美股盘前情报 | 2026年6月3日 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 来源：CNBC · Benzinga · StockTwits · The Motley Fool · TheStreet Pro · Reuters · Bloomberg · LSEG · FTSE Russell · · Yahoo Finance · The Globe and Mail 数据窗口：过去24小时（优先过去12小时） ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 【市场快照】 • S&P 500 期货：-0.3% | 纳斯达克期货：+0.4% • VIX：15.77（一年新低，中性偏低） • 恐惧与贪婪指数：66（贪婪区间） • WTI原油：约93美元/桶（伊朗紧张局势提供支撑） • 美元指数：DXY 107.2 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 【第一部分：头条新闻 — 科技/AI/半导体】  黄仁勋台北GTC（2026年6月2-3日）：最大单一催化剂 ━━━━━━━━━━━━━━━━━━━━ 来源：The Motley Fool · StockTwits · Bloomberg · Reuters · Yahoo Finance 等 1. MRVL +32.5%（周二）→ 创历史新高。黄仁勋在台上表示：“Matt（Murrett）正在打造下一家万亿美元公司。”MRVL今日盘前再涨+22%。 → 非英伟达定制AI芯片 = 新AI阿尔法。顶级科技基金经理Gary Black：“随着焦点转向定制AI ASIC，Broadcom和Marvell是最大赢家。” 2. AVGO Broadcom：与MRVL一起创52周新高。定制AI网络/DPU芯片在企业端份额提升。Gary Black称AVGO是“定制芯片转型的大赢家”。 3. HPE Hewlett Packard Enterprise：盘前创52周新高。分析师：“AI驱动的强劲季度——股价值得更高估值。”Aruba/AI基础设施 backlog 强劲。 4. 英伟达Vera Rubin平台：今日在台北GTC宣布进入量产。下代AI数据中心GPU。NVDA盘前小幅下跌0.5%，市场正在消化MRVL/AVGO等竞争对手的狂飙。 5. 台积电ADR 创纪录新高 +2.5%。台北GTC确认与英伟达深化合作。费城半导体指数（SOX）大涨5.9%至历史新高——2026年最大单日涨幅。 6. IPG Photonics、MACOM、Amkor：黄仁勋GTC台北 keynote 后集体暴涨。AI光子学 + 先进封装 = 下一大瓶颈。 7. 特朗普行政令：政府优先获得先进AI模型使用权。StockTwits称这是“利好AI基础设施”信号。微软+英伟达合作：RTX Spark + Vera CPU用于Windows AI笔记本。本周Build 2026大会即将举行。 8. MU Micron：瑞银上调目标价至1625美元（潜在上涨超100%）。HBM内存需求进入结构性超级周期。  警示信号：Michael Burry（《大空头》）警告：AI芯片狂热距离2000年互联网泡沫峰值仅差7%。图表显示需对AI momentum股保持谨慎。散户与机构仓位出现分化。 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 【第二部分：散户情绪 — StockTwits/TradingView/WallStreetBets】  过去24小时散户最热议个股： • MRVL：StockTwits热度第一。黄仁勋点名后散户涌入，看多评论主导。“MRVL冲200美元”呼声四起。 • AVGO：定制AI芯片主题带动散户兴趣。“这是2026年的英伟达”情绪升温。 • ASTS：结束两连跌。高管集体买入提振信心。SpaceX IPO传闻带来扰动，但长期逻辑完好，散户整体看多。 • RKLB / LUNR / RDW：受SpaceX IPO猜测拖累下跌。空头认为SpaceX可能蚕食发射需求。 • META：散户视作买入机会。尽管有新订阅、裁员和云计划，股价仍落后Mag 7其他公司。“META相对GOOGL/AMZN仍便宜”。 • TSLA：SpaceX合并传闻导致回调。散户大V认为牛市情景可额外增加4500亿美元估值。 • INTC Intel：今日意外大涨。投资者提问“英特尔为什么涨？”AI PC + 代工 turnaround 叙事回归。 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 【第三部分：市场主题 — AI轮动 + 小盘股突破】  AI资金从超大盘轮动至中小盘： • 《The Globe and Mail》：2026年小盘股成为AI最大赢家 • Russell 2000小盘价值股年内跑赢成长股9个百分点 • AI狂热正驱动Russell 2000异动（Benzinga） • IWM、VTWO、URTH均闪现强烈买入信号  Russell 2000指数重构进行中：美国股市总市值达75.6万亿美元，同比增加29%。指数调整带来波动与机会。  板块轮动：能源（油价93美元、伊朗因素）+ AI半导体 + 小盘价值股领涨。防御板块（公用事业、消费必需品）资金流出。  中东局势：油价维持90-95美元区间。美国-伊朗紧张升级，能源股（XOM、CVX）获地缘溢价支撑。 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 【第四部分：重点个股 — 权威数据】  MRVL — Marvell Technology 价格：约140+美元（盘前+22%）| 52周新高 催化剂：黄仁勋称“下一家万亿美元公司” Gary Black：定制AI芯片大赢家 Morgan Stanley目标价120+美元 叙事：云端定制ASIC（谷歌TPU、亚马逊Trainium客户）。数据中心定制硅片成为长期趋势。 风险：估值偏高（PS 20倍），短期超买。  AVGO — Broadcom 价格：约220+美元 | 52周新高 催化剂：定制AI网络芯片 + VMware协同 叙事：英伟达之后第二大AI芯片公司。 风险：估值已计入强劲增长。  HPE — Hewlett Packard Enterprise 催化剂：AI驱动的超预期季度，Aruba网络与GreenLake AI服务 backlog强劲。 叙事：AI基础设施 + 边缘计算 + 混合云 = 多年增长。  TSMC 价格：446.69美元（+2.5%，纪录新高） 催化剂：与英伟达深化合作，先进制程产能成为结构性护城河。  IPG Photonics / MACOM / Amkor：GTC keynote后集体暴涨。AI光子学 + 先进封装成为新瓶颈。 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 【第五部分：美股小盘阿尔法  】  IWM — Russell 2000小盘ETF 触发：小盘价值跑赢9个百分点，AI轮动确认。 催化剂：今日ADP就业数据、本周联储讲话、6月底Russell重构完成。  FLNC — Fluence Energy（AI数据中心冷却） 触发：黄仁勋强调AI数据中心基础设施，液冷/电源成下一瓶颈。 验证：GTC后股价强势上涨，机构买入量增加。  SOXL — 半导体3倍做多ETF 触发：SOX指数+5.9%创历史新高，MRVL+32%。 风险：3倍杠杆存在衰减，仅适合短期战术操作。 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 【第六部分：宏观 + 市场结构】  今日重点事件（2026年6月3日）： • ADP私营部门就业（5月）预期+18万 • ISM服务业PMI • EIA原油库存 • 联储官员讲话  主要风险： 1. 伊朗/中东升级 → 油价突破95美元 → 通胀风险 → 联储偏鹰 2. AI芯片泡沫（Burry警告）→ 板块可能出现5-10%回调 3. SpaceX/OpenAI等IPO可能抽走市场流动性 4. 对60个经济体的关税压力 5. 台海紧张导致供应链中断 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 【第七部分：机构资金流向】 • AI/基础设施：MRVL、AVGO、HPE、TSM —— 机构大量买入 • 能源：XOM、CVX —— 地缘溢价流入 • 小盘价值：IWM、VTWO —— 2026年首次显著机构轮动 • 流出：公用事业、消费必需品、REITs ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 【投资总结 — 三种情景】  牛市情景：定制AI主题延续 → MRVL/AVGO/TSM领涨 → 小盘爆发 → IWM突破250+ 催化剂：MU、AMD财报超预期 + ADP就业数据温和  熊市情景：Burry警告兑现 → AI芯片见顶 → MRVL/AVGO回调 → 纳斯达克回调3% 催化剂：ADP强劲 + 通胀升温 → 降息预期推迟  基准情景：AI基础设施长期牛市 → 半导体高位震荡 → 小盘分批轮动 → VIX维持15-18 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━  免责声明：本内容仅供参考，不构成任何投资建议。数据来源于公开英文财经媒体。历史表现不代表未来结果。投资前请咨询持牌财务顾问。

Who, in their wildest delusion, is trying to weaponize 'overcapacity' — while the 'China Dividend' is actually powering the world? 谁在妄言“产能过剩”？“中国红利”正在驱动全球！ Recently, the phrase "China's overcapacity" has been appearing with striking frequency. In North America and Europe, it comes up in think tank reports, trade hearing testimonies, and political speeches — all in remarkably uniform tone. But interestingly, the louder the outcry, the more the actual market moves in the opposite direction. This summer, European consumers rushed to buy Chinese air conditioners — in the first half of this year, China's air conditioner exports to the EU surged 43.2% year-on-year, hitting a record high. Meanwhile, developers worldwide are flocking to Chinese large-language models — on OpenRouter, a multi-model aggregation platform, all top-five models by weekly calls come from Chinese companies. One vote is cast with wallets; the other, with code. If there really were "overcapacity," why would the market still be scrambling for these products? The so-called "China overcapacity" is, in the final analysis, a false proposition that cannot withstand market scrutiny. Recently, China's Ministry of Commerce released a 12,000-word position paper titled "The So-Called 'Overcapacity' Issue: China's Position," which gets to the root of the "overcapacity" narrative. The document systematically responds to relevant international discussions and firmly pushes back against the erroneous argument that "Chinese subsidies cause overcapacity." One sentence in the paper cuts straight to the heart: "A large export volume and a trade surplus of a country are the inevitable outcome of the international division of labor in the era of economic globalization." The weight of this statement becomes particularly clear when viewed against the broader coordinates of global economic development. In psychology, there is a concept called "attribution theory": people tend to make "internal attributions" for their successes — crediting their own efforts — but lean toward "external attributions" when facing problems, shifting blame to outside factors. This tendency is especially pronounced among trade protectionists. When their industries fail to keep pace with the times, fall behind, or suffer from domestic hollowing-out, they point fingers outward and blame China. When 80% of U.S. chips are exported and two-thirds of Boeing aircraft are sold overseas, that is touted as industrial strength. When the EU runs hundreds of billions of dollars in trade surpluses year after year in automobiles, pharmaceuticals, and cosmetics, that is called normal international trade. Yet when China exports a lot, why is that suddenly "overcapacity"? Why is that "dumping at low prices"? Different yardsticks yield different conclusions — this is double standards in action. What is more telling is that by now, the "overcapacity" hype has long ceased to be a purely economic issue; it has been fully reduced to a tool of geopolitical rivalry in the West. The U.S. enacted the Inflation Reduction Act, doubling down on exclusionary local protectionism; the EU introduced the Industrial Accelerator Act, directly tying subsidies to local production. They pay lip service to fair competition, but in reality, they are erecting targeted trade barriers — a classic case of saying one thing and doing another. No matter whether the trade protectionists in North America and Europe are genuinely misguided or willfully obtuse, China's attitude remains clear and steadfast: we openly embrace healthy global market competition, but we will never allow normal commercial rivalry to be deliberately escalated and twisted into geopolitical confrontation. If we cut through the fog of rhetoric and take an honest look at the reality of international trade, we will see that the sustained growth of China's exports is rooted in the genuine global demand for green transition and industrial upgrading. Take the new energy sector as an example: Chinese photovoltaic modules now hold over 90% of the global market share, and Chinese power batteries account for more than 70% of global shipments. This is not ineffective excess capacity piling up in the domestic market; rather, it is the large-scale industrial advantages of China precisely matching the urgent global need for green transformation. More importantly, the global demand for new energy is far from peaking. The International Energy Agency projects that from 2025 to 2030, global installed renewable energy capacity will double, with nearly 80% of that growth coming from solar photovoltaic. By 2030, annual global sales of new energy vehicles are expected to reach ten times the 2023 level. Therefore, China's industrial expansion overseas has never been a "market shock" — it is a development dividend that benefits the entire world. The numbers speak for themselves: over the past decade and more, China's contribution to global economic growth has remained steady at around 30%. China has been the world's second-largest importer for 17 consecutive years and is a major export destination for nearly 80 countries. It has implemented zero-tariff treatment for 63 countries and is the world's first major economy to offer zero-tariff coverage for all African countries that have diplomatic relations with it, as well as for all least-developed countries that have diplomatic ties with it. China is the only country in the world that hosts an International Import Expo, which has been successfully held for eight sessions, with cumulative intended transaction volume exceeding $580 billion. During the 14th Five-Year Plan period, China's cumulative imports surpassed $90 trillion. The trade surplus may stay in China, but the industrial dividends, jobs, and development opportunities have genuinely benefited the entire world. Trade barriers may block Chinese products, but they cannot stop the advance of technological progress, the momentum of industrial rise, or the market's choice as expressed through its own votes: China is not only an increasingly robust "world factory" but also a vibrant "world market." China's modern industrial development is not "China Shock 2.0" for the world — it is "China Opportunity 2.0." 最近，"中国产能过剩"这六个字，出现的频率有点高。在北美与欧洲，智库提，贸易听证会提，政客演讲也提，调子出奇地一致。但有意思的是，嘴上喊得越凶，真实的市场，越是朝着相反的方向走。 今年夏天，欧洲的消费者争相抢购中国空调 —— 上半年中国对欧盟空调出口额同比暴涨 43.2%，创下历史新高。全球的开发者追捧中国的大模型 —— 多模型聚合平台 OpenRouter 的周度调用量榜单上，排名前五的产品全部来自中国企业。一个用钱包投票，一个用代码投票。 如果真的"过剩"，市场为什么还在抢？所谓"中国产能过剩"，说到底，是个经不住市场检验的伪命题。 最近，中国商务部发布了一份1.2万字的立场文件——《关于所谓"产能过剩"问题的中方立场》，一口气把“产能过剩论”的根给刨了。这份文件系统回应了国际上的相关讨论，正面回击了"中国补贴造成产能过剩"的错误论调。 文件里有一句话，直击要害："一个国家出口规模大、存在贸易顺差，是经济全球化下国际产业分工造就的必然结果。"这句话的分量，放在全球经济发展的坐标系里看，会格外清楚。心理学有个“归因理论”：人们看待成绩时喜欢“内部归因”，认为是自己努力的结果；面对问题时则倾向“外部归因”，把矛盾归咎于外部因素。在贸易保护主义者身上，这种现象尤其明显。一些国家的产业发展跟不上时代了，落伍被淘汰了，或者是国内产业空心化了，就把锅往外甩，说是中国的问题。 美国芯片80%用于出口，波音飞机三分之二销往海外被称作产业优势；欧盟汽车、医药、化妆品常年保持数百亿贸易顺差被叫做正常国际贸易。同样是出口多，到了中国这里，怎么就成了"产能过剩"？怎么就成了"低价倾销"？标准不同，结论当然不同，这就是双重标准。 更值得深思的是，炒到现在，"产能过剩"早就不单纯是一个经济问题了，它已经彻底沦为西方地缘博弈的工具。美国出台《通胀削减法案》，大搞本土排他性保护；欧盟推出《工业加速器法案》，把补贴和本地生产直接绑定。嘴上标榜的是公平竞争，实际构筑的却是针对性的贸易壁垒，典型的“说一套，做一套”。 不管北美与欧洲的贸易保护主义者是真糊涂还是装糊涂，中国的态度始终清晰而坚定：我们坦然接受全球良性市场竞争，但绝不允许正常的商业竞争被刻意升级、扭曲成地缘政治对抗。 拨开舆论的迷雾，正视国际贸易的现实就会发现，中国出口的持续增长，根源在于全球绿色转型和产业升级的真实刚需。以新能源产业为例，中国光伏组件全球市场占有率超过 90%，动力电池全球出货量突破 70%。这不是国内市场积压的无效过剩，而是中国产业的规模化优势，精准匹配了全球绿色转型的迫切需求。更关键的是，全球新能源市场的需求，远远没有触顶。国际能源署预测，2025年到2030年，全球可再生能源装机总量将翻一番，其中近80%的增量来自太阳能光伏；到2030年，全球新能源汽车的年销量，将达到2023年的十倍。所以，中国产业出海从来不是什么 "市场冲击"，而是普惠全球的发展红利。 数据就摆在那里：过去十多年，中国对世界经济增长的贡献率常年稳定在30%左右。中国进口规模连续17年居全球第二，是近80个国家的主要出口目的地；对63个国家实施零关税，是全球首个对所有非洲建交国和所有建交的最不发达国家实现全覆盖零关税的主要经济体；中国是全球唯一举办国际进口博览会的国家，已成功举办8届，累计意向成交额超5800亿美元；“十四五”时期累计进口规模超90万亿元…… 贸易顺差留在中国，但产业红利、就业岗位、发展机遇，实实在在惠及了全世界。贸易壁垒，可以挡住中国的产品，却挡不住技术进步的脚步，挡不住产业崛起的大势，更挡不住市场用脚投票的选择：中国不仅是日益强大的“世界工厂”，更是活力旺盛的“世界市场”。中国现代化产业发展对世界不是“中国冲击2.0”，而是“中国机遇2.0”。 #China# #Jiangxi# #世界经济看中国# #赣出新精彩#

Last wk despite WTI +10% & ylds +2-7 bps across the curve, S&P/Nas/R2K +0.1%/+0.4%/+0.1%. $NVDA acquisition of Hugging Face & $Meta release of Muse Spark 1.3 last wk make both names more attractive into year-end. My view is that LLMs increasingly bifurcate into 90%+ usage of open-source/ open-weight models in the future as companies optimize the right models for the right task. Since the focus on controlling AI expenditures, the Silicon Data token cost has fallen over 50% since late May but the weekly usage of tokens in models across OpenRouter has increased by 3.6x over this same time. In addition, enterprises are increasingly focused on making sure their own proprietary data does not leak out when they use third party closed frontier LLMs. Hugging Face is the premier central collaborative platform, repository, and toolkit for open-source and open-weight AI with over 18 million developers. Nvidia has three customers that accounted for 44% of their revenues over the past six months and their largest customers are increasingly designing their own ASICs and in some cases selling them externally. A more diversified customer base that owns their own AI compute stack instead of renting from the big cloud service providers would help Nvidia with both of these issues. With this acquisition, Nvidia is in an even better position to sell enterprises a complete alternative AI stack (from the model to chips) where the customer will own their own data. Valuation is also compelling. Nvidia trades at a 15x CY27 PE versus their own guidance for 70% revenue growth and the Big 3 public cloud service providers at 21-23x for 15-26% total revenue growth. The S&P trades at 19x for 9% revenue growth for comparison. Nvidia is also up “just”24% versus the Semiconductor Index up 66% following underperformance last year at up 39% versus 42%. As for Meta, the stock is down 7% year-to-date after being up just 13% last year driven largely by concerns that 1) they can only monetize their near doubling in AI capex spend through efficiencies in their own business and 2) they were falling behind in the AI model race. The launch of the Muse Spark 1.3 API last week, catapulted Meta back to near frontier status (Top four in the Artificial Analysis Intelligence Index out of 10 models) but with aggressive token pricing (Bottom four in Cost per Task.) Open-weight versions of the Muse Spark lineup are coming soon. This will give the company another way to monetize their aggressive capex plans. This follows Meta's settlement in late August with state AGs on their youth addiction trial which was another overhang on the stock. Trading at 16x CY27 PE for 20% revenue growth is compelling with the settlement and Spark 1.3 launch as catalysts. From a broader market perspective, I recommend caution between now and the US mid-terms for reasons I have fleshed out in prior posts including: 1) Don’t Fight the Fed given I believe a 9/16 hike is likely 2) September -0.5% on avg & up only 48% of the time 3) S&P drawdowns of 10% in lead-up to mid-terms 4) Bipartisan pushback against datacenter expansion 5) Iran dragging out hostilities through US mid-terms

ZEROBASE WEEKLY 8.10-8.16 ZBT came under pressure this week, trading overall in the $0.085–$0.105 range. It opened near $0.105 on August 10, gradually pulled back through mid-week (dipping toward the mid-$0.08s), and closed the period around $0.083–$0.087. Despite the notable decline of roughly 15–20% from the weekly open, trading volumes remained decent on most sessions and liquidity conditions stayed relatively stable, with bid-ask spreads holding at reasonable levels. Crypto markets traded lower and range-bound this week amid fading post-jobs momentum, mixed ETF flows, regulatory delays, and persistent Middle East supply risks. Total cryptocurrency market capitalization drifted from roughly the $2.21–$2.22T area early in the period toward approximately $2.16T by the weekend, reflecting modest net outflows in risk appetite. Bitcoin opened the week near $64,800–$65,000 on August 10 (with intraday highs above $65,300), then ground lower through mid-week, touching lows near $62,500–$62,800 before stabilizing. It closed the period around $62,900–$63,100, for a net weekly decline of roughly 2.7–3.5% from the August 10 levels. Ethereum moved in a tighter band, starting near $1,900–$1,910, dipping below $1,870, and finishing near $1,870–$1,880 — a weekly loss of approximately 1.5–2%. Derivatives metrics pointed to cautious positioning. Open interest held relatively steady-to-soft (total crypto OI near $117B by weekend), 24-hour liquidations stayed moderate outside of brief volatility spikes, and funding rates on major pairs hovered near neutral to mildly negative, consistent with reduced leverage appetite in thin summer liquidity. Macro and geopolitical developments supplied the main headwinds and occasional relief. Ongoing U.S.-Iran tensions and the Strait of Hormuz disruption remained central: negotiations between Iran and Oman on temporary shipping arrangements stayed incomplete, with Tehran continuing to demand compensation, sanction relief, and an end to the U.S. naval blockade. Houthi strikes on Saudi facilities added to supply concerns. Brent crude rose from the mid-$80s early in the week toward the high-$80s (settling near $88.50 by August 16), while WTI climbed into the low-to-mid $80s — a weekly gain of more than 5% for both benchmarks after the prior week’s decline. Higher energy prices reinforced inflation stickiness concerns even as other data softened. July CPI data released on August 12 came in line with expectations and provided limited relief: headline CPI rose 0.1% month-over-month (3.4% year-over-year, down from 3.5%), while core CPI rose 0.2% MoM (2.5% YoY). Energy prices continued to ease on a monthly basis but remained elevated annually. The print, following the previous week’s weak nonfarm payrolls (-23,000), supported the view that the Fed could stay on hold longer, yet it failed to catalyze a sustained crypto rally as liquidity remained light and regulatory overhang persisted. U.S. equity markets finished the core trading week (through August 14) mixed. The S&P 500 posted a modest gain of approximately 0.4% (closing near 7,830), the Nasdaq Composite edged higher by roughly 0.1–0.2% (near 26,730), while the Dow Jones Industrial Average declined about 0.6%. Technology and semiconductor names showed dispersion amid AI-spending scrutiny and oil volatility; overall risk assets consolidated after the prior week’s stronger advance driven by the soft labor report. Institutional flows shifted from the prior week’s strong inflows. U.S. spot Bitcoin ETFs had recorded roughly $850–$865 million in net inflows over the preceding five sessions (August 3–7). This week opened with a notable outflow of approximately $145 million on August 10, followed by small positive or negative prints and further net outflows (including roughly -$61 million on August 12, -$131 million on August 13, and -$58 million on August 14), leaving the period net negative for BTC products. Ethereum ETF flows were more mixed, with intermittent modest inflows offsetting earlier redemptions. Regulatory caution added pressure: the SEC cancelled a planned meeting on crypto rules, and the Senate entered recess without advancing the Clarity Act (now eyed for September). The Crypto Fear & Greed Index remained firmly in Fear territory, fluctuating mostly in the 26–35 range (ending near 34). On-chain data offered a more constructive contrast to the soft price action. Large holders (“strongest hands”) continued to accumulate: the number of wallets holding ≥10,000 BTC reached a six-month high near 90, and addresses in the 10–10,000 BTC cohort added substantial volume (earlier estimates pointed to ~$1.5 billion equivalent accumulation since late July). Whales recorded one of the larger single-day accumulations in recent months (over 46,000 BTC on one notable session), while smaller/micro wallets distributed. Some dormant supply (2010–2017 vintage) moved, though residual non-clustered activity remained elevated relative to July. Exchange inflows from whales early in the week signaled selective distribution readiness, yet the overall rotation toward larger, longer-term holders continued to build support at current levels. In summary, the August 10–16 period delivered a measured pullback and consolidation in spot prices. Soft CPI and lingering hopes for a less restrictive Fed stance were outweighed by fading ETF momentum, regulatory delays, thin liquidity, and elevated energy prices from the unresolved Hormuz disruption. Higher oil is likely to keep near-term inflation sticky, yet the labor-market cooling already underway points to limited room for aggressive further tightening. With on-chain accumulation by large holders providing a floor and ETF flows showing early signs of stabilization potential, the market remains in a cautious consolidation phase within a still-complex macro and geopolitical backdrop.

$LITE Q3’26 EARNINGS HIGHLIGHTS 🔹 Revenue: $808M (Est. $809M) 🟡; +90% YoY 🔹 EPS: $2.37 (Est. $2.27) 🟢; +316% YoY 🔹 Gross Margin: 47.9% (Est. 45.2%) 🟢 🔹 Operating Margin: 32.2% 🔹 Cash, Cash Equivalents & Short-Term Investments: $3,172.3M Q4 FY26 Guide: 🔹 Revenue: $960M (Est. $935M) 🟢; +100% YoY 🔹 EPS: $2.85-$3.05 (Est. $2.74) 🟢; +1% to +8% YoY 🔹 Operating Margin: 35.0%-36.0% Other Metrics: 🔹 Series A Convertible Preferred Stock Issuance: Contributed approximately $2.0B to cash in March 2026 🔹 Acquisition: Business acquired in March 2026, details not disclosed 🔹 Restructuring Charges: $1.1M in Q3 FY2026 related to reduction in force 🔹 Cloud Light Escrow Settlement: $27.5M completed during nine months ended March 28, 2026 Commentary: 🔸 “Lumentum delivered an exceptional third quarter, with revenue growing 90% year over year to a record $808 million.” 🔸 “While our top line growth continues to garner headlines, the more impressive part of our recent performance has been our margin expansion.” 🔸 “In fiscal Q3, gross margin improved by 540 basis points on quarter and operating margin by 700 basis points.”

# Decision Points in AI Agent Development 🎯 **The Hook** Should your AI agent remember everything it hears? Memory write eagerness is one of the most delicate dials in agent design. Write too aggressively and you get memory pollution. Write too conservatively and your agent never learns. Getting this balance right is critical, and the stakes are higher than most teams realize. 📋 **Overview** Memory write eagerness controls how aggressively an agent persists information gathered during interactions into long-term memory. Think of it like a database INSERT: it is a quasi-irreversible operation. When an LLM's speculations or a user's ambiguous statements get written as facts, every future session references them as established truths. Errors become self-reinforcing. This is memory pollution, and it is one of the hardest problems to debug in production agent systems. 🔍 **Decision Points** This dial is primarily driven by two variables: 🔹 **Input Trust** — When end-user free-text is the primary source, the risk of injection and misinformation is high, so raise the write gate threshold. In admin-controlled input environments, you can afford to be more aggressive. 🔹 **Failure Cost** — In healthcare, legal, and financial domains, persisting incorrect facts leads to severe consequences. For an internal chatbot, a minor memory error can be corrected without much harm. Higher failure cost means stricter write suppression. 🔹 **Accountability** — When you need to explain "why was this stored in memory" after the fact, tracking provenance and confidence scores becomes essential. 💡 **Key Details** A practical three-tier framework for write decisions: ✅ **Auto-write** — Facts explicitly stated by the user ("My name is Tanaka," "I use Python") ⚠️ **Write after confirmation** — Information inferred from user behavior ("You seem to prefer Python" — confirm with the user before persisting) 🚫 **Never write** — LLM-generated speculation, unverified external sources, ephemeral context Attach confidence tags to memory entries and downrank low-confidence entries during retrieval. This limits pollution damage without completely blocking writes. Build deduplication into your write pipeline as well. Check new candidates against existing entries using cosine similarity (0.90-0.95 threshold), and overwrite same-entity same-attribute entries with the latest value. ⚖️ **Trade-offs** 📉 Too conservative — The agent never learns. Users repeat their preferences session after session, always getting default behavior. "I already told you this" becomes a recurring frustration. For use cases requiring long-term relationship building, this is a dealbreaker. 📈 Too aggressive — The biggest risk is hallucination persistence. "A-san probably lives in Tokyo" gets stored as "A-san lives in Tokyo" and treated as confirmed fact in all future sessions. Even more dangerous: prompt injection persistence. A single-session attack becomes a persistent injection when written to memory, affecting all future interactions. 🛠️ **Use Cases** 🏥 **Healthcare / Legal / Finance** — Extremely high failure cost. Minimize writes, record only explicitly confirmed facts, and always track provenance and confidence. 💬 **Customer Support** — Need to accumulate user preferences and history, but free-text input carries injection risk. Auto-persist only information confirmed through repeated interactions (2+ matches). Use a quarantine period for implicit preferences before promoting them. 🏢 **Internal Knowledge Bots** — Want to capture organizational tacit knowledge ("this API breaks if you pass this parameter"). Admin-controlled input allows more aggressive writing, but periodic "memory audits" where users review stored information maintain long-term quality. Never forget audit trails. Tracking when, what, and from which source each write occurred makes it possible to identify and fix the root cause when memory pollution is detected. #AIAgents# #SoftwareArchitecture#

🆕 A new open-source 1.6 TRILLION-parameter MoE that scores 0.1 points behind Claude Opus 5 on an agent benchmark has been released. @NexEcosystem released Nex-N2.5-Max. It beats DeepSeek V4 Pro 0813 on all but one comparable benchmark. This thing is enormous! Stats ... 🧠 1.6 TRILLION parameters 🔀 Mixture-of-Experts 🤖 Built for agents + long-horizon workflows 💻 Coding + tool use 📚 262K context 📝 Text-only ⚖️ Apache 2.0 📦 Weights are already on Hugging Face Nex's reported results 👇 AutomationBench 🟣 Claude Opus 5 → 50.3 🔥 Nex-N2.5-Max → 50.2 🔵 GPT-5.6 Sol → 45.8 🟢 Qwen3.8-Max → 39.8 And on BrowseComp: 🔥 Nex-N2.5-Max → 92.6 🟣 Claude Opus 5 → 90.8 Terminal-Bench 2.1 → 86.1 Toolathlon Verified → 74.7 😁 You're probably NOT running this one under your desk. The official model repository is about 1.65TB, and Nex's recommended deployment uses this monster hardware .... 🎮 16× H200 🖥️ 2 nodes ⚡ TP16 + expert parallelism But here's why I'm posting the monster first. It has a 35B multimodal little brother. And THAT one may actually belong in your home AI lab. 👀🔥

September returns for S&P 500 over 25 years: 2025: +3.50% 2024: +2.00% 2023: −4.90% 2022: −9.30% 2021: −4.80% 2020: −3.90% 2019: +1.70% 2018: +0.40% 2017: +1.90% 2016: −0.10% 2015: −2.60% 2014: −1.60% 2013: +2.97% 2012: +2.40% 2011: −7.20% 2010: +8.80% 2009: +3.60% 2008: −9.10% 2007: +3.60% 2006: +2.50% 2005: +0.70% 2004: +0.90% 2003: −1.20% 2002: −11.00% 2001: −8.20% 25-year average: −1.2%