斯马克

here is the thing

[斯马克 @sssmarkkk]

Great, in the era of information explosion we need such articles to focus on the knowledge that can be reused across many years (the essential parts) instead of learning every new stuff.

[川处安 @leoshen0]

https://t.co/Q9h9ISi1L2

[斯马克 @sssmarkkk]

So in the future, everyone will become the lawmaker of their own country.

[gKev1n @GKev1n]

为什么你用 AI Agent 越用越累？因为你把自己活成了系统最大的瓶颈。

SpaceX 工程师 Lauren Tan 这场 1 小时的分享，直接戳破了大多数人的 Agent 幻想：

“连单个 Agent 的输出你都无��信任，就别妄想调动上百个 Agent 协同作战。”

如果你没有把“验证机制（Verification）”沉淀成工具与技能，那你就是那个肉眼找 Bug 的人工瓶颈。

从“每行代码都要人肉盯盘”，到“单月自动化合并 1,000 个 PR”，她的核心演进路径只有一条：
人肉盯防 ➔ 规则验证 ➔ 技能封装（Skills）➔ 评测体系（Evals）➔ 自动合并（Auto-merge）

对 Agent 的信任从来不是一种主观感觉，而是你必须亲手搭建的代码基础设施。

干货密度吊打绝大多数天价编程课，强烈建议 Bookmark 收藏细看。👇

[斯马克 @sssmarkkk]

My bachelor’s professor told me if you want to understand a topic, just read 100 papers. Same logic here: old strategy, new case.

[EnHeng嗯哼.Ai @EnHeng456]

花了四个多小时看完@justinsuntron ✖️@AlanShao111 整场访谈，网友总结的这些观点，我觉得很值得年轻人认真思考（https://t.co/NOe2PoPn9a 整合了这场播客的���字稿）

未来生活要全面AI化，多与Claude、ChatGPT交流，用AI梳理人生方向、分析问题、辅助决策，通过API接入全球不同的大模型，利用模型之间竞争白嫖大模型使用

健康管理也要成为人生的长期投资。寿命即财富，身体才是最大的本金。建议每天抽出1至2小时运动，主动管理健康。（我自己一年时间从200斤减到146斤，同时尽量保住了肌肉）

眼光要放到全球，学会赚全世界的钱，摆脱单一市场的内卷。保持地理位置和事业上的机动性，逐水草而居，哪里有机会就去哪里

做任何重大决定之前，先充分了解信息，再进行选择。比如买房，先看100套房子，��整个市场摸清楚，再决定买哪一套。信息越充分，决策失误的概率往往越低

同时要理解几个长期趋势：传统文化正在数字化，Crypto仍然值得长期关注，AI会越来越深入地参与人类决策。评价一件事要看事实和逻辑，不能只凭一个人的身份、立场或动机判断对错

我觉得整场访谈最值得记住的，其实是最后一个观点：不断升级认知，把世界当成一个持续更新的版本，保持学习，持续适应变化

未来真正重要的能力，是面对一个不断变化的世界，依然能够快速理解新事物、调整自己，并抓住新的机会

[斯马克 @sssmarkkk]

@leo_xiaolei I think Chinese related company wont work right？

[斯马克 @sssmarkkk]

3/3
So at the paradigm level, it may not be an entirely new loop. What feels more interesting is how it redefines what the intermediate simulator can be, and how much it can do.
In a sense, recent World Model works are revisiting problems previously limited by simulator fidelity.

[斯马克 @sssmarkkk]

CurrentWorld-0 feels like a learned/generative version of Real2Sim2Real: Real → Sim → rollout/eval/train → Real. The loop is familiar; the novelty is redefining “Sim” as a controllable world model with multi-view, force/tactile, and intervention.

[斯马克 @sssmarkkk]

2/3
The main difference is that the “Sim” is no longer primarily built from explicit physics, but replaced by a learned world model, with cross-embodiment, multi-view, force-tactile signals, human intervention, and branching rollouts integrated into the same framework.

[斯马克 @sssmarkkk]

1/3
After watching CurrentWorld-0, my first impression is that it feels like a learned / generative version of Real2Sim2Real.

The overall loop is quite familiar:

Real → Sim → rollout / evaluation / training → Real

[斯马克 @sssmarkkk]

@Tasihi_89 怎么断点续学以及换个非mac电脑还可以用吗

[斯马克 @sssmarkkk]

@himanshustwts what did you use to make this

[斯马克 @sssmarkkk]

@heygurisingh Can the model be sim2real？Hope it
does

[斯马克 @sssmarkkk]

@realDonaldTrump Trump hit China with a stick.
China hit back.
Trump thought, “Hmm, that didn’t work.”

Then he tried the same stick on his dog—
and realized, “Oh, this works great!”

So he went back to China…
for another stick.

[斯马克 @sssmarkkk]

@cb_doge @elonmusk ready for X-sora training😁？

[斯马克 @sssmarkkk]

@100trillionUSD The conclusion of this chart is to buy at the end of the blue phase, hold during the red phase, and sell at the turning point from red to yellow. At this moment, we are in the early stages of the red phase. Is that true？

[斯马克 @sssmarkkk]

@pengchujin u3路由器老毛子固件可以安装openclash吗

[斯马克 @sssmarkkk]

@CryptoDoggyCN @JimmyKudo0 既然你都办了香港的银行卡，那应该可以用香港的币安直接c2c成港币现金把。

[斯马克 @sssmarkkk]

@100trillionUSD is btc bull also following the usa election?

[斯马克 @sssmarkkk]

@elonmusk 翻译成英文：实现我的想法

[斯马克 @sssmarkkk]

@OGdukeneurosurg is it just a idea demo or real and successful application already?

[斯马克 @sssmarkkk]

@ain9artSD how to make it