探索 | Thread Easy - 展开 Twitter 线程｜阅读、总结与创作

I think the fact that DeepSeek cares so little about token efficiency should terrify competitors. They could have Opus-like CoTs but they are actively NOT using best length penalty tricks. Anthropic, not DeepSeek, wants “economical AI” on METR tasks. They really want infinity.

We're in a race. It's not USA vs China but humans and AGIs vs ape power centralization. @deepseek_ai stan #1, 2023–Deep Time «C’est la guerre.» ®1

Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)

Sun Dec 07 02:18:23

RT @PandaTalk8: 那些优秀的人，都有一个特质，就是他们能想清楚终局。我2024年初开始认真发推，最主要原因是我看到了很多大佬晒他们的推特广告分成收入，我当时看到了两点可能性。

Believing is seeing

Yangyi

Sun Dec 07 02:14:38

The difference between the Soviet Communists and the California Communists? The Soviets wanted to win

Silicon Valley entrepreneur (cofounder@ TrialPay, Yub, Affirm, Point, TXN), investor (General Partner @a16z), husband, father and sarcast (one who is sarcastic)

Alex Rampell

Sun Dec 07 02:12:23

信息量非常大，整理了一天：一份关于“AI 真正是如何被使用”的百兆级数据报告基于 100 万亿 Token 的真实 LLM 使用数据《State of AI 2025》报告深度解读它揭示了： 2025年 AI正在变成会“思考和行动”的智能体（Agentic Inference）角色扮演和编程，占据AI使用的近90%； “中型模型”正在吞噬大模型市场；推理型模型（Reasoning Models）已成主流；中国开源力量正在迅速崛起 ...

人们都在用AI干什么？ 100万亿 Token 给出的 AI 启示：《State of AI 2025》报告深度解读详细报告内容解读：https://t.co/O7sbsS4wQv

小互

Sun Dec 07 02:11:57

RT @indie_maker_fox: 🚨 Nextjs发现高危安全漏洞 https://t.co/5t01yx4zCe Mkdirs模板完全无事，MkSaaS模板所有仓库所有分支一大早都已升级修复，建议大家立即同步代码，或者参考文档进行升级。漏洞非常危险，因为跟R…

🔥 The best AI SaaS boilerplate - https://t.co/VyNtTs0jSX 🚀 The best directory boilerplate with AI - https://t.co/wEvJ1Dd8aR 🎉 https://t.co/bh1RxeERuY & https://t.co/zubXJCoY92 & https://t.co/tfQf8T7gGF

Fox@MkSaaS.com

Sun Dec 07 02:10:29

RT @vansinhu: Hugging Face Weekly Paper Trends @_akhaliq (Gen by nana-banana-pro)

AI research paper tweets, ML @Gradio (acq. by @HuggingFace 🤗) dm for promo ,submit papers here: https://t.co/UzmYN5XOCi

AK

Sun Dec 07 02:09:28

探索

最新在前，按卡片方式浏览线程

探索

最新在前，按卡片方式浏览线程

I think the fact that DeepSeek cares so little about token efficiency should terrify competitors. They could have Opus-like CoTs but they are actively NOT using best length penalty tricks. Anthropic, not DeepSeek, wants “economical AI” on METR tasks. They really want infinity.

RT @PandaTalk8: 那些优秀的人，都有一个特质，就是他们能想清楚终局。我2024年初开始认真发推，最主要原因是我看到了很多大佬晒他们的推特广告分成收入，我当时看到了两点可能性。

The difference between the Soviet Communists and the California Communists? The Soviets wanted to win

RT @indie_maker_fox: 🚨 Nextjs发现高危安全漏洞 https://t.co/5t01yx4zCe Mkdirs模板完全无事，MkSaaS模板所有仓库所有分支一大早都已升级修复，建议大家立即同步代码，或者参考文档进行升级。漏洞非常危险，因为跟R…

RT @vansinhu: Hugging Face Weekly Paper Trends @_akhaliq (Gen by nana-banana-pro)

探索

最新在前，按卡片方式浏览线程

探索

最新在前，按卡片方式浏览线程

I think the fact that DeepSeek cares so little about token efficiency should terrify competitors. They could have Opus-like CoTs but they are actively NOT using best length penalty tricks. *Anthropic*, not DeepSeek, wants “economical AI” on METR tasks. They really want infinity.

RT @PandaTalk8: 那些优秀的人， 都有一个特质， 就是他们能想清楚终局。 我2024年初开始认真发推， 最主要原因是我看到了很多大佬晒他们的推特广告分成收入， 我当时看到了两点可能性。

The difference between the Soviet Communists and the California Communists? The Soviets wanted to win

RT @indie_maker_fox: 🚨 Nextjs发现高危安全漏洞 https://t.co/5t01yx4zCe Mkdirs模板完全无事，MkSaaS模板所有仓库所有分支一大早都已升级修复，建议大家立即同步代码，或者参考文档进行升级。 漏洞非常危险，因为跟R…

RT @vansinhu: Hugging Face Weekly Paper Trends @_akhaliq (Gen by nana-banana-pro)

I think the fact that DeepSeek cares so little about token efficiency should terrify competitors. They could have Opus-like CoTs but they are actively NOT using best length penalty tricks. Anthropic, not DeepSeek, wants “economical AI” on METR tasks. They really want infinity.

RT @PandaTalk8: 那些优秀的人，都有一个特质，就是他们能想清楚终局。我2024年初开始认真发推，最主要原因是我看到了很多大佬晒他们的推特广告分成收入，我当时看到了两点可能性。

RT @indie_maker_fox: 🚨 Nextjs发现高危安全漏洞 https://t.co/5t01yx4zCe Mkdirs模板完全无事，MkSaaS模板所有仓库所有分支一大早都已升级修复，建议大家立即同步代码，或者参考文档进行升级。漏洞非常危险，因为跟R…