【文章标题】：[AINews] Poolside gets $12B reverse-execuhire to NVIDIA; founders stay for $1B, employees go for $6B, Infraco scaling to 7GW neocloud
【文章标题翻译】：[AINews] Poolside 以 120 亿美元反向“高管雇佣”交易加入 NVIDIA；创始人留下获 10 亿美元，员工离开获 60 亿美元，Infraco 扩展至 7GW 新云
【文章正文】： Less than a month ago we had just featured
Poolside’s Model Factory with Eiso Kant
on the pod (following
our Paper Club
coverage):
【正文翻译】：不到一个月前，我们刚刚在播客中介绍了 Poolside 的 Model Factory 与 Eiso Kant（此前还有我们的 Paper Club 报道）：
It appears that Jensen really, really liked Poolside too, as he went from
investor
to doing licensing their factory and hiring 109 of their employees:
【正文翻译】：看来 Jensen 也非常非常喜欢 Poolside，因为他从投资者变成了授权他们的工厂并雇佣了 109 名员工：
Unless things changed drastically, this accounts for the
overwhelming majority of the technical Poolside employees
:
【正文翻译】：除非情况发生巨大变化，这占了 Poolside 技术员工的绝大多数：
Eiso Kant [01:52:31]: We are hiring on every possible role in applied research and engineering in the company, from training all the way to evals to post-training architecture. Like, we are still in a world where, individuals can have massive impact. And I think our pitch to join us —I think we are one of the places where it’s the highest ratio to individual to impact, Right? L
ess than 70 people built this model. Less than 115 between engineering and researchers
, like, together did this effort, and that’s a very broad definition ‘cause I put myself in the 115 list.
【正文翻译】：Eiso Kant [01:52:31]：我们正在招聘公司应用研究和工程领域的每一个可能职位，从训练一直到评估再到后训练架构。我们仍然处于一个个人可以产生巨大影响的世界。我认为我们吸引人们加入的理由是——我认为我们是个人与影响力比率最高的地方之一，对吧？不到 70 人构建了这个模型。工程和研究人员加起来不到 115 人，共同完成了这项工作，而且这是一个非常宽泛的定义，因为我自己也算在那 115 人名单里。
As the founders say, this is “not an acquisition and not an acquihire”:
【正文翻译】：正如创始人所说，这“不是收购，也不是人才收购”：
We’ve been calling the Windsurf-Google and Character-Google and Scale-Meta and Instacart-OpenAI deals
execuhires
because usually the executives go leaving the employees with a rich payout but holding the company remaining, but this is a first time it is happening the other way around. The action amounts to founders pivoting the company extremely hard to SOMETHING, and finding an EXTREMELY comfortable golden parachute for investors and employees to continue on with the original mission or stay aboard for the new pivot:
【正文翻译】：我们一直将 Windsurf-Google、Character-Google、Scale-Meta 和 Instacart-OpenAI 的交易称为“高管雇佣”，因为通常高管离开时会给员工留下丰厚报酬但公司仍然存续，但这是第一次反过来发生。这一行动相当于创始人极其剧烈地将公司转向某个方向，并为投资者和员工找到了一个极其舒适的金色降落伞，让他们可以继续原有使命或留在新方向上：
For the last 3 ½ years we’ve been directionally correct in a race where capital requirements went vertical.
【正文翻译】：在过去三年半里，我们在一场资本需求直线上升的竞赛中方向是正确的。
At the end of last year, we had a 6 week window in which to raise $2 billion dollars to pay for a 40,000 GB300 cluster coming online in January.
【正文翻译】：去年年底，我们有一个 6 周的窗口期来筹集 20 亿美元，用于支付一月份上线的 40,000 个 GB300 集群。
We didn’t close it in time, and we lost the cluster
.
【正文翻译】：我们没有及时完成融资，失去了那个集群。
and:
【正文翻译】：以及：
We also know that at 10,000-20,000 GB300s we would produce a great model that could rival the current frontier.
【正文翻译】：我们也知道，拥有 10,000-20,000 个 GB300，我们就能训练出一个可以媲美当前前沿水平的优秀模型。
But the scale of next year’s frontier models requires far more than an order of magnitude larger cluster
. And for this the constraint today is not only capital, it is
physical data center space and contracted compute
.
【正文翻译】：但明年前沿模型的规模需要的远不止大一个数量级的集群。而如今这方面的制约不仅是资本，还有物理数据中心空间和已签约的算力。
The compute needed to be at the frontier of the current model recipe is going vertical, and as the world accelerates along the axis of Recursive Self Improvement this will only become more evident.
【正文翻译】：要处于当前模型配方前沿所需的算力正在垂直上升，随着世界沿着递归自我改进的轴加速，这一点只会变得更加明显。
To this end, the PIC infraco, spun out in Jan 2026, is also interesting in its ambitions…
【正文翻译】：为此，2026 年 1 月分拆出来的 PIC infraco 的雄心也很有趣……
We’re confused too, and the founders say they are “not ready to share the updated vision”, but everyone here is coming out with a lot of money so we’re just interested to see what’s next for everyone on the 3 different directions emerging from OG Poolside.
【正文翻译】：我们也很困惑，创始人说他们“还没准备好分享更新后的愿景”，但这里每个人都带着很多钱离开，所以我们只是很感兴趣想看看从 OG Poolside 衍生出的 3 个不同方向上每个人的下一步是什么。
The only hints left to us:
【正文翻译】：留给我们的唯一线索：
We wholeheartedly believe that everything economically valuable, scientifically interesting and a lot of what will be personally meaningful is going to be underpinned by Al.
【正文翻译】：我们全心全意地相信，一切具有经济价值、科学意义以及许多对个人有意义的事物都将由 AI 支撑。
The world has not yet reached 0.1% of this transition
….
【正文翻译】：世界尚未达到这一转变的 0.1%……
… We believe
human level capabilities of intelligence will be fully commoditized by open source models
, while super intelligence will likely
not be
.
【正文翻译】：……我们相信，人类水平的智能能力将被开源模型完全商品化，而超级智能很可能不会。
The world has two types of economically valuable problems, those that are intelligence bound, and those that are
experiment bound
. The first are problems which we can solve by scaling up intelligence e.g. building software, doing accounting, solving a math theorem. The second are ones that require real world experimentation to progress, and
no amount of increased intelligence without experimental results will make progress
. We could put 100,000 of the world’s brightest minds together to solve cancer but without a real world experimental feedback loop, they likely never will.
【正文翻译】：世界上有两类具有经济价值的问题：一类受限于智能，另一类受限于实验。第一类是我们可以通过扩展智能来解决的问题，例如构建软件、做会计、证明数学定理。第二类则需要现实世界的实验才能取得进展，没有实验结果，无论增加多少智能都无法取得进展。我们可以把世界上 10 万个最聪明的头脑聚在一起攻克癌症，但如果没有现实世界的实验反馈循环，他们很可能永远无法成功。
Today’s model revenue is from coding and soon from all of knowledge work. In the future, companies who can go beyond human level capabilities will tap into
revenue coming from scientific discoveries
where there is a true data moat derived from real world experimentation. In our humble opinion, Al’s ultimate value will not derive from the first kind, that will become a low margin commodity, but it will from the second.
【正文翻译】：今天的模型收入来自编程，很快将来自所有知识工作。未来，能够超越人类水平能力的公司将获得来自科学发现的收入，这些发现拥有源自现实世界实验的真正数据护城河。在我们看来，AI 的最终价值不会来自第一类问题，那将变成低利润商品，而会来自第二类。
Al will become the world’s most valuable scientific discovery engine
.
【正文翻译】：AI 将成为世界上最有价值的科学发现引擎。
Fascinating. Sounds like we could not have timed
our AI for Science podcast
better.
【正文翻译】：太棒了。听起来我们的 AI for Science 播客时机再好不过了。
AI News for 8/19/2026-8/20/2026. We checked 12 subreddits,
544 Twitters
and no further Discords.
【正文翻译】：AI News 2026 年 8 月 19 日至 8 月 20 日。我们查看了 12 个 subreddit、544 条 Twitter，没有其他 Discord。
AINews’ website
lets you search all past issues. As a reminder,
AINews is now a section of Latent Space
. You can
opt in/out
of email frequencies!
【正文翻译】：AINews 网站可让你搜索所有往期内容。提醒一下，AINews 现在是 Latent Space 的一个版块。你可以选择接收或退订邮件频率！
AI Twitter Recap
【正文翻译】：AI Twitter 回顾
OpenAI and Anthropic Expand the Agent Product Surface
【正文翻译】：OpenAI 和 Anthropic 扩展智能体产品面
OpenAI pushed several desktop and builder features in one wave
:
【正文翻译】：OpenAI 在一波更新中推出了多项桌面和构建者功能：
launched an
Apple Messages plugin
for ChatGPT Work/Codex on Mac, enabling message search, catch-up, drafting, and sending from the desktop app.
【正文翻译】：@ChatGPT 为 Mac 上的 ChatGPT Work/Codex 推出了 Apple Messages 插件，支持从桌面应用搜索消息、查看遗漏、起草和发送。
also added
collaborative editing for ChatGPT Sites
, with teammates sharing a project while Codex manages git/CI;
shared read-only conversation links
and
PR-context sharing
further push ChatGPT/Codex toward being a coordination surface, not just a chat UI. On the API side,
transparent backgrounds in GPT-Image-2
are now in preview for reusable design assets.
【正文翻译】：@OpenAIDevs 还为 ChatGPT Sites 添加了协作编辑功能，队友可以共享项目，同时 Codex 管理 git/CI；共享只读对话链接和 PR 上下文共享进一步推动 ChatGPT/Codex 成为协调界面，而不仅仅是聊天 UI。在 API 方面，GPT-Image-2 中的透明背景现已提供预览，用于可复用的设计资产。
OpenAI’s desktop memory/workflow features continue rolling out geographically
:
【正文翻译】：OpenAI 的桌面记忆/工作流功能继续按地区推出：
said
Computer History
and cross-app memory are now available in the
EEA, UK, and Switzerland
for Pro/Business/Enterprise Mac users, with
Record & Replay
also live there. Together, these features point to a product strategy of capturing user workflows on-device and turning repeated actions into reusable skills.
【正文翻译】：@OpenAIDevs 表示，计算机历史记录和跨应用记忆现已在欧洲经济区、英国和瑞士面向 Pro/Business/Enterprise Mac 用户开放，Record & Replay 也已上线。这些功能共同指向一种产品策略：在设备上捕获用户工作流，并将重复操作转化为可复用的技能。
Anthropic made its agent platform more composable and production-ready
:
【正文翻译】：Anthropic 使其智能体平台更具可组合性和生产就绪性：
announced general availability for
computer use, browser tool, Skills API, and Files API
on the Claude Platform. The
Skills API
adds versioned reusable procedures; the
Files API
now supports expiration control,
5x higher rate limits
to
500 RPM
, and
1 TB/org
. Anthropic also published an
AG-UI adapter for Claude Managed Agents
, mapping chat threads to managed sessions and streaming text, tool calls, and thinking into custom UIs.
【正文翻译】：@ClaudeDevs 宣布 Claude 平台上的计算机使用、浏览器工具、Skills API 和 Files API 全面可用。Skills API 增加了版本化的可复用流程；Files API 现在支持过期控制、速率限制提高 5 倍至 500 RPM，以及每组织 1 TB。Anthropic 还发布了适用于 Claude Managed Agents 的 AG-UI 适配器，将聊天线程映射到托管会话，并将文本、工具调用和思考流式传输到自定义 UI。
Model Economics, Usage Limits, and the Enterprise Shift Toward Open Models
【正文翻译】：模型经济学、使用限制以及企业向开放模型的转变
- AT&T became the clearest public case study yet for hybrid routing
-
the most consequential enterprise datapoint in the set came via
, summarizing AT&T’s internal AI deployment:
40% of employee AI usage already routes to open models
, with a target of
60–70%
;
coding costs are down 56%
for only a
2% quality drop
, at
45B tokens/day
. That supports the increasingly common view that frontier closed models remain reserved for the hardest tasks, while “good-enough” open models eat the broad middle of enterprise demand.
【正文翻译】：AT&T 成为混合路由最清晰的公开案例研究：这组数据中最重要的企业数据点来自 @Hesamation，总结了 AT&T 内部 AI 部署：40% 的员工 AI 使用已经路由到开放模型，目标是 60–70%；编码成本下降 56%，质量仅下降 2%，每天处理 450 亿 token。这支持了越来越普遍的观点：前沿闭源模型仍保留给最困难的任务，而“足够好”的开放模型则吞噬了企业需求的广阔中间地带。
explicitly framed this as a warning sign for OpenAI/Anthropic’s enterprise moat, while
welcomed AT&T to open models.
【正文翻译】：@amir 明确将此视为 OpenAI/Anthropic 企业护城河的警示信号，而 @ollama 欢迎 AT&T 加入开放模型。
Pricing pressure is intensifying across closed-model distribution
:
【正文翻译】：闭源模型分销的定价压力正在加剧：
announced
GPT-5.6 Sol at 50% off
through Router, and both
and
amplified the temporary discount for GitHub Copilot / VS Code users. At the same time, user sentiment suggests supply constraints are surfacing as usage caps rather than degraded quality:
complained that a
$200/mo OpenAI Pro plan
could be exhausted in a single heavy Codex day, and
noted it was possible to continue consuming substantial tokens after hitting the stated cap. The broader signal: labs are still searching for the right product boundary between high-end model access and economically sustainable agentic usage.
【正文翻译】：@eglyman 宣布通过 Router 提供 GPT-5.6 Sol 五折优惠，@github 和 @code 都放大了针对 GitHub Copilot / VS Code 用户的临时折扣。与此同时，用户情绪表明供应限制正以使用上限而非质量下降的形式显现：@bridgemindai 抱怨 200 美元/月的 OpenAI Pro 计划可能在重度使用 Codex 的一天内耗尽，而 @theo 指出在达到规定上限后仍可能继续消耗大量 token。更广泛的信号是：各实验室仍在寻找高端模型访问与经济可持续的智能体使用之间的正确产品边界。
Open-weight adoption and distribution continue to broaden
:
【正文翻译】：开放权重模型的采用和分发继续扩大：
said
Kimi K3
is now rolled out to over half its subscription base with
US/EU hosting
and
zero data retention
. On the open ecosystem side,
and
highlighted
Gemma surpassing 1B downloads
, while
@_philschmid
launched an
Awesome Gemma
repo aggregating variants, deployment guides, and fine-tuning recipes.
【正文翻译】：@ollama 表示 Kimi K3 现已向超过一半的订阅用户推出，提供美国/欧盟托管和零数据保留。在开放生态系统方面，@Google 和 @osanseviero 强调 Gemma 下载量已超过 10 亿次，而 @_philschmid 推出了一个 Awesome Gemma 仓库，汇总了各种变体、部署指南和微调配方。
Multimodal and Agent Benchmarks: Muse Spark, GLM-5.3, Gemini 3.7 Flash
【正文翻译】：多模态与智能体基准：Muse Spark、GLM-5.3、Gemini 3.7 Flash
Meta’s Muse Spark 1.2 had a strong benchmark day across multimodal/agentic evals
:
【正文翻译】：Meta 的 Muse Spark 1.2 在多模态/智能体评估方面表现强劲：
presented demos spanning
visual coding, robotics planning, and audio-visual understanding
, and previewed
WildArtifactBench
, an internal eval using
win rates and Elo from human/agentic judges
for practical multimodal tasks. Third-party measurements were favorable:
reported
+2.1% net improvement
in Agent Arena, up from
0.9%
in v1.1, with particularly strong
Bash Recovery (+11.4%)
;
placed Muse Spark 1.2
1 for Video-to-Website¶
,
2 for Image-to-HTML¶
, and
3 for Image-to-Frontend¶
, while noting it sits on the
price-preference Pareto frontier
.
【正文翻译】：@AIatMeta 展示了涵盖视觉编码、机器人规划和音视频理解的演示，并预览了 WildArtifactBench，这是一个内部评估，使用人类/智能体裁判的胜率和 Elo 评分来评估实用多模态任务。第三方测量结果良好：@arena 报告 Agent Arena 净改进 +2.1%，高于 v1.1 的 0.9%，其中 Bash Recovery 尤其强劲（+11.4%）；@DesignArena 将 Muse Spark 1.2 评为 Video-to-Website 第 1、Image-to-HTML 第 2、Image-to-Frontend 第 3，并指出它位于价格-偏好帕累托前沿。
Zhipu’s GLM-5.3 keeps showing up in agentic/code evals