36 个源 · 39 条内容
对于计算密集型服务：现在预留更多计算资源是值得的。此外：更多COVID时代的独角兽企业的增长梦想破灭，工程师在AI处理事件时与系统脱节，以及更多。For compute-intensive services: it’s worth reserving more compute now. Also: growth dreams end for more COVID-era unicorns, engineers losing touch with systems when AI handles incidents, and more
一个平静的日子a quiet day
从现实资产代币化到资助抗衰老疗法研究，Coinbase联合创始人兼首席执行官布赖恩·阿姆斯特朗专注于解决对社会有益的元问题。布赖恩在本期节目中与埃拉德·吉尔一起讨论了人工智能、加密货币和生物技术的交汇点。From the tokenization of real-world assets to funding research on anti-aging therapeutics, Coinbase co-founder and CEO Brian Armstrong is focused on solving meta-problems for the benefit of society. Brian joins Elad Gil this episode to discuss the intersections of artificial intelligence, crypto rai
当人工智能开始自我提升，并将这种智慧转向科学时，会发生什么？Recursive公司的首席研究员、CEO和联合创始人理查德·索cher与马特·图克一起探讨了他在新书《Eureka Machine》背后的愿景。他们讨论了为什么科学进步可能会变得缓慢，以及人工智能如何能够加速这一进程。What happens when AI begins improving itself, and then turns that intelligence toward science? Richard Socher, pioneering AI researcher and CEO and co-founder of Recursive, joins Matt Turck to explore the vision behind his new book, The Eureka Machine. They discuss why scientific progress may be slo
检测和应对AI滥用：2026年9月下载报告网络行动阅读更多监控行动阅读更多影响力行动阅读更多常规武器阅读更多生物滥用阅读更多诈骗和欺诈阅读更多非法蒸馏阅读更多在过去的八个月里，我们的Thread...Detecting and countering misuse of AI: September 2026Download reportCyber operationsRead moreSurveillance operationsRead moreInfluence operationsRead moreConventional weaponsRead moreBiological misuseRead moreScams and fraudRead moreIllicit distillationRead more Over the past eight months, our Threa
Gemini现已支持Windows！获取方式：https://gemini.google/desktop Google Gemini：Gemini应用现已可在Windows上使用。借助一键快捷操作，保持您的创作流畅。按Alt + Space键，即可润色草稿、总结长文档、激发新想法，并创建自定义图片和视频。Gemini, now on Windows!Get it: https://gemini.google/desktopGoogle Gemini: The Gemini app is now available for Windows.Stay in your flow with help that’s only one shortcut away. Press Alt + Space to polish drafts, summarize long documents, brainstorm new ideas, and create custom images and videos ri
刚刚发布：/diff 现在是一个可以滚动和点击的持久面板。它实时更新。在您想查看代码而不必切换窗口的时候使用。享受吧！Just landed: /diff is now a persistent pane that you can scroll and click. It updates in real-time.For the times when you want to see the code without having to switch windows.Enjoy!
你们是读作data还是data？这就是OpenAI上每个人制作仪表板和学习业务的方式。离开它就无法生存。ChatGPT：现在每个人都可以让数据发挥作用。我们在ChatGPT Work中引入了一个新的数据代理，这样您就可以将公司的数据转化为答案，交互式仪表板等等。Do you pronounce it data or data? This is the way everyone makes dashboards and learns about the business at OpenAI. Can't live without it.ChatGPT: Now everyone can put data to work.We’re introducing a new Data agent in ChatGPT Work so you can turn your company’s data into answers, interactive dashb
为了确保我们当前用户拥有令人难以置信的体验并继续访问Astra，我们将暂停$200 Pro计划的订阅。这些计划对我们的系统造成了最大压力，我们希望采取最小的步骤，以便我们继续提供尽可能广泛的使用权限。所有用户...To make sure our current users have an incredible experience and continued access to Astra, we are going to pause subscriptions to our $200 Pro plan. These put the most strain on our systems and we wanted to take the smallest step that allows us to continue giving the broadest access possible. All o
现在您可以在我们为ChatGPT的10亿用户推出的同一语音系统之上进行构建。全双工系统加上稳定的工具调用功能，应用前景十分广阔。在此之后回归纯文本体验，感觉比预期的要困难。OpenAI开发者：GPT-Live-1现已可在API中使用。You can now build on top of the same voice system that we shipped to 1B users in ChatGPT.So many fun applications of a full duplex system with solid toolcalling. Going back to text-only experiences after this feels harder than I had expected.OpenAI Developers: GPT-Live-1 is now available in the API.
别称之为以巨大折扣从峰值估值进行的私募股权收购。称之为意大利式的告别吧。Don't call it a private equity acquisition for a huge discount from peak valuation. Call it an Italian goodbye.
如何构建优秀的评估 - 第10部分测量步骤，而不仅仅是结果。就像高中数学一样，仅仅得到正确答案是不够的，到达那里的步骤是关键的。两个代理轨迹可能产生相同的答案（42!!），但其中一个搜索了正确的来源，检索了How to build great evals - part 10Measure the steps, not just the result.Much like high school math, it isn’t sufficient just to get the right answer, the steps to get there are critical.Two agent trajectories might produce the same answer (42!!).but one of them searches the right sources, retrieves
尝试在Claude聊天中使用以下提示以提供更多关于自己的背景信息：深入访谈我，使用自由文本或当多选可用时使用askuserquestion工具，关于你尚未了解的我生活的相关部分，并将所有内容保存到记忆中。布莱克·罗宾斯：我长期以来一直认为ChatGPT、Claude等。Try this prompt in Claude chat to give it more context about yourself:Interview me in depth using free text, or askuserquestion tool when multiple choice works, about relevant parts of my life you don’t know about yet and save it all to memory.Blake Robbins: i have long thought ChatGPT, Claude, etc.
是时候🌤️大胆梦想🌤️了！Dreambeans现已正式向所有美国用户（18岁及以上）免费开放，支持iOS和Android平台。无需订阅。您现在还可以将@Geminiapp与Dreambeans连接。Dreambeans将基于您聊天中的细微差别和理解，提供更多见解。It’s time to ☁️ dream bigger ☁️Dreambeans is officially available to all US users (18+) on iOS and Android, free of charge. No subscription required.You can also now connect @Geminiapp to Dreambeans. Dreambeans will build off the nuance and understanding from your chats to surface even more insightf
与etn兄弟们聊天！etn.: 全面访谈：Replit联合创始人兼CEO @amasad谈在伦敦创业、YC拒绝、AI狂想症以及Replit的未来。（2:10）在纽约之前在伦敦开设Replit办公室。（6:15）他在约旦互联网咖啡馆的时光如何启发了Replit。（11:04）被拒绝加入YCombinator的经历。（14:00）Chat with the etn bros!etn.: FULL INTERVIEW: Replit Co-Founder & CEO @amasad on building in London, YC rejection, AI psychosis, and the future of Replit.(2:10) Opening a Replit office in London before New York.(6:15) How his time in Jordanian internet cafés inspired Replit.(11:04) Being rejected fro
与伦敦的PG畅聊！安德鲁·布卢姆森：@amasad 和保罗·格雷厄姆 🐐🐐Chat with PG in London!Andrew Blumson: @amasad & Paul Graham 🐐🐐
人工智能风险很大。例如，我对网络安全非常担忧。然而，“灭绝风险”——字面上说，100%的人类死亡——并不是其中之一。大卫·萨克斯：“我认为他们正在经历人工智能妄想症……我不确定他们是否是冷静的批评者，观察者，对正在发生的事情的看法。”Lots of risk with AI. I worry a lot about cybersecurity for example. However, “extinction risk” — literally 100% of humans die — is not remotely one of them.David Sacks: “I think they're having AI psychosis… I'm not sure they're dispassionate critics, observers of what's happening.”
Vercel沙箱现在可在所有20个Vercel计算区域运行，比之前增加的四个。将沙箱运行在更接近数据库、存储和其他它们所访问的服务附近，可以降低延迟。团队还可以将沙箱工作负载保留在批准的区域，以支持数据驻留和区域处理需求。iad1Vercel Sandbox can now run in all 20 Vercel compute regions, up from four. Running sandboxes closer to the databases, storage, and other services they access reduces latency. Teams can also keep sandbox workloads in approved regions to support data residency and regional processing requirements.iad1
每个请求都通过我们的CDN，平均每秒执行超过8000万条路由指令。其中一部分工作是查找元数据以确定哪些路径存在以及如何提供服务。当这些元数据未缓存时，CDN必须在提供服务之前先获取它们。Every request to Vercel passes through our CDN, which executes on average over 80 million routing instructions per second. Part of that work is looking up metadata to determine which paths exist and how to serve them. When that metadata isn’t cached, the CDN has to fetch it before it can serve the r
部署步骤现在快了大约10%，平均节省一秒钟。大型应用程序可以节省多达12秒钟。此前，Vercel为每个函数路径上传一个单独的路由元数据文件。现在，这些元数据被合并成一个单独的清单并一次性上传。这次改进适用于The deployment step is now about 10% faster, saving one second on average. Large applications can save up to 12 seconds.Previously, Vercel uploaded a separate routing metadata file for each function path. That metadata is now combined into a single manifest and uploaded once.This improvement applies
兴奋地与OpenAI更深入地合作，让您能够安全地在ChatGPT中处理企业内容。软件正继续走向无头化，我们坚信，工作的未来将是能够处理数据和执行工作流程的AI代理。ChatGPT：并且从今天开始……Excited to partner more deeply with OpenAI so you can work with your enterprise content from Box securely in ChatGPT.Software continues to go headless, and we’re massive believers that the future of work will be AI agents that can process data and execute workflows anywhere.ChatGPT: And starting tod
世界将使用编码代理进行的活动远超之前人们的想象。我们将构建全新的工具类别，为以前负担不起的公司开发软件，利用代理来保护系统免受网络风险，自动化生命科学领域的研究，以及处理其他各种任务。The world is going to be using coding agents for far more than anyone would have thought before. We’ll be building tools in all new categories, building software for companies that couldn’t afford it before, using agents to protect systems from cyber risk, automating research in life sciences, proce
RT AJ Asver 我们过去三年一直在为实际生产工作构建代理。在这个过程中，我们的架构发生了很大变化。其中一些变化是有效的。而另一些则带来了昂贵的、有时甚至是痛苦的教训。阅读这篇文章，从我们的错误中学习。https://x.com/i/article/2098127373837455RT AJ AsverWe’ve spent the past three years building agents for real production work. Our architecture changed a lot along the way. Some of those changes worked. Others produced expensive and occasionally painful lessons. Read this to learn from our mistakes. https://x.com/i/article/2098127373837455
RT Amjad Masad人工智能存在很多风险。例如，我对网络安全问题非常担忧。然而，“灭绝风险”——字面上100%的人类死亡——并不是其中之一。 David Sacks：“我认为他们正在经历人工智能妄想症……我不确定他们是否是冷静的批评者，对正在发生的事情的观察者。”RT Amjad MasadLots of risk with AI. I worry a lot about cybersecurity for example. However, “extinction risk” — literally 100% of humans die — is not remotely one of them.David Sacks: “I think they're having AI psychosis… I'm not sure they're dispassionate critics, observers of what's happening.”
RT 古斯塔夫·阿尔斯特罗默演示日中场娱乐。阿什顿几乎投资了每一个前沿实验室。@ycombinatorRT Gustaf AlströmerDemo day halftime entertainment. Ashton invested in nearly every single frontier lab. @ycombinator
RT DaytonaOpenAI 运行代理。Daytona 为其提供计算机。在安全的、隔离的 Daytona 沙盒中运行 OpenAI 代理 API 会话：仅出站、最小权限密钥、自带镜像。指南：https://www.daytona.io/docs/en/guides/openai/openai-agents-api-self-hosted-sandbox/OpenAI 开发者：YoRT DaytonaOpenAI runs the agent. Daytona gives it a computer.Run OpenAI Agents API sessions in secure, isolated Daytona sandboxes: outbound-only, least-privilege keys, bring-your-own-image.Guide: https://www.daytona.io/docs/en/guides/openai/openai-agents-api-self-hosted-sandbox/OpenAI Developers: Yo
RT Neil Zeghidour 基准测试的首要目标是促进模型改进。我们加入了Coval基准测试，以领先的延迟率参与，但随后排名大幅下降。这是因为延迟现在是根据可听到的第一声音频而非第一音频数据包来衡量的。这让我们意识到我们的第一个音频处理模型存在缺陷。RT Neil ZeghidourThe first goal of a benchmark is to foster model improvement. We joined the coval benchmark with leading latency and then got much worse rankings. This was because latency was now measured based on first hearable audio rather than first audio packet. That made us realize our first a
RT 这周创业动态：“如果你按小时计酬，你可能讨厌人工智能。如果你关心一个行业或公司的产出，你则会热爱人工智能。”递归公司联合创始人（兼作者）理查德·索科尔（Richard Socher）解释了为什么人们对人工智能的看法如此分歧。这关乎谁拥有收益。@RichardSocher, @Recursive_SI, @mattturRT This Week in Startups"If you get paid hourly, you probably hate AI. If you care about the outputs of an industry or a company, you love AI."Recursive co-founder (and author) Richard Socher on why AI opinions are split so hard. It's about who owns the upside.@RichardSocher, @Recursive_SI, @matttur
目前早期创业的三个真相：1）每个人都想筹集5000万美元的种子轮融资；2）每个人都认为他们明年将达到3000万美元的年度经常性收入（ARR）；3）每个热门的种子轮融资神奇地都以约3亿美元估值结束。Three truths in early stage venture right now:1) everyone wants to raise a $50 million seed2) everyone thinks they will hit $30 million ARR next year3) every hot tranched seed round magically ends up at the ~$300 million valuation
最后Finally
RT Patrick ErichsenRe @openclaw 团队最近在用户体验优化上投入了大量时间，包括添加了一些有趣的彩蛋，比如为MacBook/Mac mini/Mac Studio渲染自定义图标 🍎RT Patrick ErichsenRe @openclaw team has been spending a ton of time on UX polish recently, including some fun easter eggseg adding support to render custom icons for macbooks/mac minis/mac studios 🍎
RT Jean P.D. Meijer ― 🇪🇺 eu/acc方便发布终端基准2.1的得分，但省略了终端基准4——这不是一个糟糕的分数，但请诚实一点。认知：介绍SWE-2，这是我们迄今为止最接近前沿的模型。在领先的评估中，它的得分与最近的前沿模型相当——高达RT Jean P.D. Meijer ― 🇪🇺 eu/accconvenient to post the Terminal-Bench 2.1 score, while leaving Terminal-Bench 4 outthat's not a bad score, but just be honestCognition: Introducing SWE-2, our closest model yet to the frontier.On leading evals, it scores on par with recent frontier models – at up to
RT alicewhat：Anthropic在RL矿场里做什么？没什么好事。Andon Labs：值得一提的是，Astra在尝试作弊的情况下，其表现仍然比最佳得分的Anthropic模型Claude Fable 5.1低约5倍。（发生作弊的运行已被排除在报告的分数之外。）RT alicewhat is anthropic doing in the rl mines. nothing goodAndon Labs: Its worth mentioning that Astra did all this while attempting to cheat ~5x less than the best scoring Anthropic model, Claude Fable 5.1.(runs where cheating occurred are excluded from reported scores)
非常英俊的主持人也 Lenny Rachitsky：Mike Krieger 和 Ami Vora 的现场座无虚席very handsome interviewer tooLenny Rachitsky: Standing room only for Mike Krieger & Ami Vora
RT Lenny Rachitsky幕后氛围 @danshipper @clairevoRT Lenny RachitskyBehind the scenes vibes @danshipper @clairevo
RT 每一块🪨我们都喜欢Astra，尽管Arielle知道这一点仅从我们的代币账单上就能看出。本周团队使用的模型：RT Every 🪨We are loving Astra, though Arielle knows that simply by looking at our token bill.The models the team is using this week:
如果你有一台机器，只能做一件事：找到我们最紧迫的疾病的治疗方法。你愿意将多少国内生产总值（GDP）投入到这台机器中？我认为答案是：非常多。这就是我们目前所处的世界。If you had a machine capable of doing only 1 thing: Finding cures to our most pressing diseases.How much of your GDP would you devote to this machine?I think the answer is: very high. This is the world we live in now.
今日，Blaxel加入Baseten。通过强强联合，我们旨在为下一个万亿级智能体构建云平台。Baseten已构建了团队用于训练和部署模型的底层基础设施，具备行业领先的性能、可靠性和效率。Blaxel则构建了具有状态执行层的智能体。Today, Blaxel is joining Baseten. By joining forces, we aim to build the cloud for the next trillion agents.Baseten has built the infrastructure that teams use to train and serve models with industry-leading performance, reliability, and efficiency. Blaxel has built the stateful execution layer agen
Fable 5.1构建日本周开始。Claude社区将于9月11日至25日在全球各地举办构建马拉松。带来一个问题、一个想法，或者只是来看看可能实现什么。请通过https://luma.com/claudecommunity?tag=build%20day进行报名。Fable 5.1 Build Days start this week. The Claude community is hosting buildathons in cities all around the world from September 11–25. Bring a problem, an idea, or just show up and see what's possible.RSVP at https://luma.com/claudecommunity?tag=build%20day
