guyu retweeted
今天雪球上讨论非常高的一个问题,也是我想问的。
当你熟读巴菲特、段永平,坚信买股票就是买公司,等腾讯从约476跌到400附近,觉得好公司终于有了好价格,于是买入。八年后,股价约376 。你没有追高,只是提前八年到达了目的地。
当你看好能源革命,等宁德时代从高点跌了约30%,又看到新能源汽车销量增长、龙头份额稳固。估值已经很便宜了,未来几年增速也很确定了,股东回报也加起来了,你买入,仍然跌 30%。
当你看见阿里巴巴从约319美元跌到150美元附近,跌了一半多,又宣布扩大回购,觉得安全边际和股东回报都有了,于是买入。后来跌到约63美元,你仍然亏了58%。
当你看好人工智能,又坚决不追泡沫,等智谱从2980港元跌到1500港元附近,觉得腰斩已经释放风险,ai 是未来十年的大事,终于买入。后来跌到500港元附近,你又亏了约67%。你避开了第一场腰斩,赶上了后面的两场。
当你看好黄金,长期觉得只有货币通胀才能解决所有问题,又怕追高,于是决定等回调;觉得紫金矿业既有资源、又有现金流,比直接买黄金更划算,于是买入。它能够给你跌到 10pe 以下。
当你反复亏钱,不断思考到底错在哪里,最后无数个痛彻心扉,痛定思痛,最终决定投资美股。要去有鱼的市场。
你最终选择投资最核心的英伟达,等它从约 240 跌到 180,已经回调了 25% ,已经是明年 13 倍,14 倍 pe 了,还有 50% 以上的回购。你觉得AI需求还在、技术壁垒还在,股东回报在,市场也对,于是买入。结果。。。。
你不追高,不碰垃圾股,不听小道消息;有研究,有耐心,熟读各种投资经典,还专门等了回调。
那么,错在哪里?
guyu retweeted
Bu Çin filmi beni mahvetti bugün. Özellikle sonlara doğru birkaç kez durdurmak ve balkona çıkıp hava almak zorunda kaldım.
guyu retweeted
Elon Musk once said:
“Don't ever attach yourself to a person, a place, a company, an organization, or a project.
Attach yourself to a mission, a calling, and a purpose only.
That's how you keep your power and your peace.”
guyu retweeted
翻了一下最近两个月,我用Agent干活最爱说的几句Prompt:
1. 从第一性原理出发,找到最本质原因,修复一下。
2. 先问一些不清晰的问题,对齐我的喜好,然后再出3版方案。
3.不要默认我提出的问题定义就是正确的。你先分析我真正想解决的底层需求,判断我有没有把症状误认为原因、把手段误认为目标。如果存在更值得解决的问题,直接指出来
4. 不要只考虑这个改动产生的直接效果。继续推演它的二阶、三阶影响,尤其关注对其他模块、用户行为、系统复杂度和长期维护成本的潜在影响。
5. 其他领域里有没有类似的方案,用跨领域的方式全面思考,找到最优解。
6. 先假设你当前认为最合理的方案是错误的。站在最强反对者的角度,寻找能够推翻它的证据、反例和边界情况。然后重新评估方案,看看有没有更好的解法。
7. 深入分析整个代码库的 Git 历史、关键 PR 和架构演变过程。找出哪些设计是有意为之,哪些是历史遗留,哪些是为了解决已经不存在的问题而引入的。识别可以安全移除的历史包袱。
8. 假设明天有一个完全不了解项目、但是能力很强的工程师来接手。请评估他理解整个项目需要付出的认知成本,找出那些命名混乱、抽象过度、职责不清晰、需要大量上下文才能理解的代码,并提出简化方案。
9. 全面分析最近XX天里的所有的PR,找出最慢的那几个,看看有没有导致卡顿、可以优化的地方。
10. 全面复盘最近XX时间的 Bug、返工、回滚和重复修改。不要只统计发生了什么,要分析这些问题背后有没有共同的根因。把能够预防的共性问题沉淀成项目规则、自动化或者Skill,避免我们反复支付同样的成本。
John Malkovich made over 100 movies, including Con Air, which grossed more than $200 million. When Bernie Madoff was arrested in 2008, Malkovich lost his entire Madoff account, reportedly worth about $2.3 million. It was pretty much everything he'd ever made.
In the video below, he describes the moment he found out.
He had just come home from directing a play in Mexico City when he saw a photo of Madoff in handcuffs.
He walked in and calmly told his wife he was going around the corner to buy a pack of cigarettes. He had quit smoking.
"We have a little problem," he told her. "Bernie Madoff just got arrested."
The court-appointed trustee later approved a claim for only $670,000, reportedly the amount he had originally put in. Everything his money had supposedly earned was gone.
At first, he says, he wanted someone to blame. Then he told himself to shut up. Most people who were ever born will never have a tenth of what he lost.
At 72, he's still working, and his new film is already getting Oscar buzz.
Do you believe a Hollywood star really suffers after losing millions, or is it nothing compared to what regular people go through?
guyu retweeted
转一个老哥的发帖。
基本属实的想法:
老黄走了,新法案落地了,从下个月开始,中国彻底没有海外的卡用了。所有的国模的训练都有可能会暂停,东南亚的百万卡的中国集群要彻底被切断了,无论是H100,H200+,还B200,B300+。当年华为为了一己之私,强行阻断了本来两国都已经批准的H200进口,拍着胸脯说自己能供给上,现在彻底被卡了。等华为明年能真正开始供给的时候,美国的AI在前方看不见的地方了。
老黄真是对中国最友好的CEO,过去几年用各种方法全力游说解除对中国的管制,并且对中国厂商的各种行为睁一只眼闭一只眼。各种沸腾还骂他,现在老黄也要下台了。你要知道,过去几年支撑中国大模型发展的,包括DeepSeek,支撑他们的其实是老黄,而不是华为。要不是老黄,根本就不会有国模的进展。
现在对中国最友好的老黄要下台了,之后怎么样?
华为的昇腾,单卡算力太低,即使有超节点,也很难训练大模。而且国内HBM+,良率极低,灰色通道进口的HBM数量极少,根本没法大规模供应。东南亚集群一旦断了,下一代大模型的路就会被堵死很长一段时间。
guyu retweeted
Investors don’t realize gold ALWAYS crashes during an oil-driven crisis.
It happened in all major oil shock in the last 50 years - 1973, 1979, 1991, 2001, and 2022.
In this thread, I'll break down exactly why it happens and what institutions expect happens next:
guyu retweeted
#China’s Oct data: Industrial production 4.8% (miss), retail sales 0.4% (miss), fixed asset investment -8.1% (worse), property -21.6% terrible. Same old pattern — high-tech/export strength papers over anemic domestic demand and a deepening property slump. “Stable” on paper, unbalanced in reality. No rebalancing, just more of the same.
guyu retweeted
BREAKING: China’s central bank officially bought +26 tonnes of gold in October, its biggest monthly purchase since August 2023.
This follows +23 tonnes acquired in September, and marks its 24th consecutive month of gold purchases.
For perspective, in 2025, China’s gold purchases did not exceed +5 tonnes in any month.
Year-to-date, China has officially acquired +129 tonnes of gold, lifting its holdings to a record 2,435 tonnes.
As a result, China now holds just 2 tonnes less gold than France, the world’s 4th-largest gold holder.
China’s demand for gold shows no signs of slowing.
guyu retweeted
For Chinese policy-makers, this ought to be the most anxiety-inducing graph of the month, if not the year.
guyu retweeted
We proudly gave @OpenAI our NVIDIA DGX-1 supercomputer to work on #AI's toughest problems. nvda.ws/2aRtOhE









































