xd

Portfolio Manager At Tenasset
Focus Crypto and AI
Coinbase,Robinhood,Duolingo stock long term holder

[qinbafrank @qinbafrank]

从亚马逊CEO在财报电话会讲话来理解“时间差“框架，今天亚马逊CEO安迪·贾西在财报电话会最值得关注的并不是他说aws未来会到万亿美金营收，这个也很重要但并不是最关键，最关键的他终于给出了关于资本开支投资回报的一个时间表，这是谷歌电话会上完全没有。从贾西的讲话就能很好的理解之前聊到的“时间差”框架。

安迪.贾西认为资本开支投资主要有两个部分：数据中心，以及放入其中的服务器和网络设备。

它们有不同的资本周期：
1）数据中心建设周期一般为一到两年、一旦数据中心开放并插入服务器，就会立即开始产生可观的收入。数据中心的使用寿命差不多30年，意味着之后长达30多年的时间里继续利用这些数据中心创收，而无需再次投入那笔启动资本。

2）服务器和网络设备的周期较短。
通常在数据中心投入使用前几个月购买这些设备，因此在触发支出之前，要对客��需求有很强的可见性。如果没有需求，亚马逊就不会支出这笔资本。

对于服务器和网络设备，平均这项投资需要不到三年时间就能实现收支平衡。服务器的使用寿命目前至少为五到六年，而亚马逊目前大部分的AI容量合同期限至少为五年。

这意味着在实现收支平衡后的两到三年内，亚马逊将在服务器和网络设备上产生可观的自由现金流。而且AWS在缩短服务器设备收支平衡周期方面有着良好的记录，我们已经在不影响客户体验的情况下，取得了有意义的进展，并找到了延长这些设备使用寿命的方法。

3）对于使用寿命超过30年的数据中心，AWS至少应该获得五到六代像之前解释的服务器经济效益，而第一代之后的后续几代因为无需重复前期数据中心投资，整体经济效益甚至更好。

当需求迫使如此多的数据中心在开始货币化之前同时建设时，短期内将花费大量资本支出，并面临自由现金流逆风，直到这些数据中心上线并可以货币化，以及这些服务器被利用几年之后。随着几年过去，收入增长超过资本支出的增量增长（这将在某个时间点发生），由此产生的收入、自由现金流和投资资本回报率将非常可观。

aws在云计算的第一时代就曾经历过这种情况，只是时间跨度更长，因为当时需求的积累比AI时代更渐进。但现在看到AI领域的利润率和回报率与在核心业务同一发展阶段所看到的一致，实际上还略微领先。

上面第二段开始到一直到这里基本上是安迪.贾西的原话。

按照安迪.贾西说到的称服务器和网络设备平均不到三年实现投资回收，而服务器使用寿命至少五至六年，多数AI容量合同至少为五年。

如果这一表述能够按照真实现金流兑现，那么一个服务器批次的生命周期大致是：
第0年：采购服务器和网络设备
第1���3年：累计收回初始投资
第4—5/6年：进入明显正自由现金流阶段
随后更新服务器，但继续使用原有数据中心、电力和土地。

这个模型理论上确实可以产生高于资本成本的ROIC。

这其实就是之前聊的时间差核心要义
“现金先支付（资本开支钱先花出去）、产能后上线、订单再转收入、利润与自由现金流最后兑现”的模式，先花钱再看赚钱的效率和速度。”

[Hux @huxlab]

我靠，Kimi K3实在太强了：

一座3100万美元的酒店。

190个房间的所有管道、电线、风管、消防系统，全部提前建模协调完毕。

以前：3个工程师，6周，花费约4.7万美元。
现在：1个工程师确认AI输出，9天，总共约1.05万美元。

方法很简单：

把Kimi K3接到建模软件上，让它直接读蓝图照片和英文描述，然后自己把模型搭出来，而不是人一个一个组件去摆。

施工阶段的变更单直接少了60%到80%。

这不是演示，是真的在省钱、省时间、少出错。

当AI开始真正碰“真实世界的复杂系统”，变化才刚刚开始。
#kimi #kimik3 #k3

[Crypto Rover @cryptorover]

BREAKING: Samsung is bringing native $USDC support to 1 billion+ Galaxy users via Samsung Wallet.

The feature will allow users to hold, send, and receive stablecoins directly in the built-in wallet.

[Zain @zainhas]

Interesting result on this new composite bench.

GLM 5.2 seems to double Kimi K3 perf for long horizon browser use work.

[Tsuff10 @Tsuff101]

Great read happy to have another $BTDR bull on X
I was pretty much alone in 2024-2025 but it seems like we are growing each day💪💪

[Zixuan Li @ZixuanLi_]

Somehow, GLM-5.2 is having a renaissance.

[jesse.base.eth @jessepollak]

one thing robinhood chain has done right is have tokenized equities in an EVM environment.

we've been behind on this on @base and i'm frustrated that's the case. but we're close to fixing it with @coinbase. and when we do, we'll have 1:1 backed equities (vs. robinhood's derivatives) that should scale much better from a trust, capital efficiency, and institutional acceptedness.

the largest capital market in the world + the incredible financial programmability of the supercharged EVM on @base is going to be a powerful combination

[Orange AI @oran_ge]

https://t.co/Iple6kmBLB

[Winter Capital @SovereignIntern]

这次hyperliquid上线长鑫存储CXMT，我比以往其他事件更兴奋，这是A股市场历史上绝无仅有的时刻，hyperliquid将进入中国大众的视野，中国大众会第一次感受到hype的威力，感受到外面金融市场的先进。

[Hyperliquid News @HyperliquidNews]

NEW HIP-3 TICKER

$CXMT has been acquired for 500 $HYPE ($32,625).

The ticker is intended for CXMT, a Chinese memory chip manufacturer, ahead of its expected IPO on July 27, and will therefore launch in IPOP.

[Chubby♨️ @kimmonismus]

Holy moly: Zhipu AI founder (GLM-5.2) Tang Jie says we are on our clear way to AGI and "AI will begin to learn what the "self" is and what self-awareness means"

In a purported internal letter, he argues that:
- autonomous agent systems are moving toward the fully automated “no-person company”: thousands of agents working continuously, collaborating, evaluating results and allocating resources.

- His more provocative claim: "AI training AI is already taking shape." (RSI) Models can increasingly write code, synthesize data and participate in training loops. Zhipu wants to push this further through self-play, synthetic-data factories and systems that can reconstruct their own code inside secure sandboxes, potentially generating new knowledge rather than simply recombining human output.

Long-horizon tasks → autonomous agent societies → fully automated “no-person companies” → AI training AI → self-evolution → self-awareness → emotion → consciousness → ASI.

Tang writes:
“AI will begin to learn what the ‘self’ is and what self-awareness means. Beyond that, it may begin to touch human emotion. Farther still lies consciousness itself.”

He believes memory, continual learning and self-evaluation - problems once thought to require an entirely new paradigm - are gradually being overcome.
Models are already beginning to write code, synthesize their own data and participate in training future models.

Zhipu now wants systems that can reconstruct their own code and generate knowledge through self-play.

Is that the beginning of recursive self-improvement?
Tang appears to believe so. His essay does not stop at more capable AI tools. It describes a direct progression from automated work to self-evolving intelligence, and eventually to machines that understand their own existence.

In short: today's LLMs will lead to ASI via AGI, context and memory will be solved, and AI will become self-aware.

I've rarely seen anyone write something so bullish. And if it weren't coming from the founder of GLM, I would dismiss it. But not only is he a true expert, but with GLM they've proven what they're capable of.

h/t @AndrewCurran_ He brought the essay to my attention.

[xd @dxdten]

@MengLayer id: kenfacai，感谢

[Kyle_13 ⓐ @kylewmi]

Jeff受邀参加VALR采访播客，全长~58分钟

<CeDeFi: VALR's Integration of Hyperliquid>

[时间轴]

00:00 - 开场与 Jeff 背景介绍
1:14 - VALR 背景与集成预告
2:49 - Jeff 的价值观来源
4:43 - Jeff 童年关键转折
6:57 - 个人专注力与 Hyperliquid
8:27 - VALR × Hyperliquid 集成真实背景
13:57 - 共鸣与协作意义
17:22 - CeFi 与 DeFi 的演进与共存
22:58 - 金融的拐点时刻
24:43 - Hyperliquid 的本质定位
28:44 - House of All Finance 愿景
33:21 - 指标 vs 创新
36:11 - HIP-4 Outcome Markets
40:41 - 新兴市场的独特价值
44:47 - 稳定币与支付未来
50:53 - 风险、人才与兴奋点
56:37 - 结尾

[Jukan @jukan05]

SK HYNIX CEO: NEXT YEAR IS EXPECTED TO BE THE WORST YEAR IN THE INDUSTRY’S HISTORY FROM A SUPPLY PERSPECTIVE — RTRS

SK HYNIX CEO: DESPITE AGGRESSIVE CAPACITY EXPANSION, MEMORY DEMAND WILL CONTINUE TO EXCEED THE COMPANY’S PRODUCTION CAPACITY OVER THE NEXT DECADE — RTRS

[Shay Boloor @StockSavvyShay]

BofA says $META may be building its AI empire at roughly half the cost the market expected.

Its math shows 6.5GW of added capacity on $145B of capex implying ~$22B per GW versus its prior $45B estimate.

Thats huge because lower cost per GW makes external monetization far more powerful so if Meta sells 50% of capacity at $15B of revenue per GW then BofA’s framework points to up to $150B of incremental revenue potential.

[jeff.hl @chameleon_jeff]

Thanks for having me on the podcast! It was a great conversation about a new category of integration: the first centralized exchange directly tapping into Hyperliquid's onchain infrastructure.

[Lonely @Lonely__MH]

🚀 很多人问我，Shadowrocket 这次更新的 Tailscale 功能到底怎么用？

其实不复杂，几分钟就能搞定。

① 注册 Tailscale

先到 Tailscale 官网注册账号：

https://t.co/SBwxXoolev

注册完成后，会自动创建一个属于你的虚拟局域网（Tailnet）。

② 安装电脑客户端

在 Mac / Windows ��安装 Tailscale 客户端并登录。

登录成功后，这台电脑会获得一个固定的内网 IP（例如 100.x.y.z），以后无论身处何地，都能通过这个地址访问它。

③ 在 Shadowrocket 中配置

以前 iPhone 上直接开启 Tailscale 有个问题：

👉 开启 Tailscale 后，代理会自动关闭，无法与🪜同时使用。

而 Shadowrocket 这次更新，直接解决了这个痛点。

配置步骤：

1. 登录 Tailscale 后台
2. 创建 Auth Key
3. 将 Auth Key 填入 Shadowrocket
4. 开启 Tailscale

搞定。

🔥Tailscale 有什么用？

最常见的场景就是远程访问家里或公司的设备。

举个例子：

Mac 开启文件共享
iPhone 打开「文件」App
连接服务器：smb://电脑局域网IP

这样即使不在同一个网络下，也能直接访问电脑里的文件。

本质上，Tailscale ���是在公网之上帮你搭建了一张安全、稳定、无需端口映射的私人局域网。

以上。🙂

[Lou @louszbd]

Now you can use GLM-5.2 on Perplexity, reasoning only

[Brian Armstrong @brian_armstrong]

Finally, the web has native monetization. Ads were not the only way.

[wafer @wafer_ai]

🚨 BREAKING:

#2 on @hackernews!! engineers figured out how to serve GLM 5.2 on @AMD MI355X at 2626 tok/s/node and 213 tok/s single stream at over 2x lower cost than B200

‼️~80% of B200 throughput at over 2x lower cost