xDai💙~🎭

At the beginning of X, there was no name, and everything is vanity was needed to break the stubbornness.

[xDai💙~🎭 @xdaixdai]

这个问题接续得很好，正好把“为什么会获得支撑”和“为什么反而剧烈波动”这两个看似矛盾的现象串联起来。

当价格从高点回落至长期均线（如200日均线）时，波动率飙升，是多空力量在此处空前集中、分歧巨大的必然结果。黄金和白银当前的表现，就是教科书式的演绎。

[概率鹿梦｜AI Wealth Edge @DeerLucid]

投资女王徐新：我做了31年投资，发现这样的人更容易成功。

[Eden @Web3Eden01]

DeepSeek 这波融资停摆，绝对是今年 AI 圈最大的黑天鹅。
梁文锋内部讲话泄露这事儿，性质极其恶劣。
这不仅仅是得罪了一个创始人的问题，这是把国产算力的家底和地道全给暴露了。

简单看几个流出的硬数据，你就知道老美看了会多开心：
1. 算力总量：
DeepSeek 拥有约 2 万张 H100 等效卡。
2. 硬件差距：
华为 950 性能落后英伟达 4 倍，周期落后 2 年。
3. 模型规模：
国内几十 B 参数 vs 美国 800 B 参数，这断层看得人绝望。

更离谱的是，会上还聊到了如何搞到那些受限的芯片。
这内容流到海外，几乎就是在提醒对面：
这里的出口管制还有漏洞，赶紧补上。

梁文锋叫停融资，是一种愤怒，更是一种自保。
对于 DeepSeek 这种走技术流的公司，数据和算力渠道就是命根子。
现在命根子被自己信任的投资人当成谈资发到了网上，这谁受得了？

那些为了发个朋友圈、显摆一下自己有内部料的投资代表，这次是真的玩脱了。
当生意场上的信任感彻底崩塌，损失的可能不仅是这一轮融资，而是整个国产 AI 环境的生存空间。
这次调查，大概率要抓典型了。

[美股大数据 @StockWe_Com]

下周，美股真正的焦点来了：财报超级周。

AI、云计算、半导体、消费巨头将集中交成绩单。

市场现在已经不再满足于“业绩增长”，更关注一个问题：

巨额AI资本开支，能不能真正变成收入和利润？

几个重点公司值得关注：

周一
📌 $NVDA 产业链相关： $AMKR 、 $RMBS
关注先进封装、HBM、AI服务器供应链需求是否继续强劲。

周二
📌 $MSFT 微软
重点看 Azure 增速、AI服务收入，以及资本开支是否继续扩大。

📌 $META
关注AI基础设施投入能否提升广告效率，市场更关心投入回报率。

📌 $AMD
关注AI GPU订单、MI系列出货，以及能否继续扩大市场份额。

周三
📌 $QCOM 高通
看AI手机、边缘AI需求是否开始贡献增长。

📌 $LRCX 泛林集团
半导体设备核心公司，关注晶圆厂投资周期是否继续。

📌 $ARM
关注AI服务器CPU和芯片授权增长。

周四
📌 $AAPL 苹果
重点看AI功能落地、iPhone需求，以及供应链变化。

📌 $AMZN 亚马逊
关注AWS云业务增速，以及AI投入是否开始产生回报。

📌 $MU 美光
存储周期核心公司，重点看HBM需求、DRAM价格和AI服务器需求。

周五
📌 $XOM 、 $CVX
关注能源价格和宏观环境变化。

这一周最大的看点不是谁财报最好。

而是市场会重新评估：

AI到底是在创造利润，还是只是在推动资本开支。

如果云巨头证明AI投入正在变现，硬件链可能继续受益；

如果资本开支增长快于收入增长，市场可能重新审视整个AI估值。

财报周，决定的不只是个股，而是下一阶段AI交易的方向。

#美股

[爱吃折耳根的Ace @_zheergen]

量化交易系统进阶必备：Python 语言底层四大核心思想

昨天群友问我：已经会用 Pandas、NumPy 写策略了，但想进一步写出真正高性能、可维护的量化交易框架、风控系统或复杂回测引擎，该从哪里提升 Python 能力？

我直接掏出我珍藏很久的 James Powell 讲师在 PyData 大会上的经典演讲《So you want to be a Python expert?》

我问他，想提升到这个层级，Python 语言层面最该先掌握哪些核心思想？

他给出的内容非常硬核且实用：

1️⃣ 数据模型与协议驱动：通过魔术方法（len、add 等）让自定义类具备原生数学运算能力
2️⃣ 元类与类构建钩子：如何用 Metaclass 或 init_subclass 强制约束派生类行为
3️⃣ 装饰器与动态包装：从手动计时到高阶带参装饰器的完整演进
4️⃣ 生成器 + 上下文管理器：现场手写类似 contextlib 的底层实现，优雅处理 setup/teardown

更有意思的是，主讲人全程用 dis 反编译字节码 + 现场手写代码，把抽象概念讲得异常清晰。

不过这个视频不讲量化策略，而是聚焦语言底层，适合已经熟练掌握 Pandas/NumPy、想提升框架设计能力的进阶开发者。

其适合想阅读 Zipline、Backtrader、Qlib 等开源量化库底层，或自己搭建大型交易系统的人，会很有启发。

建议直接收藏，认真看完会对写复杂 Python 量化系统有很大帮助。

[小牛 @Xiaoniu6161]

什么是先进封装，保证你能看懂。

[Jensen Huang @JensenHuang]

For my first post, I’m sharing a letter @NVIDIA signed on why open models matter.

AI will transform every industry, power every company, and be built by every country.

Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.

The world needs both frontier closed models and frontier open models.

https://t.co/AUKzoQ5Ikb

[Jurrien Timmer @TimmerFidelity]

Inflation remains elevated, which suggests that the Fed is more likely to raise rates than lower them.

[The Kobeissi Letter @KobeissiLetter]

We need more power.

Data center power demand in the US is projected to surge +253% from 2026 levels, to a record 194 gigawatts by 2035.

One gigawatt is roughly equivalent to the capacity of a single traditional nuclear reactor.

As a result, data centers are expected to account for ~20% of US electricity consumption by 2035, up from ~6% today.

To put this into perspective, data centers are estimated to consume ~12% of US electricity by 2030.

Much of this growth is concentrated in a handful of grid regions, such as PJM, which serves the District of Columbia and 13 states, including Virginia, Pennsylvania, and Ohio.

Electricity is the next AI bottleneck.

[Lukas Ekwueme @ekwufinance]

Markets are at all time highs... they are delusional

we’re losing:

- ~30% of fertilizers
- ~20% of LNG
- ~10% of oil
- ~30% of helium

These aren’t isolated commodities.... they sit at the core of entire production chains:

- Petrochemicals -> fertilizer -> food production
- Petrochemicals -> sulfur -> mining (copper, uranium, nickel)
- Petrochemicals -> plastics -> cars, electronics
- Petrochemicals -> drugs, rubber, textiles
- Helium -> semiconductors / AI chips
- Gas -> power generation
- Diesel -> transportation

So this isn’t just an energy problem... it’s a full-spectrum supply shock hitting food, industry, tech, transportation and power at the same time.

Our world doesn't function without Hormuz supplies.

And there is no policy tool that can replace missing physical supply.

[Neil Sethi NEW account, please refollow @neilsethinew]

Analysts at Signum Global Advisors led by Andrew Bishop have created a "TACO index" to attempt to discern when Pres Trump might de-escalate the Iran situation.

The index is comprised of Brent crude, the 10-year yield, the number of Hormuz crossings, and the SPX.

In backtesting they found that a 2.3 to 3.4 standard deviation move — or an average of 2.9 standard deviations -- has seen action from the President in the past.

Putting that in today’s terms, the analysts find it’s not yet time to TACO — but it’s getting closer. “Extrapolating linearly would suggest that a TACO could happen as early as July 22 and ‘should’ happen no later than July 30 (unless conditions materially improve, which seems unlikely) – with history suggesting July 26 as most likely,” they said.

[0x鸣人 @LuBtc888]

梁文峰的量化软件如何收割散户？

48小时完整割韭闭环，散户亏5%-7%全是精准算计

全程无死角拆解量化软件如何把散户拿捏得死死的，看完再也不敢盲目追涨！

量化操盘全流程：
1. 提前3天锁定小盘券商股，监控股吧情绪，抓准散户追涨信号
2. 9:15挂虚假大单诱多，让散户挂市价抢筹
3. 尾盘0.5秒极速撤单，开盘直接套牢追高盘
4. 慢拉股价营造突破假象，暗中融券布局
5. 散户情绪峰值瞬间砸盘，制造回调幻觉
6. 反复假托单+砸盘，逼散户批量止损割肉
7. 低位接回筹码，次日出货+还券，完美闭环

量化靠软件、数据、通道优势精准收割，
你以为的回调，全是算法设下的陷阱；
你冲动的加仓，都是给量化送筹码。

完整血腥收割细节全在视频里，
散户必看，守住自己的本金👇

[王局志安 @wangzhian8848]

梁文峰四小时交流会内容：
1，英伟达那套生态一年内就废了，华为的卡一年后完全能平替英伟达，贵一倍两倍都无所谓。
2，谁想着靠AI暴利谁就死，拿得少的人一定打败拿得多的人，OpenAI就是因为想独占才被围攻。
3，奇点不是什么爆炸，是个漫长的温水煮青蛙过程，而且顺序不能乱——先解决持续学习，再到自我迭代，最后才做机器人。
4，Ai技术路线是这样：语言模型 → 思维链(CoT) → Agent → 持续学习 → 智能奇点(自我迭代) → 具身智能，现在处于Agent走一半，持续学习是下一个必须跨过去的坎，跨过去才能进入AI自己迭代自己的奇点阶段。
5，克制是核心战略，主动放弃C端流量争夺战，不做超级App，短期商业化只是保底，全部资源主线聚焦AGI。
6，中美AI差距就一年，但国内只用了人家二十分之一的算力，人才根本不缺，差的就是卡。
7，公司一半的核心研究员都在标数据，这就是现阶段最重要的事，高端数据瓶颈不是钱，是时间。
8，最坏情况技术冻结了，光卖API也够撑起一家上市公司，商业化根本不用愁。
9，不加班、没KPI、没组织架构，全靠愿景驱动，因为做研究就得松弛，逼太紧反而没用。
10，国产芯片产能问题今年明年后年都卡，但五年后肯定解决了。
11，国内几十家做基模的公司太多了，最后就剩三四家，而且谁也别想有暴利。
12，视频生成、3D、世界模型这些东西跟智能上限没关系，纯粹是商业噱头，坚决不做。
13，员工期权拿够、团队稳住，AGI（通用人工智能）就一定能成，其他全是小事，最多晚个一年半载。
14，中国AI最终拼的是“系统性便宜”，就跟制造业一样，成本优势是结构性的。
15，持续学习要是先做出来了，通用智能就很简单，让AI自己帮自己研究就行。

[Global_Macro @Marcomadness2]

Crude back to March extreme backwardation

[Charlie Bilello @charliebilello]

Price increases since start of the Iran war...
Sulfur: +142%
European Natural Gas: +97%
Heating Oil: +62%
Jet Fuel: +44%
Diesel: +38%
Gasoline: +36%
Rice: +36%
WTI Crude Oil: +30%
Brent Crude Oil: +29%
Diammonium Phosphate: +25%
Cotton: +23%
Wheat: +18%
Palm Oil: +14%

[Holger Zschaepitz @Schuldensuehner]

The energy shock is reaching the long end: higher oil prices are helping drive long-term bond yields sharply higher as investors reprice inflation risk. UK 30y yields have climbed to 5.73%, US yields to 5.14%, Japan to 3.89% and Germany to 3.65%. Most striking: Japan now yields more than Germany after decades of near-zero rates.

[Global_Macro @Marcomadness2]

The deflated 2x Bubble

[Daniel Fang @fang_danie121]

昨天，沪市一恒通光电投资者跳楼，据说，该投资者见亨通光电涨势一直非常喜人，遂在6月中旬，梭哈该股，结果，该投资者入市即最高点，该股从6月中旬一直阴跌到现在。
--------------------
老目评论：从技术角度来说，该投资者至少犯了两大错误。
1，选择的入场点就是风险最大的位置，因为均线的位置彼此张开，这意味着入市肯定会遭受回调风险。所以如果入市，最佳点是四根均线收缩，一根K线救可以穿过。
2，没有设置止损，减少亏损同样也是一种盈利。
当然，在布尔什维克国家的股市，只能买涨，不能买跌，它的绝大多数股票基本又都是适合做空，这天然救决定了在中共国想要靠炒股挣钱等同于童话。

[Jurrien Timmer @TimmerFidelity]

With the Mag 7 losing its magnificence, the market continues to broaden out in a constructive way. Even though the S&P 500 cap-weighted index has not made a new high since June 2, this is preferable to a scenario in which the mega caps drag the index sharply lower. Sideways is fine (for a while).

[Visioner @visionergeo]

‼️‼️🇫🇷🇺🇲🇮🇷 BREAKING | The sudden decision by Air France to ground all commercial flights to Saudi Arabia and the United Arab Emirates for a full week delivers a massive shock to global aviation and signals an acute, immediate escalation in the Middle East crisis.

This drastic operational halt cuts off two of the most critical commercial and transit hubs in the Persian Gulf, reflecting an intelligence assessment that the airspace over the region is no longer safe for Western commercial carriers. By preemptively clearing its airliners out of the Gulf corridor, Air France is reacting directly to the impending threat of a high-intensity kinetic conflict, likely anticipating that imminent U.S. military strikes against Iran could trigger massive, uncoordinated retaliatory missile salvos and drone swarms across the entire theater.

This suspension is poised to trigger an immediate domino effect across the global aviation industry. As a premier European carrier takes this extreme precaution, other major international airlines will face intense pressure to rapidly reroute or cancel their own flights to avoid the airspace entirely. The financial and logistical fallout will be instantaneous, choking vital travel corridors, snarling international supply chains, and signaling to the global markets that the Persian Gulf is on the absolute brink of a full-scale, hot war.

See the latest updates with us: @visionergeo

[Bull Theory @BullTheoryio]

CHINA JUST SPENT $9 BILLION TO STOP ITS STOCK MARKET FROM COLLAPSING.

Two state owned funds, China Reform Holdings and China Chengtong, put nearly $9 billion into Chinese stocks over the weekend.

Before this, Chinese stocks had a brutal two weeks. Rising Middle East tensions, a global AI-chip selloff, and an $8.6 billion IPO from chipmaker CXMT drained cash out of the market.

Together, these wiped out $1.48 trillion from the Chinese stock market.

Regulators met with market participants before the state funds stepped in.

This is China's usual move during a bad selloff: state money buys just enough to stop the bleeding.

This rescue pushed prices up, but investors aren't fully convinced yet.

For the recovery to last, buying needs to spread beyond state backed stocks into private companies and financials.

That didn't happen last time China tried this, in 2021, when a similar $1.3 trillion crash came right back days after the government stepped in.