AI前沿每日脉动

17SEP

AI前沿每日脉动

AI Frontier Pulse · 中英双语版 Bilingual Edition

2026.09.17 · 周四刊

14 位 Builder27 条推文1 期播客1 篇博客

ChatGPT WorkApplied AIOpen WeightsTokens per WattTrust & Access

Richard Liu · 2026

Curated by Richard Liu · Snapshot follow-builders-2026-09-17-v1

今天最强信号：GPT-5.6 Sol 只用两项运行设置变化，就在 ARC-AGI-3 达到 SoTA。模型能力不再能脱离 harness、compaction 与跨 context 执行单独评估。The strongest signal: two runtime setting changes put GPT-5.6 Sol at the top of ARC-AGI-3. Model capability can no longer be separated from harness design, compaction, and cross-context execution.

@sama X 原文12.2K ❤ · 545 RT · 1K 💬

Sam Altman：big 🚢 this week and then for devday 🚢🚢🚢🚢🚢🚢Builder 动态 · Source Excerpt

Sam Altman 原文要点：big 🚢 this week and then for devday 🚢🚢🚢🚢🚢🚢

Source-grounded excerpt from Sam Altman: big 🚢 this week and then for devday 🚢🚢🚢🚢🚢🚢

@sama X 原文12.2K ❤ · 545 RT · 1K 💬

Claude：Salesforce in Claude is now available in beta.技能化工作流 · Source Excerpt

Claude 原文要点：Salesforce in Claude is now available in beta.

Source-grounded excerpt from Claude: Salesforce in Claude is now available in beta.

@claudeai X 原文8.9K ❤ · 634 RT · 285 💬

02 / 15

Thibault Sottiaux：2026 is the year of efficiency and Tuesday is for random sBuilder 动态 · Source Excerpt

Thibault Sottiaux 原文要点：2026 is the year of efficiency and Tuesday is for random swag drops on your door handle

Source-grounded excerpt from Thibault Sottiaux: 2026 is the year of efficiency and Tuesday is for random swag drops on your door handle

@thsottiaux X 原文4.9K ❤ · 63 RT · 1.2K 💬

Thariq：I was not expecting things to go this way, but I think MCPBuilder 动态 · Source Excerpt

Thariq 原文要点：I was not expecting things to go this way, but I think MCPs are better than CLIs for most integrations.

Source-grounded excerpt from Thariq: I was not expecting things to go this way, but I think MCPs are better than CLIs for most integrations.

@trq212 X 原文2.4K ❤ · 109 RT · 209 💬

03 / 15

AI 降低执行成本，却没有消除品味与经验：资深创始人的判断继续复利，设计工具也必须把审美、易用性和乐趣一起产品化。AI lowers execution cost without eliminating taste or experience. Seasoned judgment compounds, while design tools must productize aesthetics, ease, and fun together.

@danshipper X 原文1.5K ❤ · 74 RT · 50 💬

Dan Shipper：we almost never test new foundation models but we've been Builder 动态 · Source Excerpt

Dan Shipper 原文要点：we almost never test new foundation models but we've been testing this for ~a week @every and it's pretty wild.

Source-grounded excerpt from Dan Shipper: we almost never test new foundation models but we've been testing this for ~a week @every and it's pretty wild.

@danshipper X 原文1.5K ❤ · 74 RT · 50 💬

Guillermo Rauch：Excited to formally introduce Vercel Labs.Builder 动态 · Source Excerpt

Guillermo Rauch 原文要点：Excited to formally introduce Vercel Labs.

Source-grounded excerpt from Guillermo Rauch: Excited to formally introduce Vercel Labs.

@rauchg X 原文1.1K ❤ · 47 RT · 66 💬

04 / 15

Amjad Masad：The AI naming curse strikes again.Builder 动态 · Source Excerpt

Amjad Masad 原文要点：The AI naming curse strikes again.

Source-grounded excerpt from Amjad Masad: The AI naming curse strikes again.

@amasad X 原文659 ❤ · 70 RT · 33 💬

Amjad Masad：If you’re curious about Effective Altruism, the origins ofBuilder 动态 · Source Excerpt

Amjad Masad 原文要点：If you’re curious about Effective Altruism, the origins of the cult, their tactics, and their influence network — check out my conversation with Tucker from two years ago:

Source-grounded excerpt from Amjad Masad: If you’re curious about Effective Altruism, the origins of the cult, their tactics, and their influence network — check out my conversation with Tucker from two years ago:

@amasad X 原文646 ❤ · 49 RT · 26 💬

Guillermo Rauch：The future is multi-model.产品组织 · Source Excerpt

Guillermo Rauch 原文要点：The future is multi-model.

Source-grounded excerpt from Guillermo Rauch: The future is multi-model.

@rauchg X 原文380 ❤ · 10 RT · 35 💬

Josh Woodward：New updates in @Gemini_Notebook for students, my 2 favorit低延迟模型 · Source Excerpt

Josh Woodward 原文要点：New updates in @Gemini_Notebook for students, my 2 favorites: 1) Have a live, spoken Q&A with your class materials in ~100 languages 2) Record lectures and notes on the go, the…

Source-grounded excerpt from Josh Woodward: New updates in @Gemini_Notebook for students, my 2 favorites: 1) Have a live, spoken Q&A with your class materials in ~100 languages 2) Record lectures and notes on the go, the…

@joshwoodward X 原文329 ❤ · 19 RT · 22 💬

05 / 15

今天的深度信号来自官方博客：AI 工具正在把工作过程转成更可协作的界面。Today's depth signal comes from the official blog: AI tools are turning work into more collaborative interfaces.

Claude Blog 原文链接

Claude for Small Business launches new workflows, integrations, and training programsToday's Deep Blog

今天的博客信号来自官方长文：AI 工具正在把上下文、工具调用和协作流程整合到更完整的工作界面中。

Today's blog signal points to AI tools integrating context, tool use, and collaboration into more complete work interfaces.

Claude Blog 原文链接Sep 15, 2026

06 / 15

PODCAST DEEP DIVE

AI 的下一轮竞争不只发生在模型层，也发生在冷却、电网、数据中心、供应链与每瓦 token 数构成的物理系统里。The next AI race is not only at the model layer, but across cooling, grids, data centers, supply chains, and tokens per watt.

Training Data YouTube 完整播客

Box's Aaron Levie: On Reinventing Yourself in the AI Age and Enterprise DiffusionPodcast Deep Dive

今天的播客把产品、基础设施和商业化放在同一条线上：AI 不只是模型能力，也是速度、成本和交付系统。

Today's podcast connects product, infrastructure, and commercialization: AI is not only model ability, but speed, cost, and delivery systems.

Training Data YouTube 完整播客2026-09-15

07 / 15

算力上线即被消耗New Compute Is Consumed Immediately

新增容量上线即被工作负载吸收，过早放慢建设通常会带来负面意外。

New capacity is consumed immediately, and slowing the build too early repeatedly creates negative surprises.

Compute YouTube 完整播客

关键指标：每瓦 token 数The Metric Is Tokens Per Watt

定制芯片与模型工作负载协同设计，目标是在电力约束下最大化每瓦可生成的 token。

Custom silicon and model workloads are co-designed to maximize tokens generated per watt.

Efficiency YouTube 完整播客

数据中心必须新增电力Data Centers Must Add Power

大规模 AI 基础设施需要同步投资发电、输电、变电站与配套容量。

Large-scale AI infrastructure requires new generation, transmission, substations, and supporting capacity.

Grid YouTube 完整播客

算力策略是 All of the AboveThe Strategy Is All Of The Above

多种云、芯片、数据中心与供应路径并行推进，减少单点约束。

Clouds, chips, data centers, and supply paths advance in parallel to reduce single-point constraints.

Portfolio YouTube 完整播客

08 / 15

Amjad Masad：The AI naming curse strikes again.Builder 动态 · Source Excerpt

Amjad Masad 原文要点：The AI naming curse strikes again.

Source-grounded excerpt from Amjad Masad: The AI naming curse strikes again.

@amasad X 原文659 ❤

Amjad Masad：If you’re curious about Effective Altruism, the origins ofBuilder 动态 · Source Excerpt

Amjad Masad 原文要点：If you’re curious about Effective Altruism, the origins of the cult, their tactics, and their influence network — check out my conversation with Tucker from two years ago:

Source-grounded excerpt from Amjad Masad: If you’re curious about Effective Altruism, the origins of the cult, their tactics, and their influence network — check out my conversation with Tucker from two years ago:

@amasad X 原文646 ❤

Guillermo Rauch：The future is multi-model.产品组织 · Source Excerpt

Guillermo Rauch 原文要点：The future is multi-model.

Source-grounded excerpt from Guillermo Rauch: The future is multi-model.

@rauchg X 原文380 ❤

Josh Woodward：New updates in @Gemini_Notebook for students, my 2 favorit低延迟模型 · Source Excerpt

Josh Woodward 原文要点：New updates in @Gemini_Notebook for students, my 2 favorites: 1) Have a live, spoken Q&A with your class materials in ~100 languages 2) Record lectures and notes on the go, the…

Source-grounded excerpt from Josh Woodward: New updates in @Gemini_Notebook for students, my 2 favorites: 1) Have a live, spoken Q&A with your class materials in ~100 languages 2) Record lectures and notes on the go, the…

@joshwoodward X 原文329 ❤

Amjad Masad：This is cool, but if your output domain is known in advancBuilder 动态 · Source Excerpt

Amjad Masad 原文要点：This is cool, but if your output domain is known in advance, why not just train a model to produce logprobs over enums?

Source-grounded excerpt from Amjad Masad: This is cool, but if your output domain is known in advance, why not just train a model to produce logprobs over enums?

@amasad X 原文326 ❤

Guillermo Rauch：Safari 27 supports JSPI, a magic WebAssembly capability foBuilder 动态 · Source Excerpt

Guillermo Rauch 原文要点：Safari 27 supports JSPI, a magic WebAssembly capability for sync native code to suspend on an async 𝙿𝚛𝚘𝚖𝚒𝚜𝚎.

Source-grounded excerpt from Guillermo Rauch: Safari 27 supports JSPI, a magic WebAssembly capability for sync native code to suspend on an async 𝙿𝚛𝚘𝚖𝚒𝚜𝚎.

@rauchg X 原文284 ❤

09 / 15

今日数据概览Today Stats

收录 Builder：14 位
总推文数：27 条
播客节目：1 期
博客文章：1 篇

最高互动：Sam Altman：big 🚢 this week and then for devday 🚢🚢🚢🚢🚢🚢 · 12.2K ❤
第二高互动：Claude：Salesforce in Claude is now available in beta. · 8.9K ❤

The snapshot includes 14 builders, 27 tweets, 1 podcast, and 1 blog post.

follow-buildersSnapshot follow-builders-2026-09-17-v1

5 条关键洞察5 Key Takeaways

ChatGPT Work 正从任务助手变成可调度上下文、应用与协作的新计算入口。

ChatGPT Work is becoming a computing surface that orchestrates context, apps, and collaboration.

模型越强，企业越需要连接数据、监管、人类决策点与现实反馈回路的 applied AI 层。

Stronger models increase demand for an applied-AI layer spanning data, regulation, human decisions, and real-world feedback.

开放权重从开发者偏好升级为产业领导力与生态扩散战略。

Open weights are moving from developer preference to an industrial strategy for leadership and ecosystem diffusion.

算力竞争的关键约束转向物理世界：电力、冷却、建设速度与每瓦 token 数。

Compute competition is increasingly constrained by power, cooling, build speed, and tokens per watt.

跨应用执行放大了价值，也把 Gmail、日历与办公数据的信任问题推到产品核心。

Cross-app execution expands value while making trust around Gmail, calendars, and office data a core product issue.

10 / 15

本周之声 VOICE 01

只要允许模型持续推理，并借助 canonical compaction 跨多个 context window 工作。

“You just have to allow it to reason and work over multiple context windows.”

@sama X 原文12.2K ❤ · 545 RT · 1K 💬

11 / 15

本周之声 VOICE 02

最好的路径，是赋能科学家，而不是试图由我们自己解决所有问题。

“The best way to do this is for us to empower scientists.”

@thsottiaux X 原文4.9K ❤ · 63 RT · 1.2K 💬

12 / 15

本周之声 VOICE 03

我总会先创建 design.md，写清颜色、字体以及其他设计准则。

“I always create a design md file with guidelines for color, typography, and more.”

@amasad X 原文646 ❤ · 49 RT · 26 💬

13 / 15

播客之声 PODCAST VOICE

AI 是一场“50 年的一夜成功”。

“AI's fifty year overnight success.”

Kevin Kelly · AI & I by Every YouTube 完整播客

14 / 15

AI前沿每日脉动

AI Frontier Pulse · 2026.09.17

本期收录 14 位 Builder · 27 条推文 · 1 期播客 · 1 篇博客

ChatGPT Work · Applied AI · Open Weights · Tokens per Watt · Trust & Access

感谢阅读 · Thank You For Reading

Richard Liu · AI前沿每日脉动 · 2026 · Snapshot follow-builders-2026-09-17-v1