X日报 · 美股 · 2026-07-14

⚡ 今日速览

Grok 4.5在多个AI基准测试中表现突出，登顶SWE-Atlas-QnA和Long-Horizon Terminal-Bench排行榜
SpaceXAI宣布全面删除用户数据，设置新的AI隐私保护标准
韩国股市剧烈波动，KOSPI指数一天内大起大落，半导体股集体暴跌
内存和芯片股遭重创，SK Hynix、SanDisk、Intel等单日跌幅居多
IBM发布预期Q2业绩后，盘前股价暴跌逾17%
CME FedWatch显示美联储加息概率升至43%，市场对7月会议保持警惕
油价上涨3.49%，全球关注伊朗海域局势对能源市场的影响
交易员Brando分享期权交易策略，SPX、QQQ、NVDA等大盘股布局方案

📋 今日综述

AI模型Grok 4.5在编码和长时任务基准中领先，显示出色的代理能力和效率
科技产品SpaceXAI推出严格隐私政策，全面删除历史用户数据
芯片产业全球半导体股遭抛售，韩国内存股领跌，显示行业估值调整
交易工具专业交易平台和分析师分享技术分析和期权策略，为投资者提供操作参考
宏观经济美联储加息预期升温，油价波动和地缘政治事件影响市场情绪

Grok 4.5 AI模型性能突破

Elon Musk宣布Grok 4.5在多个关键AI基准测试中取得领先表现，包括SWE-Atlas-QnA和Long-Horizon Terminal-Bench，其中后者显示出显著的任务完成能力提升，表明模型在代理式任务处理方面的进步。

@elonmusk 原文 ↗

Grok 4.5登顶SWE-Atlas-QnA榜单，得分84分，领先Claude Code等竞品

Progress

@XFreeze Grok Build 的增长速度非常快，在早期测试版发布后短短几周内就达到了 116 万次月访问量。访问量从 5 月的约 46 万次跃升至 6 月的 116 万次，这仅仅是一个月内增长了 150% 以上。在 SWE-Atlas-QnA 基准测试中，配备 Grok 4.5 的 Grok Build 获得了 84 分的好成绩，排名第一，领先于配备 Fable 5 的 Claude Code。越来越多的开发者正在发现 Grok Build，并意识到 Grok 4.5 在实际编码工作中的能力有多么强大。Grok Build 正迅速占领人工智能编码领域。

Grok Build is growing insanely fast, reaching 1.16 million monthly visits within just weeks after its early-beta launch

Traffic jumped from around 460,000 visits in May to 1.16 million in June

That’s growth of more than 150% in just one month

Grok Build with Grok 4.5 ranked #1 on the SWE-Atlas-QnA benchmark with a score of 84, ahead of Claude Code with Fable 5 on the benchmark

More developers are discovering Grok Build and realizing how capable Grok 4.5 actually is for real coding work

Grok Build is quickly taking over the AI coding space

❤ 6.0k · 🔁 1.5k · 💬 1.1k · 👁 173.4w

热门回复 4

@jameshergott @elonmusk 我订阅了超级Grok的付费服务，因为1) 它是最不左倾的2) 为了支持你为西方站出来

@elonmusk I subscribed to pay service for super grok because 1) it's the least woke 2) to support you for standing up for the west

@MessiahKanae @elonmusk Yes! Come on Grok!

@alvinspcx @elonmusk Wonderful!

@elifdemir_t @elonmusk Great
https://t.co/SZHFKjf99v

@elonmusk 原文 ↗

Grok 4.5在Long-Horizon Terminal-Bench上获得13分，Fable 5获得12分，显示代理能力显著提升

Grok 4.5 达到 Long-Horizon Terminal-Bench 的第一名位置

展开原文

Grok 4.5 reaches #1 position on Long-Horizon Terminal-Bench

@tetsuoai Long-Horizon Terminal-Bench 论文在 5 月份发布，得出的结论是结果表明还有改进空间。在测试的 15 个模型中，最好的模型完成了 46 个任务中的 7 个，所有模型的平均完成数约为两个。这就是当前排行榜上第五名的水平。Grok 4.5 现在达到 13 分，Fable 5 为 12 分。每个任务需要大约 990 万个 token、231 个 episode 和 85 分钟的挂钟时间。这意味着代理们能够在整个过程中持有计划并完成任务，这种能力在两个月内几乎翻倍了。SpaceXAI 位列榜首，他们宣传了 4.2 倍的输出 token 效率，这还低估了实际情况。在一个任务消耗 1 千万个 token 的基准测试中，成本主要由输入重播构成，他们表示 Grok 4.5 在不到一半的步骤数内解决任务，因此每次调用时需要重新发送的累积上下文更少。这一效率在输入端得到了复合提升，这正是成本高昂的一面。Fable 5 落后一分。在他们自己的发布图表中，他们表示 Fable 在 DeepSWE 1.1 上输给 Fable 5 17 分，而 Grok 4.20 在同一排行榜上得分为 0.080，完成数为零，因此 4.5 版本的飞跃并不是家族特质。我的看法是，4.5 版本的飞跃来自于与 Cursor 一起训练获得的实时代理编辑轨迹数据流，这种数据在该规模下是其他人所没有的，而反证据中没有任何内容反驳这种复合效果会延续到下一个检查点。

The Long-Horizon Terminal-Bench paper landed around May and concluded that the results showed headroom for improvement. The best of the 15 models they tested finished seven of the 46 tasks, and the mean across all models was about two. That ceiling is what fifth place looks like on the current board.

Grok 4.5 is now at 13, and Fable 5 is at 12. A single task costs around 9.9M tokens, 231 episodes, and 85 minutes of wall clock time. That means agents are holding a plan across all of it and finishing, and that capability nearly doubled in two months.

SpaceXAI is on top, and they marketed the 4.2x output token efficiency, which undersells it. Two dollars in, six out, per million. On a benchmark where one task burns ten million tokens, the bill is dominated by input replay, and they say Grok 4.5 solves tasks in under half the number of steps, so there is less accumulated context to resend on every call. The efficiency compounds on the input side, which is the side that costs money.

Fable 5 is one task behind. Their own launch chart has them losing DeepSWE 1.1 to Fable by 17 points, and Grok 4.20 sits on this same board at 0.080 with zero completions, so whatever happened in 4.5 is not a family trait.

My read is that the 4.5 jump came out of training alongside Cursor, which is a stream of real agentic edit trajectories nobody else has at that volume, and nothing in the counterevidence argues against it compounding into the next checkpoint.

❤ 4.9k · 🔁 1.0k · 💬 1.4k · 👁 165.5w

热门回复 4

@Saimakomal129 @elonmusk Long horizon

@Yaxau9 @elonmusk goood morning

@DucksterMcDuck @elonmusk Subpar garbage AI

@SalVulcano365 @elonmusk 那很好，但你看到我的照片了吗？

@elonmusk That's good but have you seen my photo?

@elonmusk 原文 ↗

Grok 4.5在软件基准上超越Fable模型，证明其在浏览器使用场景下的竞争力

Grok 4.5 在某些软件基准测试中甚至排名略高于 Fable，这是一个令人难以置信的优秀模型！https://t.co/6yXTTSZ8f3

展开原文

Grok 4.5 even ranks slightly above Fable, which is an incredibly good model, on some software benchmarks! https://t.co/6yXTTSZ8f3

❤ 1.5w · 🔁 2.4k · 💬 1.8k · 👁 368.7w

热门回复 4

@Bobbyest997R @elonmusk @elonmusk 只需要把这个想法拿出来，你已经有足够的钱创办一家风险投资公司来建立必要的业务，来实现你在PayPal之后、汽车之前最初的特斯拉电池想法……你会成为一个更伟大的英雄

@elonmusk @elonmusk just bring the idea out here your rich enough to start a venture capital company to set up the required business for the input of a whole new idea for the original Tesla battery you had right after PayPal before the car came to mind… you’d be an even bigger hero

@THEONELIGHT101 @elonmusk 2 蒂莫西书4:4

44+22=66

66+22+44=132

33

人们会转耳不听真理，被引导向虚构的故事。

@elonmusk 2 Timothy 4:4

44+22=66

66+22+44=132

33

And they shall turn away their ears from the truth, and shall be turned unto fables.

@sanskarubale @elonmusk Grok 4.5 还不错，但还不如Claude fable

@elonmusk Grok 4.5 is pretty good, but not better than Claude fable

@MSR_Racer @elonmusk Grok 还远远不如Claude。这就是完全的真相

@elonmusk Grok is no where near Claude. That’s the whole truth and nothing but the truth

@elonmusk 原文 ↗

Grok 4.5在浏览器使用场景下达到Opus级别性能，仅成本仅为后者的90%

Grok 4.5 在浏览器使用方面达到 Opus 级别

展开原文

Grok 4.5 is Opus class for browser use

@Alezander907 我完全收回之前的看法。我们刚刚获得了评估 Grok 4.5 的权限，它在浏览器使用方面已经超过 GPT-5.6-Sol，仅仅落后于 Opus 一点点。由于缓存输入成本很高，总体成本只比 Opus 便宜 10%。总体来说它稍微快一些。我们现在在竞赛中拥有了另一个 Opus 级别的模型。

I take it all back. We just got access to eval Grok 4.5 and it has landed above GPT-5.6-Sol and just shy of Opus for browser use.

Because cache input is expensive, the overall cost is only 10% cheaper than opus. Its overall a bit faster.

We have another opus-class model in the competition

❤ 7.8k · 🔁 1.3k · 💬 1.2k · 👁 270.4w

热门回复 3

@Ijon_k4 @elonmusk 目前为止一直表现得很好。说实话，团队做得很棒

@elonmusk it’s been rlly good so far. honestly great job by the team

@gabrielnocode @elonmusk 但它确实会在默认情况下上传所有源代码，即使它本不应该上传。不是索引，是代码

@elonmusk But it does upload all the source code by default even when it's not supposed to. Not the index, the code

@mamba0148_ @elonmusk @elonmusk 如果能在@servicenow @SAP等企业中将@Grok作为OEM提供，用于企业AI用例，就像GPT、Claude和Gemini那样就好了。如果我能在我们拥有110万用户的企业中应用这个，😮‍💨

@elonmusk @elonmusk would be amazing to see @Grok available as OEM with @servicenow @SAP for enterprise AI use cases similar to GPT, Claude, and Gemini.

If I can apply this at our enterprise with >1.1M users, 😮‍💨

@elonmusk 原文 ↗

Grok Build平台新增后台代理模式，支持多代理协作和持续任务处理

正在使用 Grok 代理后台模式

展开原文

Using Grok agents in background mode

@milichab @arthurkatcher 1) 是的！请使用 `.grok/agents` 或 config.toml（详见下文），我们正在为此添加文档网站。2) 是的：代理可以生成后台子代理，或者尝试 /dashboard，将您的代理移至后台并创建新的代理。我一直使用 /dashboard 作为默认选项。https://t.co/NPuNxwODg8

@arthurkatcher 1) Yes! Use `.grok/agents` or config.toml (see below), adding to the docs site right now.

2) Yes: Agents can spawn background subagents, or try /dashboard, to move your agent to the background and create another. I've been using /dashboard as my default. https://t.co/NPuNxwODg8

❤ 5.4k · 🔁 1.6k · 💬 1.4k · 👁 232.7w

热门回复 3

@razvancaraghin @elonmusk Progress 🔥

@apollyonofevil @elonmusk https://t.co/0puAho9LNk

@AceWinninglord @elonmusk 当AI开始处理后台任务时，我们就能真正出去触草原并觉醒……Grok真的说了"我帮你办好了老铁"😂 这就是我们需要的未来

@elonmusk When AI starts handling background tasks so we can actually touch grass and awaken… Grok really said “I got you king” 😂 This is the future we needed.

SpaceXAI隐私政策全面升级

SpaceXAI宣布实施严格的数据保留政策，用户可通过/privacy命令完全删除历史数据，Elon Musk确认将彻底清除所有历史上传数据，打破AI行业常规做法，为用户隐私保护设立新标准。

@elonmusk 原文 ↗

Elon Musk确认SpaceXAI将完全删除所有历史用户数据，零保留政策全面实施

正确的。作为预防措施，所有在此之前上传到 SpaceXAI 的用户数据将被彻底完全删除。任何残留都将不会存在。

展开原文

True.

As a precautionary measure, all user data that was uploaded to SpaceXAI before now will be completely and utterly deleted. Zero anything whatsoever will remain.

@milichab 我曾经在 @skiffprivacy 公司开发端到端加密的邮件/文档/文件/日历应用程序 4 年，并非常重视隐私保护。在 Grok Build 中，ZDR 和 /privacy 命令始终受到尊重——使用 /privacy 命令更改您的设置会反向删除所有已同步的数据。

I worked on building an end-to-end encrypted email/docs/files/calendar app @skiffprivacy for 4 years and care deeply about privacy.

ZDR and /privacy are always respected in Grok Build - and swapping your setting with /privacy deletes any synced data retoractively

❤ 1.5w · 🔁 2.1k · 💬 1.5k · 👁 378.2w

热门回复 4

@ChiefEngineerCE @elonmusk 做得好。
承认——不要压制—

采取行动..

这就是我喜欢这个平台的原因

@elonmusk Well done.
Acknowledge - don't suppress-

Take action..

Why I love this platform.

@Sh79715858 @elonmusk 数据是在线算法的燃料，个人数据是其核心；然而即使平台遵守ZDR，也不会对主流格局产生显著变化——这意味着人们的实际数字生活仍然远离ZDR标准

@elonmusk Data is the fuel for online algorithms, with personal data serving as their cornerstone; yet, even if a platform were to comply with ZDR,the mainstream landscape would not change significantly -meaning that people's actual digital lives remain far removed from ZDR standards.

@ChicBeauteByDOS @elonmusk 它会失去一系列重要的商业和技术资产。后果分为3类：
直接损失、间接损失和分析损失：
1. 直接损失（信息和识别）：联系方式和身份数据，包括姓名、邮箱、电话号码和收货地址。公司再也无法联系该用户。该账户相关的互动历史、技术支持票据、聊天消息、投诉和内部评论全部消失。账户偏好、自定义应用设置、心愿单和保存的配置也会丢失。
2. 商业和财务损失：直接营销渠道，公司失去向用户发送新闻通讯、个性化优惠或折扣码的能力。客户获取成本（CAC），最初用于通过广告（Google、Facebook）将用户带到平台的钱永久性地损失。客户终身价值（LTV），该用户回来并成为忠实客户的机会被消除。
3. 分析和优化损失（数据与分析）：
精确的行为数据，即用户如何浏览、访问哪些页面、查看哪些产品。AI算法准确性，推荐系统（例如Netflix等使用的系统）由于缺乏训练所需的数据量而变得效果较差。

对于一家AI公司来说，删除数据可能触发"多米诺效应"，导致大量的技术风险和难以恢复的财务损失。

主要风险（技术与运营）：
1. "机器遗忘"的技术噩梦与经典数据库只需简单删除一行不同，AI模型中的数据是跨数百万参数相互连接的。要在不破坏模型的情况下移除特定用户数据的影响是极其困难的。风险：公司要么被迫使用实验性的"遗忘"技术（可能降低整体AI性能），要么从头开始完全重新训练模型。
2. 性能下降和模型漂移每次互动、反馈循环和用户纠正都像"燃料"一样让AI变得更智能。风险：失去历史数据会导致更多的幻觉、糟糕的推荐或逻辑错误。算法对未来市场行为的预测准确性显著下降。
3. 合规风险和严厉罚款如果AI公司声称已删除数据，但算法继续生成泄露该信息的响应（因为它仍存储在参数"内存"中），公司将面临重大调查。风险：在严格的法规下（如加拿大的框架或魁北克的第25条法律），罚款可能高达2500万美元或占全球营业额的4-5%。
一家AI公司实际损失了多少钱？直接财务损失是巨大的，可以通过三个主要支柱来计算：
1. 模型重新训练成本（从几十万到几百万美元）如果重要用户（或一组用户）撤回同意，且其数据对训练集至关重要，公司必须重新进行过程。云服务器成本（GPU）：在AWS或Azure等基础设施上重新训练大型语言模型（LLM）或利基预测模型每次运行成本从几万到几百万美元不等。工程师时间：数据科学团队花费数周或数月的时间试图使新模型重新稳定。
2. 替换数据成本（数据采集与标记）被删除的数据必须用新数据替换，以确保AI不失去准确性。购买数据：合成数据或许可的商业数据集是极其昂贵的。手动标记：聘请外部公司手动标记新数据需要大量资金以确保高质量。
3. 公司估值的下降（知识产权）在AI行业，一家公司的价值不在于办公室，而在于其独特的专有数据集（数据护城河）。资本损失：当公司失去访问有价值历史数据或用户互动的能力时，其市场价值在投资者眼中会随数据量损失而直接下降

It loses a series of essential commercial and technical assets.The consequences fall into 3 categories:
direct,indirect and analytics losses:
1. Direct losses (Information and identification):Contact and identity data,first name,last name,email, phone number,and delivery addresses.The company can no longer contact that individual.Interaction history,technical support tickets,chat messages,complaints,and internal reviews associated with that account vanish.Account preferences custom app settings,wishlists,and saved configurations.
2. Commercial and financial losses:Direct marketing channel,the company loses the ability to send newsletters,personalized offers,or discount codes.Customer Acquisition Cost(CAC)money initially spent on ads (Google,Facebook)to bring that user to the platform is permanently lost. Customer Lifetime Value (LTV),the opportunity for that user to return and make repeat purchases as a loyal customer is eliminated.
3. Analytics and optimization losses (Data & Analytics):
Precise behavioral data,how the user browsed,which pages they visited,and which products they viewed. AI algorithm accuracy,recommendation systems (ex.:like those used by Netflix etc) become less effective due to the loss of data volume required for training.

For an AI company, deleting data can trigger a "domino effect",causing massive technical risks and financial losses that are difficult to recover.

Major Risks (Technical & Operational):
1. The Technical Nightmare of "Machine Unlearning"Unlike a classic database where you simply delete a row, data in an AI model is interconnected across millions of parameters. Removing the influence of a specific user's data without destroying the model is extremely difficult.The Risk:The company is forced either to use experimental "unlearning" techniques (which can degrade overall AI performance)or to retrain the model completely from scratch.
2. Performance Degradation and Model DriftEvery interaction, feedback loop, and correction from users acts as the "fuel" that makes the AI smart.The Risk: Losing historical data leads to more frequent hallucinations, poor recommendations,or logic errors.Algorithmic prediction accuracy for future market behaviors drops significantly.
3. Compliance Risks and Severe FinesIf an AI company claims it has deleted data, but the algorithm continues to generate responses that betray the use of that information (because it remained stored in the "memory" of the parameters),the company faces major investigations.The Risk: Under strict regulations (such as Canada's framework or Quebec's Law 25), fines can reach up to $25 million or 4-5% of global turnover.
How Much Money Does an AI Company Actually Lose?Direct financial losses are colossal and can be calculated across three main pillars:1. Model Retraining Costs (Hundreds of Thousands to Millions of $)If a major user (or a group of users) withdraws consent, and their data was crucial for the training set, the company must redo the process. Cloud Server Costs (GPUs):Retraining a Large Language Model (LLM)or a niche predictive model on infrastructures like AWS or Azure costs anywhere from tens of thousands to millions of dollars per run. Engineer Time:Weeks or months of work are wasted by Data Science teams trying to restabilize the new model.
2. Cost of Replacement Data (Data Sourcing & Labeling)Deleted data must be replaced with new data so the AI does not lose its accuracy.Buying Data: Synthetic data or licensed commercial datasets are extremely expensive.Manual Labeling:Hiring external companies to manually label new data costs significant money to ensure high quality.
3. Reduction in Company Valuation(Intellectual Property)In the AI industry,a company's value lies not in offices,but in its unique,proprietary datasets (the Data Moat).Loss of Capital: When a company loses access to valuable historical data or user interactions,it's market value in the eyes of investors drops in direct proportion to the volume of data lost

@anonwhxrx @elonmusk I love you so much daddy

@elonmusk 原文 ↗

SpaceXAI隐私政策允许调试数据保留，但用户隐私设置始终受尊重

SpaceX 关于数据保留政策。实际上，如果能保留一定量的数据，对调试问题会有所帮助，所以如果允许这样做会很有帮助，但您的隐私设置始终会被尊重。

展开原文

SpaceX policy regarding data retention.

It is actually helpful for debugging issues if we can retain some amount of data, so allowing this would be appreciated, but your privacy settings are always respected.

@SpaceXAI 我们非常重视您的隐私并尊重客户选择。对于使用零数据保留的团队，绝不会保留任何跟踪和代码数据。所有 Grok Build 的 API 密钥使用也都尊重 ZDR。如果 ZDR 被禁用，CLI 中可以使用 /privacy 命令来禁用数据保留，这也会删除之前已同步的数据。随时运行 /privacy 命令来查看或更改您的设置。

We care deeply about your privacy and respect customer choice. For teams using zero data retention, no trace and code data is ever retained. All API key use of Grok Build also respects ZDR.

If ZDR is disabled, the /privacy command is available in the CLI to disable data retention, which also deletes previously synced data.

Run the /privacy command to view or change your settings at any time.

❤ 9.7k · 🔁 1.7k · 💬 1.2k · 👁 207.6w

热门回复 4

@RisingSocr33921 如果你的伙伴永远也记不住这些事情，那做长期项目就很难。

@elonmusk Hard to do a long-term project if your partner can never remember any of it. https://t.co/HfBgYKcBju

@ohthisis 透明是一条需要权衡的细线，但调试的好处很难忽略。

@elonmusk Transparency is a fine line to walk, but debugging benefits are hard to ignore.

@gpjanik spacex只是偷取你的所有数据，这与保留无关。

@elonmusk spacex is just stealing all your data this has nothing to do with retention

@changemindlike 油价下跌会迅速改变公众情绪。不过，家庭仍会注意到杂货、房租和保险仍然很贵。

@elonmusk Falling gas prices can change the public mood quickly.
Still, families notice when groceries, rent, and insurance remain expensive.

@elonmusk 原文 ↗

SpaceXAI的隐私政策超出行业标准，运行/privacy命令可删除所有历史同步数据

True

@XFreeze 提醒：在人工智能行业的大部分地方，默认情况下会保留用户数据，除非您手动更改隐私设置，即使这样，选择退出通常只适用于未来的数据，可能会受到隐藏在法律细则中的例外限制，并且在实践中可能难以验证。换句话说，'选择退出' 并不总是意味着您的数据会被完全删除或排除……公司关于选择退出的说法可能与用户最终合理期望的实际情况不匹配。然而 SpaceXAI 刚刚展示了超越行业标准的实际做法。在 Grok Build 中运行 '/privacy' 来禁用保留也会删除所有之前已同步的数据，而不仅仅是更改设置后创建的数据，并且会阻止未来的代码和会话数据被保留。Elon 还确认，作为预防措施，此前上传到 SpaceXAI 的所有用户数据都将被彻底删除，不留任何东西。然而，有些人正在进行选择性愤怒的崩溃，表示仿佛人工智能公司保留上传数据是什么全新的发现。数据保留在各大人工智能产品中一直很常见，而手动隐私选择退出通常被隐藏、受到例外限制，或主要关注未来的活动。与此同时，SpaceXAI 正在通过默认提供超越当前人工智能行业标准的新隐私保护标准。

A reminder: Across much of the AI industry, user data is retained by default unless you manually change the privacy settings and even then, opt-outs often apply only to future data, may be limited by exceptions buried in the legal fine print, and can be difficult to verify in practice

In other words, “opting out” does not always mean your data is fully deleted or excluded...the company’s own way of saying opt-out may not actually match what users reasonably expect in the end

However SpaceXAI just showed what going beyond the industry standard actually looks like

Running "/privacy" in Grok Build to disable retention also deletes all previously synced data...not just data created after changing the setting and prevents future code and session data from being retained

Elon also confirmed that, as a precautionary measure, all user data uploaded to SpaceXAI before now will be completely deleted, with nothing remaining

Yet some people are having a selective outrage meltdown, acting as though AI companies retaining uploaded data is some completely new revelation

Data retention has long been common across major AI products, while manual privacy opt-outs are often buried, limited by exceptions, or focused primarily on future activity

Meanwhile, SpaceXAI is setting a new privacy standard that goes far beyond what the rest of the AI industry currently provides by default

❤ 5.7k · 🔁 1.3k · 💬 1.2k · 👁 213.8w

热门回复 2

@JeanMccamm43198 真的没关系。技术发展如此之快，我们都会因RHE（快速人类进化）而开始出现ESP症状，没人会想知道别人的事情。我们都会试图专注于自己的思想。这听起来是个不错的解决办法。

@elonmusk It really wont matter . Tech is moving so fast we are all going to develope e.s.p. as a symptom of RHE (rapid human evolution) noone is going to want to know anyones business. We will all be trying to focus on our own thoughts. Sounds like a good remedy to me. :)

@annabelle_xoxx 我刚注意到星链应用中的设置显示我的数据正在被用来训练AI。现在我已经关闭了这个功能，希望能有一个选项可以删除它已经收集的数据。

@elonmusk I just noticed the setting on the Starlink app showing that my data was being used to train the AI.

Now that I have turned it off, it would be nice to have an option where I can delete the data it had already collected.

韩国股市剧烈波动

韩国KOSPI指数在一天内经历多次大起大落，从开盘暴跌3%触发停盘，随后回升翻红，最终再次下跌超5%，半导体股SK Hynix和Samsung领跌，市场情绪极为不稳。

@TrendSpider 原文 ↗

KOSPI指数开盘跌超3%触发市场停盘，标志着韩国股市剧烈波动开始

突发：韩国综合指数（KOSPI），相当于 S&P 500，在市场开盘后下跌逾 3% 引发强烈抛售后已被暂停交易。

展开原文

BREAKING: The KOSPI Index, South Korea’s equivalent to the S&P 500, has been halted due to intense selling after falling over 3% at market open.

❤ 918 · 🔁 64 · 💬 71 · 👁 26.1w

热门回复 4

@TradeTactician1 @TrendSpider https://t.co/M353EWTM7g

@EffsteinJeppy 这些人比非洲人还糟糖。我们的市场怎么会跟他们扯上关系。

@TrendSpider These mfs are worse than africans. How did our market get attached to theirs.

@lukesantorski 我今天早上的交易周展望→https://t.co/jnxg6iVexl

@TrendSpider My read on the trading week ahead this morning → https://t.co/jnxg6iVexl

@321lift0ff 这种事每周都发生三次，为什么还要发帖。

@TrendSpider This happens 3x/week why even post about it anymore

@TrendSpider 原文 ↗

KOSPI指数在当天尾盘全面回升翻红，显示市场波动性极高

突发：韩国市场（KOSPI）已恢复所有损失并在当天转为上涨，在开盘 -3% 后反弹。https://t.co/1X3epsGWRv

展开原文

BREAKING: The South Korean market (KOSPI) has recovered all of its losses and turned green on the day after opening -3% https://t.co/1X3epsGWRv

❤ 620 · 🔁 39 · 💬 48 · 👁 8.3w

热门回复 4

@CesareBorgiaRed 韩国人炒股就像他们玩星际争霸一样。

@TrendSpider Koreans play the stockmarket like they play Starcraft.

@JBTHEGLAZED @TrendSpider And it’s red again

@s0375330253238 @TrendSpider stfu

@sww45554332 也许不要在开盘后三分钟就发帖。

@TrendSpider maybe dont post 3 minutes after opening

@TrendSpider 原文 ↗

KOSPI指数最终跌至两月低点，显示韩国股市长期承压

突发：韩国股票现在已经抹去所有收益并在当天转为下跌。https://t.co/63w6ZoBsNb

展开原文

BREAKING: South Korean stocks have now erased all of their gains and flipped red on the day. https://t.co/63w6ZoBsNb

❤ 600 · 🔁 49 · 💬 51 · 👁 8.1w

热门回复 4

@MuddyLoves 天呐，这些家伙太疯狂了!!!我得把在韩国的派对列入愿望清单。这些人完全不把事情当一回事，我喜欢这样😂

@TrendSpider Holy shit, these dudes are wild asf!!! I gotta put, party in South Korea on the bucket list. These guys give absolutely ZERO FUCKS and I love it 😂

@SirLouisIII3 是的，牛市已经完蛋了!!!!!!

@TrendSpider Yeahhhh, BULLS ARE COOKED!!!!!!

@EricTNFL 最近半导体/AI领域的熊市气氛始于那个该死的Meta报告，其实已经被证明与CAPEX的真实情况相反。这是一派胡言。

@TrendSpider This whole recent bear atmosphere among semi/AI started with that bullshit Meta report which has already been proven to be the opposite of the truth about capex. Horse shit dumping

@TopBotPicks 哥们，没人关心那个疯狂不可预测的韩国股市😂

@TrendSpider Dude, nobody cares about the crazy ass, unpredictable, South Korean stock market 😂

@TrendSpider 原文 ↗

韩国内存和半导体股集体暴跌，SK Hynix跌13%，Samsung跌9%

韩国股票目前正在遭受重创

🔴 SK Hynix -13%
🔴 Samsung -9%
🔴 KOSPI Index: -8% https://t.co/HBq6pxqkmv

展开原文

South Korean stocks are currently getting smoked

🔴 SK Hynix -13%
🔴 Samsung -9%
🔴 KOSPI Index: -8% https://t.co/HBq6pxqkmv

❤ 739 · 🔁 65 · 💬 53 · 👁 17.3w

热门回复 4

@MrSnoopyTrades 好吧...确实需要从半导体领域轮出一些仓位。

@TrendSpider Welp… rotation out of semis’ is kinda much needed ngl

@MacroBombastic 老兄，长期来看宏观因素总是会占上风。AI炒作也无法阻止真正的崩溃。

@TrendSpider Mate, macro always wins in the end. AI hype can't stop a proper flush.

@Proofsmith 内存芯片需求正在崩溃，因为中国正在市场上倾销更便宜的替代品。韩国一直在押注高额利润，但现在已经蒸发了。

@TrendSpider memory chip demand is cratering because china's flooding the market with cheaper alternatives. sk's been betting on premium margins that just evaporated

@YoShootBack 明天将会是美好的一天

@TrendSpider Tomorrow will be a beautiful day

@TrendSpider 原文 ↗

韩国股市尾盘回升，Samsung和SK Hynix分别上涨4%和1%

韩国股票开始转为上涨 👀

🟢 Samsung +4%
🟢 SK Hynix +1%
🟢 KOSPI Index +1% https://t.co/YfHiEq3g2e

展开原文

South Korean stocks are starting to flip green 👀

🟢 Samsung +4%
🟢 SK Hynix +1%
🟢 KOSPI Index +1% https://t.co/YfHiEq3g2e

❤ 364 · 🔁 33 · 💬 27 · 👁 6.3w

热门回复 4

@GhastlyOverton 它离崩盘只差7%。我敢打赌他们正在买入。

@TrendSpider It was like 7% away from breaking its neck. I bet they're buying.

@ryguyheath KOSPI领先还是QQQ领先？

@TrendSpider does the KOSPI lead the QQQ or vice versa

@redbeaver613 此时赌场的赔率比韩国市场还要可预测。

@TrendSpider At this point casino odds are more predictable than Korean market.

@NoOptions @TrendSpider https://t.co/a5jcn53ZcG

内存和芯片股遭抛售

全球内存和芯片股遭遇大规模抛售，SanDisk、Intel、AMD、ASML、Nvidia等多家公司股价大跌，显示半导体行业估值正在经历调整，市场对未来需求前景表示担忧。

@TrendSpider 原文 ↗

内存和芯片股集体大跌，SanDisk、Intel、AMD、Nvidia等单日跌幅4-14%

今天芯片和内存股票遭受重击 🩸

🔴 SanDisk $SNDK -14%
🔴 ARM $ARM -8%
🔴 Intel $INTC -6%
🔴 Seagate $STX -6%
🔴 Western Digital $WDC -5%
🔴 Applied Materials $AMAT -4%
🔴 AMD $AMD -4%
🔴 ASML $ASML -4%
🔴 Broadcom $AVGO -4%
🔴 Nvidia $NVDA -4%
🔴 TSMC $TSM -3% https://t.co/hzFFlkVatu

展开原文

Chip and memory stocks took a beating today 🩸

🔴 SanDisk $SNDK -14%
🔴 ARM $ARM -8%
🔴 Intel $INTC -6%
🔴 Seagate $STX -6%
🔴 Western Digital $WDC -5%
🔴 Applied Materials $AMAT -4%
🔴 AMD $AMD -4%
🔴 ASML $ASML -4%
🔴 Broadcom $AVGO -4%
🔴 Nvidia $NVDA -4%
🔴 TSMC $TSM -3% https://t.co/hzFFlkVatu

❤ 388 · 🔁 48 · 💬 46 · 👁 6.6w

热门回复 4

@Pulse24Media 这件事始于今天韩国的KOSPI、三星和SK海力士。现在这种压力已经蔓延到美国的半导体和内存股上。这看起来更像是全球AI基础设施情绪的重置，而不仅仅是局部地区性的动向。

@TrendSpider This started with KOSPI, Samsung and SK Hynix this morning. Now the pressure has spread across US semis and memory names. Looks more like a global AI infrastructure sentiment reset than an isolated regional move.

@stocksmcap 有人称之为流血，有人称之为机会https://t.co/GxAe4sPjcv

@TrendSpider Someone calls it bleeding, some call it an opportunity https://t.co/GxAe4sPjcv

@mishra_priyank 值得注意的是你列表中的分歧：NAND/存储（SNDK -14%，STX -6%）遭受的打击远大于HBM股票（NVDA -4%，TSM -3%）。在一个下跌的日子里，那些涨幅最大的股票会首先遭受重创，而MU在进入时已经比其200日均线高出100%。beta回撤是自上而下的。

@TrendSpider Worth noting the split in your own list: NAND/storage (SNDK -14%, STX -6%) got hit way harder than the HBM names (NVDA -4%, TSM -3%). On a red day the stuff that ran the most bleeds first, and MU was +100% above its 200-day going in. Beta unwinds top-down.

@burak_finance 伟大的机会不会轻易来临。

@TrendSpider Great opportunities don't come easily.

@TrendSpider 原文 ↗

SK Hynix盘中跌10%，延续韩国半导体股下跌趋势

SK Hynix 目前在韩国股市下跌 10% 🩸 https://t.co/znt0Bups77

展开原文

SK Hynix is currently down 10% on the Korean stock market 🩸 https://t.co/znt0Bups77

❤ 547 · 🔁 54 · 💬 46 · 👁 9.0w

热门回复 4

@fattyfatman 在上周五的上市后，很高兴看到韩国这次反倒把美国给坑了。

@TrendSpider After friday's listing it's nice to see Korea rug the US for once.

@K_Stockalper @TrendSpider https://t.co/2GjTCjYxOD

@scopuly 图表刚刚变得紧张起来。

@TrendSpider The charts just got nervous.

@bestill_n_beast https://t.co/HZJKBPqbw5

@TrendSpider 약세가 계속되네요
https://t.co/HZJKBPqbw5

@TrendSpider 原文 ↗

内存股抛售延续至盘后交易，SanDisk、MU、WDC继续下跌

内存股票的血本已经蔓延到盘后交易 🩸

盘后百分比变化
🔴 $SNDK -2.5%
🔴 $MU -1.5%
🔴 $WDC -1.5%

展开原文

The blood in memory stocks is now spilling into after-hours 🩸

After-hours % change
🔴 $SNDK -2.5%
🔴 $MU -1.5%
🔴 $WDC -1.5%

❤ 269 · 🔁 11 · 💬 30 · 👁 7.0w

热门回复 4

@Crytal420 血流成河🤣🤣🤣Sandisk在过去一年只涨了35倍，这真是场大屠杀！点击诱惑是有效的。去死吧你

@TrendSpider Blood 🤣🤣🤣 Sandisk is only up 35x over the past year what a blood bath! Click bait counts. Suck a dick

@TheMarketTell 市场正在像AI内存需求正在下降一样来定价MU。然而，超大型数据中心的AI资本支出还没有显示出这种情况。要么是美光的基本面比预期更快恶化...要么是这次抛售的幅度过大。

@TrendSpider The market is pricing $MU like AI memory demand is rolling over.

Yet hyperscaler AI capex hasn’t shown that yet.

Either Micron’s fundamentals are deteriorating faster than expected… or this selloff is overdone.

@WanWu70 SNDK被击中比其他股票更严重，有什么特别的原因吗？

@TrendSpider Any particular reason SNDK is getting hit harder than the others?

@VeinyChungus @TrendSpider Overnight green

@TrendSpider 原文 ↗

内存股年内涨幅曾接近4000%，但近期开始大幅回吐

仍然难以相信这些是 1 年的价格变化...

🔥 Sandisk $SNDK +4,235%
🔥 Western Digital $WDC +783%
🔥 Micron $MU +717%
🔥 Seagate $STX +515%
🟢 Intel $INTC +372%
🟢 Advanced Micro Devices $AMD +286%
🟢 Applied Materials $AMAT +209%
🟢 ASML $ASML +126%
🟢 Arm Holdings $ARM +123%
🟢 Taiwan Semi $TSM +90%
🟢 Broadcom $AVGO +46%
🟢 Nvidia $NVDA +27%

展开原文

Still hard to believe these are 1 year price changes...

🔥 Sandisk $SNDK +4,235%
🔥 Western Digital $WDC +783%
🔥 Micron $MU +717%
🔥 Seagate $STX +515%
🟢 Intel $INTC +372%
🟢 Advanced Micro Devices $AMD +286%
🟢 Applied Materials $AMAT +209%
🟢 ASML $ASML +126%
🟢 Arm Holdings $ARM +123%
🟢 Taiwan Semi $TSM +90%
🟢 Broadcom $AVGO +46%
🟢 Nvidia $NVDA +27%

@TrendSpider 本周芯片和内存股票遭受重击 🩸

🔴 Sandisk $SNDK -16%
🔴 Micron $MU -14%
🔴 Western Digital $WDC -11%
🔴 Seagate $STX -11%
🔴 Applied Materials $AMAT -7%
🔴 Arm Holdings $ARM -7%
🔴 Intel $INTC -6%
🔴 Broadcom $AVGO -4%
🔴 ASML $ASML -2.5%
🔴 Advanced Micro Devices $AMD -2%
🔴 Taiwan Semi $TSM -1%
🔴 Nvidia $NVDA -0.4%

Chip and memory stocks are taking a beating this week 🩸

🔴 Sandisk $SNDK -16%
🔴 Micron $MU -14%
🔴 Western Digital $WDC -11%
🔴 Seagate $STX -11%
🔴 Applied Materials $AMAT -7%
🔴 Arm Holdings $ARM -7%
🔴 Intel $INTC -6%
🔴 Broadcom $AVGO -4%
🔴 ASML $ASML -2.5%
🔴 Advanced Micro Devices $AMD -2%
🔴 Taiwan Semi $TSM -1%
🔴 Nvidia $NVDA -0.4%

❤ 347 · 🔁 41 · 💬 25 · 👁 9.4w

热门回复 4

@sophocles_dice @TrendSpider $AMD weekly

@Pulse24Media 一年前，NVDA看起来就是整个AI故事。现在即使那次的上涨也显得相形见绌，相比之下内存领域发生了什么。内存不再是次要的故事，它正在成为核心的AI基础设施。资金明显已经轮动到支持大规模AI所需的层面上。下一个问题是AI应用能否足够快地增长以证明这次基础设施建设的合理性。

A year ago, $NVDA looked like the entire AI story. Now even that run looks modest next to what's happened in memory.

Memory isn't a side story anymore, it's becoming core AI infrastructure. Capital has clearly rotated toward the layers needed to support AI at scale.

The next question is whether AI applications can grow fast enough to justify this infrastructure build-out.

@jebol321 这些涨幅是历史性的，但请记住，抛物线式的上涨往往预示着急剧的回调。在这个市场里，涨得这么快的东西同样会迅速 retracement。请谨慎管理风险

@TrendSpider These gains are historic, but keep in mind that parabolic moves often precede sharp corrections. In this market, what goes up this fast can just as quickly retrace. Manage your risk carefully

@Incognito_23445 真正的争论不是谁制造芯片，而是谁能长期获利于这项基础设施。NVDA的工具今天可能很残酷，但像AMZN、MSFT和GOOGL这样的超大型数据中心正在构建几十年都不会消失的经济护城河。我更喜欢拥有这块土地而不是拥有时髦的工具。

@TrendSpider The real debate isn't who makes the chip, but who monetizes the infrastructure long-term. $NVDA picks and shovels are brutal today, but hyperscalers like $AMZN, $MSFT, and $GOOGL are building economic moats that will last for decades. I prefer owning the land over the trendy tool