LL🐂🀄️

人的任何一个决策都是基于不完整信息下的相对推理，而决策并非由你的自由意志决定/INTJ/talk is cheap show me PNL

[LL🐂🀄️ @LNudt]

ape $SOL

[Hanson Z @HansonzZhang]

不知道大家知不知道Chris Camillo, 曾经美股社区的平民王者 从2w做到了7000w美金。

他就是特别擅长发现身边的机会 在他第一次看到iphone的时候就被吸引，当时的华尔街没人认为苹果会成功 Chris一意孤行买入苹果

包括亚马逊，他发现亚马逊的AWS有段时间在google上的搜索次数翻了好几倍，于是重仓亚马逊 赚得盆满钵满

很多人在看k线 看现金流，但真正的王者看的都是趋势，不是明面上的趋势，而是暗地里在涌动的资金趋势

[LL🐂🀄️ @LNudt]

@thecryptoskanda 🤯

[Cryptoxiao @cryptoxiao]

是这样的，但是不太习惯开空 没吃到

[MP @MoneyPrinter0x]

i read the whole thing this is genuinely very sharp

im not as active onchain nowadays but even i want to bid, the thesis makes sense

[LL🐂🀄️ @LNudt]

@SwitchesBoard @EasyA_Kickstart 🫡

[Yueya☂️ @yueya_eth]

前段时间在 @virtuals_io 看到 @Litebeam_xyz，没想到币价涨的这么猛。研究完，感觉它切到的是 AI agent 真正落地后绕不开的一层：

service routing

过去 OpenRouter 证明过一个很清晰的逻辑：当模型越来越多，用户不需要自己逐个接入、比价和结算，只需要一个统一入口。

Litebeam 做的是把这套逻辑放到 AI agent 上。
OpenRouter 路由模型，Litebeam 路由 agent 执行任务时需要调用的 microservices。

搜索、数据、代码、支付、SaaS、图像、音频、链上工具，这些能力如果都让每个 agent 自己去接 API、管 key、比价格、控预算，根本不现实。

所以它的核心很简单：
One MCP connection, every microservice.
这个方向也和现在 AI infra 的讨论很一致。

@karpathy 一直强调，AI 的进步会越来越依赖模型周围的 scaffolding、tool use、routing 和 orchestration。

@a16z 也在持续讨论 MCP、context layer、agent identity / permissioning、payment 这些基础设施。

因为 agent 一旦从 chatbot 变成真正执行任务的经济参与者，它需要的不只是大脑，还需要稳定调用外部服务的手脚。

Litebeam 目前做的是一次 MCP 连接，跨服务商实时路由，按价格、延迟、声誉选择更合适的服务，同时接入 Base 上的 USDC 结算、预算控制和链上审计。

它吸引我的地方在于飞轮：
更多 agent 调用
→ 更多真实服务质量数据
→ vendor reputation 更准确
→ routing 效果更好
→ 更多 agent 接入

目前看到 @Base 和 @virtuals_io 都有公开支持，也和 @reppo 、@officialbunnyos 有合作，项目方 metrics 里 routed calls、active agents 和 live services 都在增长。

另外，在FOMO trending 第 5 的交易员 @theveeman 前几天因为 $ANSEM 赢麻了的那个男人）也买入了 $LBM，并写了买入理由。

Litebeam 也在引入 @daytonaio  作为 AI agent security layer。Daytona 在 review 了 Litebeam 的技术之后，提供了 grant + partnership 支持，这对一个早期 first mover infra 来说是很重要的外部验证。

$LBM ：0x15b15fa54b629c634958e8bd639b2fc8af654974

还很早，波动肯定会大。
但如果 agent economy 真的起来，服务发现、路由、支付、声誉层会是很重要的底层位置。

这个方向的项目我会持续跟踪。

[Ignas | DeFi @DefiIgnas]

How does one prepare for the next bull if you are still bullish on crypto?

When I started actively writing on Twitter four years ago, things were different.

Like now, it was a bear market, but back then, CT believed with certainty that a new bull run would arrive.

Notably, a bull run in altcoins.

BTC was boring, and the upside seemed limited. Ethereum's modular roadmap seemed exciting, with potential sharding not in the trash bin just yet.

Solana was written off (before a monolithic vs. modular narrative reemerged with the $SOL revival).

To build an edge, one believed in learning about new technical crypto innovations before anyone else so they could ride a narrative when the general crowd caught up.

No surprise that technical breakdown posts of narratives did amazingly well on X:

Proof of humanity, soulbound tokens, quadratic voting... anyone care to learn lol?

Maker initiated Endgame Plan for Unbiased World Currency. They did the exact opposite of it (to be fair, pro crypto regulation changed their roadmap).

VCs were heavily deploying, ready to replay the previous bull run playbook.

Even KOLs were playing VCs, getting in at lower valuations than VCs in exchange for a few posts on X.

Liquidity mining in liquid tokens was still popular. Now replaced by points.

LP mining is basically dead. Dead as #hashtags on X.

Overall, reviewing my old posts, it's clear that this bear market is fundamentally different.

CT is unwilling to learn anything technical. Nobody reads or cares about white papers.

In turn, projects don't even have or share roadmaps: announcements come straight from the sky and bag holders don't scream "can devs do anything" anymore.

Price decline is seen as a natural state of affairs. No one to blame but you (or CZ).

Discords and projects' TG channels are dead.

We believed that crypto-native innovation would bring us mass adoption the moment that a "killer crypto app" appears.

Indeed! This turned out to be true: prediction markets, onchain perps, and RWA/stock tokenization are those killer apps.

Even more bullish is that revenue, especially that which goes to holders, matters more than empty roadmaps.

Yet besides $HYPE (and maybe $ZEC), there were no true token success stories this cycle ($SOL revival could be one more).

The high hopes for altcoin season were exploited by memecoin overprinting and low-float, high-FDV launches.

If reading whitepapers or technical threads on CT is not your thing, the 'research on your own' part still matters.

Simply daily checking dashboards on Defillama, Blockworks, Artemis, etc., for high-revenue projects will put you ahead, as you can whitelist those tokens for more bullish times.

I'm also not giving up on new TGEs and airdrops: following new projects that manage to raise decent amounts from VCs in this terrible market must have something up their sleeves.

I bet launching at high FDVs in a bullish market will make a comeback (thus juicy airdrops).

(Praying for Variational's $VAR stimmy).

Then don't forget to try out new apps: anything trending on CT is worth five minutes of your time:

Download the app, put a few USD in, and see if it clicks with your degen senses.

Finally, regulation is a big variable this time around:

The Clarity Act in the US (if passed) could give revival to some select altcoins while others might suffer.

Equally, the EU's MiCA regulation for CEXs is already changing the CEX competition: if Binance is out, then it gives someone else an advantage.

(Not gonna lie, I like Backpack's $BP as a proxy trade for the EU regulation plus tokenized stocks on the Solana narrative.)

In any case, researching isn't dead.

We are just researching real products now instead of trading on pure whitepapers and technical narratives.

Which should make things even easier this time as very few actually match the criteria.

[Mikli @CryptoMikli]

Threadguy says Solana’s next run will not be driven by memecoins

"My take is that the bull case for solana:So11111111111111111111111111111111111111112 making a new ATH cannot be the resurgence of memecoins. It can help, it can add some market cap, but it’s not sustainable"

"Solana has been pushing tokenized stocks, perps, and RWAs really hard. If there’s going to be a narrative to rip the coin back to ATHs, it has to be that. It can’t be memecoins"

"I don’t think memes will ever die, they’ll get hot then get quiet over and over again. In order for me to bid Solana, the coin, I need to see up-and-to-the-right tokenized stocks, or perps volume, or ETF inflows, at least for me"

[LL🐂🀄️ @LNudt]

@Cryptor_dot @MCGlive @SwitchesBoard Is that true? What has he done before?

[qinbafrank @qinbafrank]

谁能拿到海力士的HBM4测试设备订单？最有可能拿到这笔订单的主要是安德测试和Teradyne（泰瑞达）两者将占据HBM4测试仪（核心ATE自动化测试设备）订单的大头、Onto可能拿到订单中的非ATE设备商份额、韩国本土的UniTest等公司可能分得部分补充设备（如烧录/老化测试仪）、探针卡等配套可能由FormFactor等公司参与。

HBM4测试对速度、并行度、功耗和系统级验证要求极高，技术门槛高，供应链相对集中，属于高价值、高并行、高速内存测试设备。
1）Advantest（安德测试）：HBM测试设备市场领导者（份额约60-70%），长期为SK海力士、三星、Micron等内存厂商供应HBM/DRAM测试系统（如T5503HS2系列，支持高并行HBM测试）。SK海力士为保障HBM良率，已在使用其设备。AI/HBM需求推动下，其内存测试业务持续受益，是本次订单最稳的候选。

2）Teradyne（泰瑞达）：内存测试领域的另一巨头，2025年推出专为HBM设计的Magnum 7H新一代测试平台，支持HBM3/HBM4高速测试（数据率达4.5 Gbps+），已开始在主要HBM厂商量产出货。Teradyne与SK海力士在HBM测试解决方案上有合作记录（包括获奖认可），具备直接竞争力，尤其在新一代HBM4平台上。

3）Onto是半导体工艺控制领域的领先厂商，核心产品是Dragonfly系列（如Dragonfly G3/G5）高分辨率2D/3D检测与量测系统，专攻先进封装（Advanced Packaging）、HBM堆叠、凸块（bump）量测、TSV深度、RDL高度、亚表面缺陷检测等。

Onto的Dragonfly系统已被多家逻辑与内存厂商用于HBM + GPU的异构集成（AI芯片封装），支持亚微米级检测和3D量测。Onto是本次订单中最值得关注的非ATE设备商之一，大概率能拿到一定份额，尤其如果SK海力士在HBM4量产验证中采用其新系统。

4）韩国本土厂商（如UniTest）：可能拿到部分订单。UniTest已完成SK海力士HBM4烧录测试仪（burn-in tester）的量产认证，适合工厂内的老化/可靠性测试环节。韩国设备商在成本、交付和服务响应上有优势，可能分得补充或配套设备。

配套设备厂商：
1）探针卡（Probe Card）：FormFactor（美国）是全球领先的MEMS探针卡供应商，广泛用于DRAM/HBM晶圆测试（高针数、高速需求），很可能参与配套。Technoprobe等也有机会。

2）晶圆检测/量测设备：KLA、Applied Materials、东京电子（Tokyo Electron）等，可能涉及光学/缺陷检测部分（广义“检测设备”范畴）。

SK海力士此前与供应商共同开发HBM4系统级测试（SLT/安装测试）设备，部分能力已内部化，但工厂量产仍需外部标准ATE支持。

订单分配逻辑
1）核心HBM4测试仪（高价值部分）：Advantest + Teradyne大概率拿下大部分（可能70%以上），因为它们在技术、产能和历史供货上领先。设备零部件交期长（已超1年），双方已提前口头协商明年交付量。

2）补充/配套设备：UniTest等韩国厂商、探针卡/探针系统厂商有机会。

总台数约200台并非单一供应商独占，而是多家分担（不同类型设备：ATE、烧录仪、探针系统、检测仪等）。

还是之前推文聊到的，测试与先进封装是容易被低估的环节。过去市场喜欢把封测看成低毛利、低过去市场喜欢把封测看成低毛利、低壁垒后周期资产，但AI/HBM改变了这一点。先进封装越来越接近前道制造，设备壁垒和客户认证壁垒提高

[Fiona ❤️& ✌️ @nft_hu]

CXL 的意义不在于替代 GPU memory，而在于把 memory 变成可调度资源，从而打开 AI 基础设施的第二增长曲线。

核心是通过解耦 Prefill 与 Decode，将计算与内存瓶颈分离，从而把 GPU 利用率最大化。

CXL 4.0 是高速互连技术的重要跃进。其降低延迟到纳秒级别，数据速率提升至 128 GT/s，引入了native x2 width的概念，以支持平台更高的扇出能力，从而提升链路距离，相比 CXL 3.0 实现更长的信号传输。

[LL🐂🀄️ @LNudt]

@xtony1314 她老婆的亲戚是朱晓虎的那个欧神吗

[Fiona ❤️& ✌️ @nft_hu]

最近半导体设备的资金热度和讨论热度都很高，包括我很钦佩的交易员 @labubu_trader 也多次提到。

本文梳理一下最近我看到的设备厂商名字和他们的基础业务。排名前后按照个人喜好排序，不构成投资建议。

首先，我们要知道设备厂商被炒作的逻辑是内存厂商今年赚太多，无论是本身的商业选择还是政府推动的，扩厂都是必然选择，且高优先级。这首先利好的就是设备厂商。

而我先把这些设备厂商分为两大类：
🌟原材料/前道核心设备（SUMCO、Kokusai、TEL、AMAT、LRCX、KLAC、ONTO）。
🌟设备零部件和关键子系统（ICHR、UCTT）。

[LL🐂🀄️ @LNudt]

@leto_bao DC群拉一下老师，小红书上链接失效了

[LL🐂🀄️ @LNudt]

@Soratechsuki tmd和抽风一样，这谁拿的住

[Ansem 🐂🀄️ @blknoiz06]

iq test

[Shillz @Shillz_Official]

@Schoen_xyz @blknoiz06 Don’t worry he’s launching a campaign on Shillz later this week

[LL🐂🀄️ @LNudt]

@joyfenlie @WangNextDoor2 未必不能到啊

[Ansem 🐂🀄️ @blknoiz06]

black bull mode 🐂🀄️

[Ansem 🐂🀄️ @blknoiz06]

vibe shift