-shipping, orthogonal performance changes and so on). Business has gone back up and overall mental health of people is better. Slower deployments where people know what they are shipping is not a bad thing and it will never be a bad thing, in my opinion. So I am now an example of DEV who doens't use Claude to write code, and so are my colleagues, everyone is happier and business is doing good.

  • 我现在的公司领导在 2 月/3 月期间出现了 完全的 AI 精神病 式决策,对业务产生了 严重负面影响。随后 AI 被 大幅回撤,自 9 月起 禁止任何 AI 撰写的 PR。我们仅 使用 AI 作为只读工具(如在上线前发现缺陷、性能优化等)。业务回升,员工整体心理健康改善。我认为,即使是更慢的发布,只要团队清楚自己在交付什么,也不是坏事,也永远不是坏事。因此,我是一个不使用 Claude 编写代码的开发者,我的同事也是,大家更开心,业务也好。

  • It's important to note that when I say we use AI to give bug reports and other types of static analysis, the majority of those are absolutely useless and/or nitpicks. But it does find an important one every once in a while and so far that reason that makes it worth it for the business. You do not need any of this in your hobby and you certainly do not need to pay for it, any latest local model is sufficently good to aid you on this stuff. I suspect really cheap solutions for this type of problem will be integrated into IDE's soon, you certainly don't need a generalistic LLM to do this.

  • 需要注意的是,当我说我们使用 AI 给出 bug 报告和其他静态分析时,大多数情况是 完全无用 或 琐碎的批评。但偶尔会发现一个重要的问题,这正是企业认为值得的原因。你 不需要 在爱好项目中使用这些,也不 需要付费,最新的本地模型已经足够好来帮助你。我猜测针对这类问题的低成本方案很快会集成到 IDE 中,完全不需要通用的 LLM。

  • Especially if you're a new dev, do not think that every developer is using AI. They are not. You do not need to use it unless you are being mandated to use it. You are not falling behind on anything. Claude is certainly not a compiler and all the vibe coded repos have massive amounts of problems under the surface, even the reverse engineered ones which should be much simpler. Keep doing your thing and ignore the bots.

  • 尤其是如果你是新手开发者,请不要认为 所有 开发者都在使用 AI。他们并不是。除非被强制使用,否则你不需要使用它。你并没有落后于任何事物。Claude 绝对不是编译器,所有 vibe coded 仓库在表面之下都有大量问题,即使是反向工程的仓库也应该更简单。保持做自己的事,忽略机器人。