顺丰速运🧙

@Hoyooyoo 帮忙看看TEM

CRCL，短线143，中线200，长线300+。因此，100之上都建议长期持有。我不只是90非常准哈，我还做空到50块的目标位呢。

这套Claude Code教程讲得太细了 新手完全能听懂并且直接落地！ 静下心看完这一套完整教程，你就能真正用AI搭建工作流、实现自动化、处理复杂任务。 在国内这种质量的课，收费398、598元估计还有一大堆人抢着买…… 现在免费就能学到！ 适合想用AI提升效率、搭建自己专属Agent、自动化日常工作的朋友。 看完直接上手，少走很多弯路。

@Hoyooyoo 秋神看看TEM

特斯拉，圈圈是一个2break，向上则反转进入多头。(假突破则用力猛锤)

@BTCdayu 大宇讲讲TEM

《今日好文精选》 PayPal有4.3亿用户，为什么干不过USDC？ 稳定币赛道的赢家法则 CPN是 $CRCL 的AWS时刻 Circle：当开放系统面临考验 黄瓜猫：不陪WLFI玩了 帮 TAO 涨了 90% 的人，又亲手带崩了价格 王仕锐：打造医疗 AGI，让人活到150岁 t.me/btcdayu888/126…

@JKeynesAlpha @grok 总结一下

$TEM 95% of Big Pharma Didn't Get the Memo That Claude is Supposedly Going to Kill This Company 🤣 Tempus AI announced an expanded multi-year strategic collaboration with Gilead Sciences (NASDAQ: GILD) today. Under the deal, Gilead gains enterprise-wide access to Tempus' AI-powered Lens platform, unlocking broader datasets across multiple cancer indications alongside dedicated Tempus analytical services. Patrick Loerch, Senior Vice President of Clinical Data Science at Gilead, framed the partnership around combining Gilead's scientific expertise with Tempus' real world data insights in oncology to inform clinical decision making and improve care for cancer patients. This follows the expanded multi-year Merck collaboration announced March 3, where Merck gained access to Tempus' de-identified multimodal data and Lens Platform for precision oncology biomarker discovery. TEM is trading around $45 today. Down nearly 7% on the session. Down 56% from its October 2025 all-time high of $103.25. The stock is getting hit by a confluence of macro and narrative pressure that has nothing to do with the actual business. The Iran war has created a broad equity risk-off environment that has dragged growth names lower across the board. Oil prices near $110 a barrel, the Strait of Hormuz blockade, etc. have investors selling anything with duration or perceived risk. On top of that, the SaaSpocalypse narrative that follows every Anthropic announcement has cast a shadow over anything with "AI" in its name. Investors have been indiscriminately punishing anything that touches AI-driven enterprise workflows, regardless of whether the underlying business is actually threatened. The market is getting this wrong. Claude and the broader large language model ecosystem are not competitive threats to Tempus AI. They are not even operating in the same problem space. Claude Cowork automates white-collar workflows. It triages NDAs. It drafts outreach emails. It researches case law. It writes code. These are horizontal, task-level productivity functions that replace or augment human seat-based work inside SaaS platforms. That is why many SaaS names, fairly or not, are getting crushed. Their revenue models depend on human headcount using software, and Claude demonstrated that AI agents can collapse that headcount, and there's a generalized fear that they are so powerful they can just replace all software or do so in unpredictable ways that make SaaS toxic. However, Tempus does not sell seats. It does not sell per-user software licenses. It sells access to one of the world's largest libraries of de-identified multimodal healthcare data, covering clinical, genomic, transcriptomic, proteomic, and imaging records linked to over 40 million de-identified research records. It sells genomic and molecular diagnostics. It sells AI-driven analytical services that sit on top of proprietary, curated datasets that took a decade to build and that no LLM can replicate from public information. The frontier model companies are not competing with Tempus. They are approaching Tempus (as Lefkofsky noted in the recent JP Morgan conference). The relationship runs in the opposite direction of what the selloff implies. AI reasoning capabilities are rapidly commoditizing. Every major lab, Anthropic, OpenAI, Google, Meta, is converging toward similar frontier performance on general tasks. The intelligence layer is becoming commodity. The proprietary, curated, multimodal healthcare data that sits underneath the models is not. That is where Tempus lives. No amount of Claude reasoning capability can generate de-identified longitudinal patient records linked across genomics, pathology, radiology, and treatment outcomes. That data does not exist on the open internet. It exists inside Tempus' infrastructure because Tempus built the relationships with health systems, processed the samples, ran the assays, structured the records, and trained the models on top of them. The moat is the data asset. The moat is the flywheel. More diagnostic volume generates more data. More data improves the AI models. Better models attract more pharma partnerships. More partnerships generate more revenue. More revenue funds more diagnostic capacity and more data acquisition. This flywheel compounds over time and becomes progressively harder to replicate. In healthcare specifically, the barriers go beyond data. Professional diagnostics carry clinical and regulatory credibility that cannot be shortcut by a general-purpose AI. FDA-cleared products like the xR IVD RNA NGS assay, the Pixel cardiac imaging platform, the ECG-LVEF software, and the HRD-RNA algorithm all required rigorous clinical validation. Clinical reputation takes years to build. Hospital system integrations, like the David EHR co-pilot deployed at Northwestern Medicine, require deep institutional trust. Over 50% of oncologists in the United States are connected to Tempus through sequencing, clinical trial matching, and research-enabled partnerships. That network was not assembled by merely writing code or deploying a general model. It was assembled by processing millions of tumor samples, delivering actionable results to clinicians, and proving clinical utility over years of operation. Lefkofsky made this point explicitly on the Q4 2025 earnings call in February. When asked about competition from broader AI entrants in healthcare, he argued that the most durable AI business models center around two things: access to proprietary data to train models and proprietary distribution once you have a model that generates insight. He said Tempus has both at scale. Over 450 petabytes of connected multimodal data flowing from its diagnostic business, with real-time connection to outcomes and response, tracking patients longitudinally across rich molecular and imaging data. Distribution through more than 5,500 hospitals and more than 8,500 regularly ordering oncologists. That is the prime real estate for AI in healthcare. The foundation models are where this thesis becomes tangible. Tempus is building its own domain-specific foundation models and is further along than most investors realize. The first is the oncology foundation model, built in collaboration with AstraZeneca and Pathos AI under the $200 million deal announced in April 2025. This is designed to be the largest multimodal foundation model in oncology, trained on Tempus' de-identified patient records spanning clinical notes, genomics, imaging, and transcriptomics. Lefkofsky confirmed on the Q4 call that the model had a Q1 2026 deliverable where it had to hit specific benchmarks established by AstraZeneca, and that Tempus believes it hit those benchmarks and submitted the work to AstraZeneca for testing. To support this, Tempus established a dedicated compute cluster of over 1,000 NVIDIA H200 GPUs. It has since procured a second cluster of more than 500 NVIDIA GB200 GPUs, which Lefkofsky said provides greater compute power than the first cluster. That is serious infrastructure commitment for a company that most investors still think of as a diagnostics lab. The oncology foundation model is not the only one. Lefkofsky disclosed on the same call that Tempus is running additional internal foundation models across other modalities and disease areas, including radiology, pathology, cardiology, and neuropsych data. He said the company is "doubling down" on those efforts. These internal models are being trained on Tempus' proprietary data and are not dependent on external partnerships. They represent the next phase of the flywheel: as the foundation models mature, they will generate increasingly differentiated insights, which Lefkofsky said will accelerate the stacking of algorithmic value on top of diagnostic results. He framed it clearly: to the extent that Tempus is already some percentage better than competitors because it can generate more insights, that gap should be expected to grow quite a bit over the next several years as the foundation models come online. The frontier model companies are commoditizing general-purpose reasoning. Tempus is building domain-specific foundation models on top of proprietary healthcare data that no one else has. The intelligence layer commoditizes. The data layer and clinical distribution layer do not. Tempus sits on both. This is why 95% of the top 20 pharmaceutical companies by revenue are Tempus partners. The list of companies that have signed data agreements, strategic collaborations, or enterprise-level Lens platform access with Tempus reads like a directory of global pharma and biotech. AstraZeneca. Merck. Pfizer. Gilead Sciences. GlaxoSmithKline (GSK). Bristol Myers Squibb. Novartis. AbbVie. Eli Lilly. Daiichi Sankyo. Boehringer Ingelheim. Johnson and Johnson (Janssen). Bayer. Genmab. Tempus claims 95% of the top 20 pharma companies as partners, medical device, health system, and institutional partnerships with Medtronic, Illumina, Northwestern Medicine, NYU Langone Health, Median Technologies, and the ARPA-H ADAPT program. In 2025 alone, Tempus signed data agreements with over 70 customers. The $200 million AstraZeneca/Pathos AI foundation model deal demonstrates exactly how the value chain works. The AI companies and pharma companies are coming to Tempus for the data. Tempus is not losing to AI. It is selling to AI. Revenue grew 83% year-over-year in 2025 to $1.27 billion, with 30% organic growth excluding the Ambry acquisition. Diagnostics revenue grew 122% in Q4, driven by 29% oncology volume growth and 23% hereditary volume growth. Data and Applications revenue grew 25% in Q4, with Insights (data licensing) growing 70% excluding prior-year warrant effects. MRD testing volume hit 4,700 tests in Q4, up 56% quarter-over-quarter. The company guided $1.59 billion in revenue for 2026 and approximately $65 million in adjusted EBITDA, its first full year of meaningful positive adjusted EBITDA. Total remaining contract value exceeds $1.1 billion with 126% net revenue retention. Lefkofsky noted on the call that bookings have been so strong that Tempus starts 2026 with greater visibility into the data revenue build than it has ever had, describing the current demand environment as "crazy amounts of demand" for its data products. The stock is trading at roughly 5x forward revenue at an $8 billion market cap. The average analyst price target is $76, representing 67% upside from current levels. The company ended 2025 with $760 million in cash and marketable securities. This is a business that is scaling, approaching profitability, stacking enterprise pharma partnerships, building two foundation model clusters on NVIDIA's latest silicon, and expanding its diagnostic product portfolio. It is being sold because investors are scared of a chatbot and because a war in the middle east is pushing oil higher. When the Iran conflict resolves or de-escalates, and when the market realizes that Claude and its peers are not going to replace precision oncology data or any of Tempus' multimodal data and infrastructure, TEM has the fundamental profile to round-trip back toward its prior highs and much higher. The Gilead deal today, on top of Merck, on top of Daiichi Sankyo, on top of the Medtronic ALERT trial results, on top of the Median Technologies lung cancer screening collaboration, all in the first four months of 2026, tells you that the business is accelerating while the stock is declining. That is the setup for asymmetric upside. The SaaSpocalypse will not kill Tempus AI because Tempus AI is not a SaaS company. It is a healthcare data monopoly in training, with a diagnostics engine underneath it, a growing AI analytics layer on top, two foundation models in development on dedicated GPU clusters, and the most comprehensive pharma partnership network in precision medicine. The market is conflating a horizontal software disruption narrative with a vertically integrated precision medicine platform. AI is becoming a commodity. Data, clinical reputation, regulatory clearance, and institutional trust are not. That conflation will correct. And when it does, the move higher will be violent. I maintain my bullishness on Tempus, my only thought is I want to own even more at these prices. "To the extent advancements in technology and AI have failed to live up to their hype, I believe it's largely a function of timing. People underestimated the complexity and scale of the healthcare industry, and the often morassic pace at which it evolves. But the tsunami is coming and we own lots of internal farmland that is about to become beach front property. We intend to acquire more." - Lefkofsky Q4 Letter

@wufantouzi 第三国家需求增加？ 上次听到类似的话还是在gamefi

iphone13这台二手手机拿货价1800-1900 东南亚的二手商来香港采购2300-2400 新机苹果已经停产 但是二手交易非常火热 你经常看到的爱回收或者转转 就是这个行业的上游 下游商户一单几百台的买卖 商户一天也赚几万块 最后再转到非洲东南亚等地 这些地方的消费者 最终买走了这些二手手机 这栋楼相当于是香港的“华强北” 这种业务形态其实是传统的外贸行业 香港是自由贸易港 物流和税收有优势 所以这里就成了全球苹果二手手机集散地之一 为什么来这里看呢 因为以前老外是拿着港币来收货 现在都是用稳定币结算 越来越多的老外卡里存的不再是usd 而是钱包存着usdt和usdc来扫货 你问他们为什么不用港币或者usd了 他们说更加方便 这个行业里 已经有20-30%的结算采用稳定币结算 你注意到这个支付行业的范式转移机会了吗

@lufeieth 建仓，准备开始买了

@HCL2018888 忍不住想割肉还是忍不住想加仓？😂

Tempus AI(TEM）的CEO昨天在第44届J.P. Morgan Healthcare 会议上讲TEM的AI数据飞轮效应： “我们在10年前开始专注于建立这些能够去标识化并被广泛使用的大型多模态数据集（大规模的文本、图像和分子数据）。 直到今天，大约 10 年后，我们积累了一个真正独特的数据集。它涵盖了超过 4500 万名患者，超过 800 万份具有数字化标注的影像记录，我们测序了超过 400 万个样本。 在这个漏斗的最底部，有超过 35 万份记录拥有丰富的分子数据、基因组数据、转录组数据、影像数据和临床数据，这个数据集真正通过规模化运作支撑了我们要今天的绝大部分数据业务。 这种将临床诊断与驱动研究的数据集相结合的理念至今仍然有效。 这两者产生的飞轮效应（Flywheel Effect）非常深远。我们测序的患者越多，收集的数据就越多，这使我们能够提供更多的洞察，进一步增强我们的业务并允许我们收集更多数据。 那个在 6 或 7 年前开始转动的飞轮现在正在全速运转。正如你所看到的，我们不仅仅是在收集一些数据，我们实际上是在收集确定患者是否使用了正确的药物、是否参加了正确的试验，或者如何推进药物发现开发所需的全部数据。” -------------------------- TEM是非常典型的具有强AI数据飞轮效应的公司。

@lufeieth 哥 别奶了，还等着倒车接我上车呢

上周（3月11日），Circle CEO Jeremy Allaire在Canaccord Genuity第六届数字资产虚拟研讨会上做了一场信息量炸裂的主题演讲。 这篇演讲出来第二天我就读了，信息量很大，跟主持人的对话里口语化表述也不少，如果直接贴中文翻译版，我自己读起来都很费劲，找不到重点。 所以我通过Claude MAX Opus4.6 模型对这场分享全部要点的逐条结构化整理，专业表述尽量保留Jeremy的原话表达，并将口语化做了书面处理，提炼了十一个主题、41个要点。 终于完成了一件悬在心里的事。希望能帮大家节省点时间和提供一些关于CRCL的增量信息。

@lufeieth 跟战争有关

如图1，USDC和CRCL不会有人坐庄吧？ 图2，USDC流通量超过前高，达到了792亿。 看过去90天的流通量变化，很漂亮很暴力的V 反。

@lufeieth 同感 Gemini越用越拉

提问：“委内瑞拉新总统是谁？” 几个大模型的回答如图：Gemini、GPT、Claude。 启示： 尽量少用Gemini；不要抄底GOOG。

@frankyluan 目标多少

买入 amd

JUNE 2028. The S&P is down 38% from its highs. Unemployment just printed 10.2%. Private credit is unraveling. Prime mortgages are cracking. AI didn’t disappoint. It exceeded every expectation. What happened?​​​​​​​​​​​​​​​​ citriniresearch.com/p/2028gic

@wdctlllltcdw 这睡眠质量志明做到的

连续90分以上 中间不醒

@Hoyooyoo CRCL怎么看老板

CRCL put，账面赚了7倍！2倍本钱先收回，剩下的继续奔跑。💪

@wufantouzi BTC下跌肯定会对CRCL有影响，午饭为什么在看跌BTC的时间段买入CRCL？这个一直没太想明白。

这波回调现在看都不确定有没有结束 时间上比当初预计的提前太多 太快了 以至于我觉得是牛市插针还会反弹 有些东西真的是上天给你的灵感或者提示 回到8月底我在佑国club群里发的图 微博老粉应该记得： 历史上我凭感觉画过几次 分别是312/519就很离谱 画的时候完全是凭感觉随手画的 只是当时以为是月线后面发现是周线 三个高点 震荡位置 趋势形态 几乎全部对上

@_FORAB 接入的是长桥

又一币圈平台支持买美股了。 老牌加密资管平台 Matrixport，宣布上线美股购买入口，支持美元稳定币 USDT、USDC 入金，且无等待时间。 根据披露，该产品是持牌券商直连，在平台上买卖股票，有底层券商提供撮合，从而享有股票分红、投票权。 目前首期支持标的为美股主板股票、主流 ETF。 由于是美国服务券商，所以该产品不会同步 CRS 税务系统。 补充：Matrixport 是老牌的资管平台，在 2021 年获 Paradigm、IDG、DST 投资。@Matrixport_CN

@IamRamenPanda @AlphaguyTrading 最近USDC流通量也少了不少，这也是价格下跌的原因之一

@AlphaguyTrading 发型规模可以抵消利率降低

稳定币行业的另外一大利好是纳斯达克的全天候交易所，链上交易所，稳定币结算 股票一旦上链，一定需要一个稳定的“数字美元”来做结算和周转，而现在最成熟的就是USDC 不管是纽交所、贝莱德，还是未来各种金融平台，只要大家用 USDC 来结算、做保证金、做流动性，钱就会不断从它这条管道里流过 稳定币是AI之后，另一个高确定性的增长赛道 很多人看到只有币圈人奶 crcl，就看不起这只股票。但现实是只有币圈人他妈懂啊！

妈的，终于把 Chrome 上的 Gemini 开了 录个视频教大家怎么开启 如果你也跟我一样，所有的方法都用了，还是不行，而且你是 macOS 系统的话。 你直接把系统语言改成英文，它重启后就出来了。 首先就是 Chrome 的设置： 1. 把 Gemini 相关的所有开关都打开 2. 然后再把所有的 Chrome 浏览器地区设置成美国 这个流程可以用命令行去完成，也可以用现成的项目（如下面所发）去完成 然后关于谷歌账号和系统的设置： 1. 把你的 Google 账号地区改成美国 2. 把 Google 账号语言设置成英文 3. 刚才我提到的 Google 浏览器也需要改成英文 在 macOS 上，如果想把 Google 浏览器改成英文界面，需要把你整个系统语言都设置成英文