半沐
469 Posts
半沐
慢就是快，相信复利的力量 币安100%返佣邀请码: BM168
Open Trade
High-Frequency Trader
7.9 Years
11 Following
123 Followers
267 Liked
Posts
Portfolio
All
Quotes
It’s all Alibaba’s investment—what a mess, didn’t expect that, did you? Pull it out and scare you to death.
链研社lianyanshe
·
--
Alibaba has also been lucky recently: Zhongxin has a floating profit of 130 billion yuan. Today, Kimi K3 is making the charts, and Alibaba holds 36% of the shares. Part of the investment also involves participating in it in the form of providing compute power services.
If Kimi were listed in Hong Kong, with limited shares available for trading, Alibaba, Meituan, Tencent, and Xiaohongshu have all invested. Its market cap might exceed Zhipu. Then the money Alibaba earns would be more than what it earned on Changxin—just from these two, it could make 20% of Alibaba’s market value.
Source: @lianyanshe on X
#Crypto #Web3
If Kimi were listed in Hong Kong, with limited shares available for trading, Alibaba, Meituan, Tencent, and Xiaohongshu have all invested. Its market cap might exceed Zhipu. Then the money Alibaba earns would be more than what it earned on Changxin—just from these two, it could make 20% of Alibaba’s market value.
Source: @lianyanshe on X
#Crypto #Web3
kimi.k3 will be fully open sourced on July 27. Yang Zhilin is awesome, Alibaba is awesome $BABA
See translation
杨植麟: 很多天才都管我叫天才
半沐
·
--
月之暗面创始人杨植麟的成就
一、学术层面：奠定全球大模型长文本底层基础（含金量最高的长期资产）
1. Transformer-XL（2018，共同一作）
- 核心突破：解决原生Transformer固定上下文窗口、无法捕捉长距离文本依赖的致命缺陷；引入片段循环记忆机制。
- 行业意义：当今所有长上下文大模型的技术源头之一，Claude、Kimi、GPT系列长文本优化方案大量借鉴该思路。没有这项工作，百万Token长文本模型几乎无从谈起，也是后来Kimi主打超长文档的理论根基。
2. XLNet（2019，第一作者，谷歌Brain联合发布）
- 核心突破：解决BERT掩码预训练存在的信息偏差，采用排列语言建模；在20项主流NLP评测全面超越BERT。
- 历史地位：被业内公认为BERT之后预训练模型最重要进展之一，入选NeurIPS Oral；论文引用数万次，是华人学者在基础NLP领域影响力最强的工作之一。
3. 其他学术贡献
- 和图灵奖得主Bengio合作构建HotpotQA多轮推理问答数据集；
- CMU博士仅4年毕业（常规学制6年），师从苹果前AI负责人Salakhutdinov；先后在Google Brain、Meta FAIR实习，参与早期Gemini、Bard预研；
- 国内35岁以下NLP领域论文总引用长期位居第一梯队。
一、学术层面：奠定全球大模型长文本底层基础（含金量最高的长期资产）
1. Transformer-XL（2018，共同一作）
- 核心突破：解决原生Transformer固定上下文窗口、无法捕捉长距离文本依赖的致命缺陷；引入片段循环记忆机制。
- 行业意义：当今所有长上下文大模型的技术源头之一，Claude、Kimi、GPT系列长文本优化方案大量借鉴该思路。没有这项工作，百万Token长文本模型几乎无从谈起，也是后来Kimi主打超长文档的理论根基。
2. XLNet（2019，第一作者，谷歌Brain联合发布）
- 核心突破：解决BERT掩码预训练存在的信息偏差，采用排列语言建模；在20项主流NLP评测全面超越BERT。
- 历史地位：被业内公认为BERT之后预训练模型最重要进展之一，入选NeurIPS Oral；论文引用数万次，是华人学者在基础NLP领域影响力最强的工作之一。
3. 其他学术贡献
- 和图灵奖得主Bengio合作构建HotpotQA多轮推理问答数据集；
- CMU博士仅4年毕业（常规学制6年），师从苹果前AI负责人Salakhutdinov；先后在Google Brain、Meta FAIR实习，参与早期Gemini、Bard预研；
- 国内35岁以下NLP领域论文总引用长期位居第一梯队。
See translation
天才只不过是见他的门槛
半沐
·
--
月之暗面创始人杨植麟的成就
一、学术层面：奠定全球大模型长文本底层基础（含金量最高的长期资产）
1. Transformer-XL（2018，共同一作）
- 核心突破：解决原生Transformer固定上下文窗口、无法捕捉长距离文本依赖的致命缺陷；引入片段循环记忆机制。
- 行业意义：当今所有长上下文大模型的技术源头之一，Claude、Kimi、GPT系列长文本优化方案大量借鉴该思路。没有这项工作，百万Token长文本模型几乎无从谈起，也是后来Kimi主打超长文档的理论根基。
2. XLNet（2019，第一作者，谷歌Brain联合发布）
- 核心突破：解决BERT掩码预训练存在的信息偏差，采用排列语言建模；在20项主流NLP评测全面超越BERT。
- 历史地位：被业内公认为BERT之后预训练模型最重要进展之一，入选NeurIPS Oral；论文引用数万次，是华人学者在基础NLP领域影响力最强的工作之一。
3. 其他学术贡献
- 和图灵奖得主Bengio合作构建HotpotQA多轮推理问答数据集；
- CMU博士仅4年毕业（常规学制6年），师从苹果前AI负责人Salakhutdinov；先后在Google Brain、Meta FAIR实习，参与早期Gemini、Bard预研；
- 国内35岁以下NLP领域论文总引用长期位居第一梯队。
一、学术层面：奠定全球大模型长文本底层基础（含金量最高的长期资产）
1. Transformer-XL（2018，共同一作）
- 核心突破：解决原生Transformer固定上下文窗口、无法捕捉长距离文本依赖的致命缺陷；引入片段循环记忆机制。
- 行业意义：当今所有长上下文大模型的技术源头之一，Claude、Kimi、GPT系列长文本优化方案大量借鉴该思路。没有这项工作，百万Token长文本模型几乎无从谈起，也是后来Kimi主打超长文档的理论根基。
2. XLNet（2019，第一作者，谷歌Brain联合发布）
- 核心突破：解决BERT掩码预训练存在的信息偏差，采用排列语言建模；在20项主流NLP评测全面超越BERT。
- 历史地位：被业内公认为BERT之后预训练模型最重要进展之一，入选NeurIPS Oral；论文引用数万次，是华人学者在基础NLP领域影响力最强的工作之一。
3. 其他学术贡献
- 和图灵奖得主Bengio合作构建HotpotQA多轮推理问答数据集；
- CMU博士仅4年毕业（常规学制6年），师从苹果前AI负责人Salakhutdinov；先后在Google Brain、Meta FAIR实习，参与早期Gemini、Bard预研；
- 国内35岁以下NLP领域论文总引用长期位居第一梯队。
See translation
月之暗面创始人杨植麟的成就
一、学术层面：奠定全球大模型长文本底层基础（含金量最高的长期资产）
1. Transformer-XL（2018，共同一作）
- 核心突破：解决原生Transformer固定上下文窗口、无法捕捉长距离文本依赖的致命缺陷；引入片段循环记忆机制。
- 行业意义：当今所有长上下文大模型的技术源头之一，Claude、Kimi、GPT系列长文本优化方案大量借鉴该思路。没有这项工作，百万Token长文本模型几乎无从谈起，也是后来Kimi主打超长文档的理论根基。
2. XLNet（2019，第一作者，谷歌Brain联合发布）
- 核心突破：解决BERT掩码预训练存在的信息偏差，采用排列语言建模；在20项主流NLP评测全面超越BERT。
- 历史地位：被业内公认为BERT之后预训练模型最重要进展之一，入选NeurIPS Oral；论文引用数万次，是华人学者在基础NLP领域影响力最强的工作之一。
3. 其他学术贡献
- 和图灵奖得主Bengio合作构建HotpotQA多轮推理问答数据集；
- CMU博士仅4年毕业（常规学制6年），师从苹果前AI负责人Salakhutdinov；先后在Google Brain、Meta FAIR实习，参与早期Gemini、Bard预研；
- 国内35岁以下NLP领域论文总引用长期位居第一梯队。
一、学术层面：奠定全球大模型长文本底层基础（含金量最高的长期资产）
1. Transformer-XL（2018，共同一作）
- 核心突破：解决原生Transformer固定上下文窗口、无法捕捉长距离文本依赖的致命缺陷；引入片段循环记忆机制。
- 行业意义：当今所有长上下文大模型的技术源头之一，Claude、Kimi、GPT系列长文本优化方案大量借鉴该思路。没有这项工作，百万Token长文本模型几乎无从谈起，也是后来Kimi主打超长文档的理论根基。
2. XLNet（2019，第一作者，谷歌Brain联合发布）
- 核心突破：解决BERT掩码预训练存在的信息偏差，采用排列语言建模；在20项主流NLP评测全面超越BERT。
- 历史地位：被业内公认为BERT之后预训练模型最重要进展之一，入选NeurIPS Oral；论文引用数万次，是华人学者在基础NLP领域影响力最强的工作之一。
3. 其他学术贡献
- 和图灵奖得主Bengio合作构建HotpotQA多轮推理问答数据集；
- CMU博士仅4年毕业（常规学制6年），师从苹果前AI负责人Salakhutdinov；先后在Google Brain、Meta FAIR实习，参与早期Gemini、Bard预研；
- 国内35岁以下NLP领域论文总引用长期位居第一梯队。
半沐
·
--
#kimi k3#
Moon’s Dark Side co-founder, will release open-source Kimi k3. Wow— the world’s number one large language model is going open-source. How does this not kill OpenAI and Claude Code’s closed-source offerings? $BABA
Moon’s Dark Side co-founder, will release open-source Kimi k3. Wow— the world’s number one large language model is going open-source. How does this not kill OpenAI and Claude Code’s closed-source offerings? $BABA
A group of people still haven’t figured out the value of Kimi K3. Its frontend capabilities surpass Fable 5—now it’s number one in the world. Its backend capabilities are only behind Fable 5. GPT 5.6.Sol runs at the highest configuration, and the subscription price is only 5%. Currently, OpenAI is valued at $1.2 trillion. If Moonshot’s Dark Side IPOs, it could directly approach $1 trillion. Why? Because its founder is Yang Zhilin, a top-tier talent whom AI leaders around the world all regard as a benchmark—first author of Transformer-XL, first author of XLNet—someone even more impressive than Sam Altman or Dario; an outstanding figure.
And the largest shareholder of Moonshot’s Dark Side is Alibaba, holding 36% of the shares.
And the largest shareholder of Moonshot’s Dark Side is Alibaba, holding 36% of the shares.
半沐
·
--
#kimi k3#
Moon’s Dark Side co-founder, will release open-source Kimi k3. Wow— the world’s number one large language model is going open-source. How does this not kill OpenAI and Claude Code’s closed-source offerings? $BABA
Moon’s Dark Side co-founder, will release open-source Kimi k3. Wow— the world’s number one large language model is going open-source. How does this not kill OpenAI and Claude Code’s closed-source offerings? $BABA
#kimi k3#
Moon’s Dark Side co-founder, will release open-source Kimi k3. Wow— the world’s number one large language model is going open-source. How does this not kill OpenAI and Claude Code’s closed-source offerings? $BABA
Moon’s Dark Side co-founder, will release open-source Kimi k3. Wow— the world’s number one large language model is going open-source. How does this not kill OpenAI and Claude Code’s closed-source offerings? $BABA
Super awesome Kimi k3 from the Dark Side of the Moon, and his largest external shareholder is Alibaba, holding 36%. If we estimate the value of Moon's Dark Side based on OpenAI’s valuation, the value of the shares Alibaba currently holds is $500 billion—two valuations for Alibaba right now
$BABA
$BABA
A group of people still don’t understand why the Dark Side of the Moon is so awesome—what does it have to do with Alibaba? Because Alibaba holds 36% of the Dark Side of the Moon’s shares, participates in every funding round, and if it goes public it will directly break through $10,000. And that price directly creates an investment income of $360 billion for Alibaba.
Alibaba’s greatest investment isn’t investing in ChangXin Memory, but investing in the Dark Side of the Moon, acquiring 36% of the shares—Kimi K3 directly smashes through
Fable 5 is still just cheap cardboard. Kimi K3 directly built it for you—J-12. This really just blows right through everything.
半沐
·
--
The strongest model in the world this year, Kimi K3, is out—it's awesome!
The strongest model in the world this year, Kimi K3, is out—it's awesome!
Crushed through—crushed through—Fable 5 can only be called “junior” now. New architecture really is awesome!
Kimi k3 is here—outperforms GPT-5.6, and is even more formidable than Fable 5. It’s ready to break through everything. Get ready with Alibaba’s treasure.
After Kimi k3 is released, if its capabilities push into Fable 5, then its valuation is beyond limitation—directly surpassing $1.8 trillion and standing atop the world is not impossible either. Yang Zhilin isn’t a loser—he is, besides Liang Sheng, China’s number one AI figure. With the courage to take three tries for Tsinghua, who would dare to compete with him? At NVIDIA’s keynote, he was the only invited Chinese person. And it’s Alibaba, holding a 36% stake in the Moon’s Dark Side, that truly stands to profit handsomely, overflowing the bowl.
They say Alibaba’s investments are all rotten garbage, but in the AI battle, it fights straight into the future: 5% in Changxin Storage, 36% in the Moon’s Dark Side.
$BABAB
They say Alibaba’s investments are all rotten garbage, but in the AI battle, it fights straight into the future: 5% in Changxin Storage, 36% in the Moon’s Dark Side.
$BABAB
After Kimi goes live, it’s expected to move straight toward OpenAI’s valuation—$800 billion. Meanwhile, Alibaba, which holds a 36% stake in Moonshot Dark Side, is valued at only $220 billion.
AB Kuai Dong
·
--
Shareholder disclosure documents confirm that DeepSeek's valuation has exceeded RMB 350 billion. Listed company Kimpers? opened? shares (KAIrun Co., Ltd.) released an evening announcement about the progress of its indirect investment in DeepSeek, unexpectedly disclosing DeepSeek's latest valuation.
Based on the investment amount and working backward from the shares held, after its previous funding round DeepSeek's valuation has risen to approximately RMB 350.8 billion, or about $51.8 billion.
Based on the investment amount and working backward from the shares held, after its previous funding round DeepSeek's valuation has risen to approximately RMB 350.8 billion, or about $51.8 billion.
What does it mean to directly align with Claude Fable 5? It’s the world’s most powerful large model—China has one too, and its name is Kimi K3. The biggest beneficiary is Alibaba.
半沐
·
--
The Kimi of the Dark Side of the Moon is about to release k3. Its capability is more than ten times that of k2.6. It may even surpass Claude Code 4.8 and Codex 5.6, and directly align with Fable. And the biggest beneficiaries are Alibaba and Tencent. Alibaba holds 36% of the Dark Side of the Moon, Tencent holds 7%, and it's crushing.
Turns out all the investment by Alibaba is a mess, and I didn’t expect that this AI battle, Alibaba would have everything right
ChangXin Memory: 5% stake
Zhipu: 3% stake
Mysterious Dark Side of the Moon: 36% stake
Terrifyingly so
ChangXin Memory: 5% stake
Zhipu: 3% stake
Mysterious Dark Side of the Moon: 36% stake
Terrifyingly so
The Kimi of the Dark Side of the Moon is about to release k3. Its capability is more than ten times that of k2.6. It may even surpass Claude Code 4.8 and Codex 5.6, and directly align with Fable. And the biggest beneficiaries are Alibaba and Tencent. Alibaba holds 36% of the Dark Side of the Moon, Tencent holds 7%, and it's crushing.
Log in to explore more content