Tor Tor

@Thomaskong96638 没有毛主席就没有中华民族 清算毛主席就是解体中华民族 --知识分子光嘴炮爽，知道自己在推啥不？ 🤣

著名经济学家茅于轼：“他（毛泽东）掌权以后，就开始杀杀杀，他要杀人立天下，我们现在还把当成国父，还把他头像挂在天安门城楼上，中华民族不清算他，就没有希望！”。 一针见血！ 向敢于直言的茅老致敬！ 老一辈经济学家的肺腑之言，值得深思… （视频来源于TikTok 若侵权联系删除）

@JasonZX 这是在干嘛？ 就是拔拉一下包裹？

人机大战结束，Figure的实习生小哥Aime获胜，在10个小时的直播里，他说他的左前臂基本上断了。 最终得分，人类在每个包裹的分拣时间要比机器人少0.04秒，估计机器人迭代一次人类就不会再赢了： 机器人F.03：12,732 个包裹（2.83 秒/包） 人类实习生Aime：12,924 个包裹（2.79 秒/包）

@populartourist Is there any physical limitations That prohibit a 1B model to beat Opus? If not, just shut up

Qwen3.7 spotted Can't wait for Qwen3.7 1B to beat Opus 4.8

@NiKiTa_32156 全天下最没用的就是知识分子 特别是有了chatgpt以后 更加暴露了纯嘴炮的百无一用 -- 这也包括只会开会的德国政客们

德國對川普近期關於台灣和中國的無恥言論進行了充分評估，並將當前局勢與烏克蘭局勢進行比較，預測這位紅頭髮的無賴會為了自身利益而將台灣拱手讓給中國，或者在極端情況下，乾脆對台灣置之不理。 烏克蘭的局面正在重演。一旦中國發動攻擊，川普很可能會更進一步，質問：“台灣，你們明明毫無勝算，為何還要發動戰爭？” ——德國專欄作家於爾根‧瑙迪特

@mardehaym AI is replacing coding as a labor job AI is making SWE job to an engineering intensive job.

I was wrong about AI replacing developers. For a year I told clients the leverage was in the tool. Better Cursor setup, better Copilot rules, more output per engineer. Two years of building production AI teams later, that take aged badly. The original story was simple. AI is an output multiplier. Adopt the tools, ship more code, win. Every vendor sold it. Every conference talk repeated it. I repeated it on sales calls. DX surveyed 121,000 developers across 450+ companies between November 2025 and February 2026. 92.6% use an AI coding assistant. AI-authored code is now 26.9% of all production code, up from 22% the prior quarter. Productivity gains still haven't moved past 10%. Digital Applied's Q1 2026 survey of 2,847 developers found something even sharper. Reviewing AI-generated code now takes 11.4 hours per week. Writing new code takes 9.8. The cost of AI is showing up where most teams aren't measuring it. In the human attention required to keep AI-generated code from breaking in production. At Limestone, 98% of our code is not handwritten. Auth, payments, and a few domain edge cases are the exception. But every line of that 98% gets reviewed, restructured, or rejected by a senior engineer who decided what should exist before the agent ever generated it. AI replaced typing. It didn't replace thinking. The engineers who thrive aren't the fastest coders. They're the ones who can read a 200-line AI-generated diff and spot the three edge cases the model missed. They're the ones who architect before they prompt. They're the ones who can tell you why a piece of code shouldn't exist before they explain how to write it. If your AI strategy assumes the agent does the thinking, you're not building an AI-augmented team. You're building a risk surface that compounds with every commit.

@chatgpt21 I still don't understand What's the work? Are they just pretending to work?

The moment the robot passed the human worker, because the human had to take a bathroom break

@Yangtze07 @RobinSeun 美国承诺 不挑战中国共产党和习主席的领导地位

@RobinSeun 中国买买买，美国做了哪些让步？

关于此次5月13-15日特朗普访华的会谈结果。 白宫在昨天（5月17日）发布了公告，可以算是迄今为止最清楚最明确的一份公告了。 里面明确提出的事项如下： 1、两国应在公平互惠的基础上构建建设性的战略稳定关系。 2、两国领导人一致认为伊朗不能拥有核武器，呼吁重新开放霍尔木兹海峡。 3、确认实现朝鲜半岛无核化的共同目标。 4、建立美中贸易委员会和美中投资委员会。 5、中国批准首批采购200架美国制造的波音飞机。 6、除了2025年10月作出的购买大豆的承诺外，中国还将在2026年（按比例计算）、2027年和2028年每年购买至少170亿美元的美国农产品。 ------ 其余未尽内容，会在文章中分析。 稍候！

@iqbal_riqbal @haider1 @ylecun If you know it can be solved with more resources, then that's an engineering problem. Scientists work on assumptions, unknowns, that maybe 99% will fail eventually, but it's the 1% success that really makes the difference to the world

@haider1 @ylecun Yann Lecun was at Meta since 2012. Why we didn’t see this world model materializing. Neither he was able to deliver concrete evidence to convince Meta to invest resources in his idea. You know why , the results were not good enough.

Yann LeCun says that within a year to 18 months, we'll have a general method for training hierarchical world models These models would learn from video and real-world data, then help plan actions in robotics, healthcare, and other areas "then scale them toward a universal world model"

@elonmusk Who in their right mind would allow email autoresponder? You are literally useless at this point

Grok upgrades

@SydneyDaddy1 @grok 奶爸对LLM认识不深 对它的一致性抱有不切实际的期望 -- 基本上它们被训练的时候，就是一个见人说人话，见鬼说鬼话的高级模式匹配机器 -- 它们没有世界观，没有核心价值，匹配的是人类语言的平均水平，而人类的平均思维水平实在称不上高明。

根据下面的使用反馈实测了一下，果然没错 问豆包是否可以徒手打磨出5纳米的表面精度，中国工程师就可能，美国工程师就完全不可能 我用deepseek也测了一下，deepseek求生欲似乎更强烈🤣 @grok 你来评论一下这种现象

@ThrillaRilla369 I don't have anyone can be called as friend from work -- if I don't need the money, very minimum motivation left to go to work everyday😒

This is the reality of friendships at work Anyone else relate?

@bishara Was that a VP level role?

a friend of mine Tesla got him doing 10 interviews including; >screening call >1 hour interview with manager >1 hour interview with peer >1 hour interview with sr. peer >take home 24 hour test >6 hour interviews back to back >presentation assignment >1 hour interview with manager (again) >1 hour post interviews >onsite tour + presentation all day 2 full months of interview steps just to tell him you’re not the right skill set 😭

@difficultyang Who said C++ has ownership rule?

Do LLMs truly understand C++ ownership rules or have they simply memorized enough examples to present a simulacra of understanding

@__Inty__ 摸鱼吧 干啥了？

美国Figure公司的机器人连续上班已经进入第 4 天。😂

@haider1 I think his "world model" point applies to llm too -- without a world model, a llm can be easily manipulated by user prompt, context. World model is the cornerstone of true intelligence, that knows when to fight back garbage input.

Yann LeCun says LLMs are strongest in domains where language itself is the substrate of reasoning, like math and code They can solve problems, prove theorems, and write programs — but they are not creative mathematicians, software architects, or computer scientists "their role is to help humans build"

@karthikponna19 Die, then turn to dirt.

TELL ME ONE THING YOU CAN DO THAT AI CANNOT

@satireorcry1 法国总统不是一样跪舔 那个啥嫌啥黑 🤣

法国高人气时事综艺《Quotidien》昨晚节目超精彩！ 记者Paul Gasnier直接点破：美国国务卿“卢比奥”为何能「绕过」中国制裁踏上中国土地，以及他当年被制裁的来龙去脉。 最好笑的是，节目直接吐槽中共为了「不丢脸（perdre la face）」硬搞变通，结果反而更丢脸🤣 尾段政治评论员Jean还突然发出一声超嫌弃的冷笑，接着脱口：「这就是中国人！」节目效果直接拉满😂

I made something like 40k loc to product level code base at one of the big tech. I think the key that it can converge is design first, test case first, only merge when you have concrete test coverage The other secret is, the 40k loc is split to like 100 CRs, 100 randomly assigned readability reviewer reviewed it 😂😂😂😂

Am I the only one getting vibe coding fatigue? Building landing pages in 30 seconds was fun, but maintaining a complex codebase where half the logic was “vibed” into existence is an absolute headache. Feels like we traded 1 hour of typing for 5 hours of architectural debugging later. I’ve started manually writing core logic again so I actually know where the technical debt is hiding. Is anyone successfully managing large production projects with AI agents, or are we all just building disposable software?

@GuangruiH @V_brigade_Eric 霸权争议在太平洋上就解决了 打陆战都不是选项 这就是习大大想要建设强大海军 和美帝平分太平洋的高屋建瓴思想的起源

@V_brigade_Eric 核大国之间谈不上轰炸首都的选项

川普访华这事，其实无趣得很，就像围观谁谁又落马了一样无聊，还没马斯克在李老师帖下炫儿子学中文魔幻。 但有的东西，还是很明显：川普要软化对华强硬路线了，或者说，要保持一点弹性。 为什么？因为美伊冲突。 中美目前整体基调是对峙，美方更是强硬派占上风，这次川普带这么多公司CEO去，在没那么融洽的氛围下，硬是达成200架客机等交易，为何？当政绩吹？单薄了点。除了国事访问签点商单的例行公事，还有一点就是利用这些商业大佬，对冲和平衡国会包括白宫内的对华鹰派。 川普是现实主义者，虽然他打着保守派的旗号，实际上，是怎么有用有利就怎么来，意识形态和价值观只是其对外的工具，这从川普只喜欢跟赢家打交道而毫不在意对方政治光谱就能看出来。哪怕是被他骂的狗血淋头的“觉醒”派，只要上台了，愿意给川普面子，川普一样会热情似火，更别提什么独裁者了。 所以，川普其实并不喜欢国内对华鹰派一边倒，然后把他推着走，让他缺少选择和弹性。 这次川普访华，之前一切还很不明晰，但美伊冲突后就突然提速和确认了。不得不说，这场冲突，直接影响了中美关系走向。 因为这场冲突，出乎川普意料。 携活捉马杜罗的余威，川普本以为一通快刀斩乱麻的斩首，能令伊朗迅速屈服，结果却事与愿违，大大出乎了川普意料。 而川普，极讨厌长期战事，因为这几乎无一例外会成为主事人的政治噩梦。 现实是残酷的，高科技速胜越来越像时代的泡沫，即便是伊拉克战争看起来摧枯拉朽快速通关，后面残酷的治安战却依旧打了多年，更别提阿富汗那伤心地。 现在这时代，别说是两个国家之间的全面战争了，就算是强国打一个穿拖鞋的武装组织，动辄都会旷日持久相持多年。美伊冲突只是把科技速胜的泡沫戳破了，把这个现实摆在了川普面前。 虽然美军地球最强，却也没法用一场场的速胜带给川普政绩，结果必然就是，川普要从极端强硬路线回撤了。 假如之前川普真有在台海冲突中轰炸北京的想法，现在恐怕也烟消云散了。对他来说，做做交易，稳住北京，去风险又加分，何乐而不为？即便这次访华，中方规格比上次微降，川普赞扬之声却比上次更盛，原因正在于此。 对于反贼来说，这一点其实更值得深思和警醒： 美国的灯塔之光总是或明或暗，纵有千好万好，船总要自己开。 不要像阿富汗政府那样依赖美国，要像乌克兰那样自我成长。

@Suyutong 美帝看习大大搞香港 那相当于看下棋的对手自己给自己断一手气 帮忙都来不及 还主动帮你解套 想啥呢

特朗普告訴記者：習近平稱他正非常認真的考慮釋放金明日牧師。 但習稱：黎智英案很棘手。習真無恥，中共和港府構陷黎先生，自己還要稱棘手。 立即無條件釋放黎智英先生及所有良心犯！