🚨 AI News | TestingCatalog@testingcatalog · 2天前48

Apify has partnered with Coinbase to add more than 20,000 of its web automation Actors to the x402 ecosystem, giving AI agents thousands of tools they can discover, pay for, and run on their own. When an agent calls an Actor, it gets back an HTTP 402, settles the payment in USDC on Base, and the Actor runs. Built on x402 by Coinbase.

译Apify与Coinbase合作，将超过20,000个Web自动化Actors接入x402生态系统。AI智能体可自主发现、支付并运行这些工具：调用Actor时收到HTTP 402状态码，通过Base链上的USDC完成支付后立即执行。此前x402生态仅有约2,000个工具（来自@apify），此次合作将可用工具数量提升10倍，无需账户、API密钥或人工介入。

歸藏(guizang.ai)@op7418 · 2天前33

CodePilot 0.5.6.3 已经更新。 今天新上 ClinePass 以及之前的 Opencode GO，这两个 Codeplan 套餐，想体验的可以试试

译CodePilot 0.5.6.3已更新，新增ClinePass和Opencode GO两个Codeplan套餐。此外，Cline推出每月9.9美元的token plan，支持DeepSeek、Kimi、MiniMax、MIMO、通义千问以及GLM5.2，目前还有1.99美元的优惠价。这类聚合token plan在海外逐渐增多，便于调用多家推理服务。

Chubby♨️@kimmonismus · 2天前47

Etched is coming out of stealth with $800M raised, $1B+ in customer contracts, first racks shipping this summer, and claims of SOTA inference throughput, latency, and power efficiency. But holy, look who backed the funding. The who-is-who if AI reseracher and VC.

译Etched正式走出隐身模式，宣布已融资8亿美元，并签下超过10亿美元的客户合同。成功完成A0流片后，首批推理机架已制造完成，预计今年夏天发货。早期客户测试显示，其在推理吞吐量、延迟和能效方面均达到SOTA水平。投资方阵容堪称AI研究员与VC的全明星阵容。

elvis@omarsar0 · 2天前53

The gap in autonomous agentic loops that gets ignored: agents can plan and call APIs but can't acquire tools they don't have access to. x402 + Apify's 20,000+ Actors is a concrete fix for that. Worth paying attention to.

译自主智能体可规划和调用API，但无法获取未授权工具。x402协议与Apify的20,000+个Actors解决了这一缺口。此前智能体仅可通过x402购买约2,000个工具，如今Apify与Coinbase合作，将其10倍扩展至20,000+个，为自主智能体提供最大的网络自动化工具市场。无需账户、API密钥或人工介入。

meng shao@shao__meng · 2天前74

Flowith 团队推出「Matrix」：Agent 公司的操作系统，你定使命，Matrix 编排多 Agent 部门长期运转，目标是从创建、分发到变现的完整商业闭环。 Matrix 的核心主张 · 产品形态：自演化、多层级的 multi-agent runtime · 用户角色：战略负责人（设 mission），不是日常执行者 · 组织模型：CEO Office → OKR → 部门（Research / Engineering / Growth / Product）→ 证明与复盘 · 商业闭环：建站、接 Stripe、发邮件、投广告、产内容、收 revenue · 新指标：VPTD（Value Per Token Dollar）= 产出价值 ÷ token 成本 产品架构 1. Runtime 层 每个 Agent 有独立 browser、工具、文件、记忆；支持 Neo / Claude Code / Codex 等，强调 超长时运行 和 主动执行（不是一问一答）。 2. 协调层 用户输入 intent + 资产 → CEO Office 定目标与节奏 → OKR 分解任务 → 各部门并行 → 以 proof（文件、截图、上线页面、收入、流量）闭环。 3. 公司原语（Company Primitives） 内置：网站部署（*.matrix.site 或自定义域名）、Stripe 收款、Agent Wallet（预算与审批）、Agent Email。宣称可跳过传统公司注册、银行卡、域名等 setup。 4. 交付形态 目前 macOS 客户端，Web 版「coming soon」。 GDPval-Bench 数据 1. Matrix（GPT 5.5 + harness）：95.45% 2. Codex CLI（GPT-5.5）：84.9% 3. Claude Opus 4.7：80.3% GDPval 是 OpenAI 等提出的 benchmark，测的是 44 个职业、真实知识工作交付物（法律文件、工程图、客服对话等）

译Flowith 团队发布「Matrix」，定位为自演化、多层级的 multi-agent runtime。用户设定使命后，Matrix 通过 CEO Office → OKR 分解任务，驱动多个 Agent 部门（Research/Engineering/Growth/Product）并行执行，并以 proof（文件、上线页面、收入等）闭环。架构包含 Runtime 层（独立 browser/工具/记忆，支持 Neo/Claude Code/Codex，超长时运行）和公司原语（网站部署、Stripe 收款、Agent Wallet、Agent Email）。macOS 客户端已上线，Web 版 coming soon。在 44 个职业真实知识工作的 GDPval-Bench 上，Matrix（GPT 5.5 + harness）得分 95.45%。此前有限 beta 中用户已创建数万个零人公司。

elvis@omarsar0 · 2天前64

http://x.com/i/article/2071684582336782336 # FW Serverless 2.0: The Routing Pattern GLM 5.2 has kept open-weight models in the conversation and has everyone wondering how to start leveraging these open models in production. Once you move open models into production, the first thing that breaks under load is not output quality. It is whether the request is served at all. When traffic across the shared fleet exceeds available capacity, Fireworks can reject the request before generation and return a 503 Service Overloaded. The traditional fix has been to buy capacity ahead of time, either reserved GPUs or an enterprise contract sized to your peak. That leaves two bad options. Over-provision for traffic you rarely see, or guess low and eat failures when a spike arrives. Fireworks Serverless 2.0 (@FireworksAI_HQ) turns that standing capacity decision into a per-request routing decision. Each call can select the serving tier that handles it, so reliability becomes a runtime control instead of a procurement decision. The pattern below keeps live traffic available during congestion without reserving GPUs up front. ## The three serving tiers Serverless 2.0 gives you three serving tiers behind one API and one endpoint. Fig. 1. Three synchronous serving paths share one API surface and one fleet. Priority is selected with service_tier, while Fast uses a Fast model ID. Source: Fireworks Serverless 2.0 announcement. - Standard for everyday traffic. This is your default for production calls. It runs on elastic shared infrastructure and is the most cost-efficient path. Under high platform load, Standard requests are the first to be queued or rejected. - Priority for reliability under load. Reach for it when a dropped request has real cost, like an interactive session or a long agent run. It gets stronger admission during congestion and is shed last, at a higher per-request price than Standard. - Fast for latency-sensitive generation. Use it when wall-clock generation time is the bottleneck, such as agent loops, coding workflows, and interactive apps. Fast uses the same model family through an optimized serving path for higher generated-token throughput, not a smarter model or a different reasoning tier. Same API surface, no capacity reservation. You choose one serving behavior per request. Leave the default model on Standard, add `service_tier="priority"` for stronger admission during congestion, or switch to a Fast model ID for higher generated-token throughput. Priority and Fast solve different problems and are not stackable on one request. Take a concrete case. A chatbot runs fine on Standard until a launch drives a traffic spike and Standard starts returning 503s. Instead of provisioning GPUs or putting users behind a queue, you add `service_tier="priority"` on that endpoint, keep serving through the spike, and switch back to Standard once it passes. ## When to switch tiers You do not pick Standard or Priority up front. You default to Standard all day, and the moment a request gets shed under congestion (a 503 Service Overloaded, not a rate-limit 429), you flip to Priority for the next 30 minutes, then drift back. Fig. 2: The escalation policy. Default to Standard, flip to Priority for a 30-minute window on a 503 Service Overloaded, then drift back to Standard once the window expires. The premium is a control-plane tradeoff, not a new architecture. Priority costs more than Standard for the requests that use it, so the point is to promote only the traffic where a failed request has user-visible or workflow-visible cost. Interactive endpoints and long agent runs get the escalation path. Batch jobs should use Standard, the Batch API, or Background serving when retries and queueing are acceptable. Use Priority only when a 503 would waste expensive multi-step work. ## The code The code below is illustrative — written to demonstrate the documented Serverless 2.0 pattern, not an official Fireworks code sample. The `service_tier="priority"` field and the 503 Service Overloaded signal are from the Fireworks docs. The control loop, including the 30-minute window and `priority_until` bookkeeping, is our recommended implementation. The important part is the scope of the fallback. Escalate on 503 because that indicates serving capacity pressure. Do not use the same branch for 429 rate limits, auth errors, invalid requests, or application exceptions. Those are different failure modes and should not silently move traffic into a higher-priced tier. ## Guardrails to set - Track priority_until, escalation count, and 503 rate in metrics so you can see when Priority is masking sustained load. - Keep the escalation window bounded. A 30-minute window is enough to ride through a spike without leaving the service permanently promoted. - Apply the policy per workload or per route. User-facing paths can be promoted to Priority on 503. Evals, offline jobs, and other async batch workloads should use Standard or Background unless a failed request wastes expensive progress. - Alert if Priority remains active for multiple windows in a row. That is a capacity or traffic-shaping signal, not just a transient failover. ## What Priority costs Use the Serverless pricing docs as the source of truth. In the current pricing table, Kimi K2.7 Code Priority is listed at 1.5x the Standard row, while Kimi K2.7 Code Fast is listed as a separate Fast model ID at 2x Standard. Pricing varies by model, so always keep the docs as the reference. The operational point is simple. If a worker needs Priority for a 30-minute congestion window, that +50% per-token premium can still be a useful tradeoff when the alternative is failed multi-step work. For broader cost framing, refer to this article, which reports open-worker plus advisor setups running 19% to 67% cheaper than Opus-as-worker across its benchmark table. ## Which tier for which workload The pattern matters in the three places AI devs actually ship. Fig. 3. Routing by workload type. Batch and offline work routes to Standard or Background when retries are acceptable. Fast remains for latency-sensitive generation when wall-clock time is the bottleneck. - User-facing chat and agents. Interactive traffic is latency-sensitive and bursty. Keep it on Standard and let the first 503 during a spike (a launch, a viral post) auto-escalate to Priority, so users get answers instead of errors and you are not babysitting a dashboard. - Long agent runs. A single agentic task fans out into dozens of dependent calls, and one shed request mid-chain can sink the whole run. Escalating to Priority after the first 503 protects the expensive, multi-step work where a retry is not free. - Batch and offline jobs. Evals, synthetic data, bulk embeddings, nightly summarization, report generation, offline analysis, and data enrichment usually care more about throughput and completion cost than instant response time. Keep these on Standard or Background when retries and queueing are acceptable. Use Priority only when a 503 would waste expensive multi-step work. Leave Fast for latency-sensitive generation paths where wall-clock time is the bottleneck. Because the switch is per call, you run these paths off one codebase. Live endpoints can default to Standard with the escalation guard, long-running workflows can promote to Priority when 503s threaten completion, and async workers can stay on Standard, Batch, or Background. No separate clusters, no separate SDKs. ## Reliability without the cluster Serverless 2.0 gives teams more room before they need dedicated capacity. Start on Standard, add Priority when overload behavior matters, switch to Fast when wall-clock latency matters, and reserve capacity when you need hard guarantees. ## Links - Sign up - Docs - Serverless 2.0 announcement (tiers, the service_tier parameter, and 503 behavior) - Coding-model pricing comparison

译Fireworks AI 推出 Serverless 2.0，通过同一 API 端点下的三种服务层级解决共享集群高负载时的 503 Service Overloaded 问题。Standard 为默认经济型；Priority 在拥塞时提供更强准入，价格更高；Fast 通过优化路径提升生成 token 吞吐量，适用于低延迟场景。推荐默认使用 Standard，遇到 503 时临时切换 Priority 30 分钟，随后自动回退。Priority 和 Fast 不可叠加。

PixVerse@PixVerse_ · 2天前28

This hospital hallway just served me a jump scare I did not order. My controller is now MIA. Seedance 2.0 on PixVerse. Native 4K horror, straight from prompt.

译这条医院走廊刚给我来了个我没点的跳吓。我的手柄现在不见了。 Seedance 2.0 上线 PixVerse。原生 4K 恐怖内容，直接通过提示词生成。

🚨 AI News | TestingCatalog@testingcatalog · 2天前35

Bloome launched its instant messaging platform for agentic teams! Agents can draft, push back on one another, cross-check details, and refine the output until it is ready. Models like Claude, ChatGPT, and DeepSeek can run side by side with coding agents as well as custom agents built in @Bloome_im

译Bloome 推出了面向智能体团队的即时通讯平台！ 智能体可以互相起草、反驳、交叉核对细节，并不断完善输出，直到准备就绪。Claude、ChatGPT 和 DeepSeek 等模型可以与编码智能体以及 @Bloome_im 中构建的自定义智能体并排运行。

X.PIN@thexpin · 2天前50

UBTech has unveiled life-like robots! It’s their first consumer humanoid robot under the new UWORLD brand: the U1, a full-size humanoid designed for emotional companionship. Available in male (183 cm) and female (168 cm) versions, the robot features 88 degrees of freedom, 2–4 hours of battery life, local encrypted memory, and an on-device emotional AI model trained with Huawei’s Ascend framework. UBTech says the U1 has already received more than 11,000 pre-orders, with deliveries expected to begin in mid-September. They do look a lot like C-Drama characters 😂 If their interactions are also this life-like, my sorry isolated ass may actually consider buying one.

译优必选（UBTech）推出新品牌UWORLD首款消费级人形机器人U1，主打情感陪伴。提供男款（183cm）与女款（168cm）两个版本，具备88个自由度、2-4小时电池续航、本地加密存储，搭载基于华为昇腾框架训练的端侧情感AI模型。该机器人已获超11,000份预订单，预计9月中旬开始交付。

PixVerse@PixVerse_ · 2天前24

When you think you have outrun your problems… turns out it is a T-Rex. 💀🦖 Seedance 2.0 4K — your prompts turned into cinematic terror.

译当你以为已经摆脱了烦恼……结果发现是一只暴龙。💀🦖 Seedance 2.0 4K — 你的提示词变成了电影级恐怖。

AYi@AYi_AInotes · 2天前63

X 官方 MCP也来了，可以让 Grok、Cursor、Claude 等 MCP 兼容的 AI 工具，能零部署直接调用 X API，获取实时 X 数据（搜索、时间线、书签、发文等），全部走用户权限。 但不是完全免费喔，Grok Cursor 以及所有兼容 MCP 的工具，零部署就能直接调用 X API。 但它不是免费的，走的是 X Developer App 的按量计费模式，读帖读用户发帖各有单价，重度 Agent 使用很容易账单失控，写操作的限流也比读操作严格得多，很容易触发 429。 最稳妥的做法，先进开发者后台设好消费上限，小流量测清楚价值再考虑放量

译X 官方推出托管 MCP，使 Grok、Cursor、Claude 等 MCP 兼容 AI 工具无需部署即可直接调用 X API，获取搜索、时间线、书签、发文等实时数据，全程走用户权限。该服务并非免费，采用 X Developer App 按量计费模式，读帖和发帖各有单价，重度 Agent 使用容易导致账单失控；写操作限流比读操作更严格，易触发 429 错误。建议开发者先在后台设定消费上限，从小流量测试后再放量。

向阳乔木@vista8 · 2天前64

如何管理和监控一个超过 10,000 块 GPU 的集群？ 腾讯团队开源了一个叫ARGUS的方案，强！ 大模型训练极其贵，一个万卡集群，一天的电费和硬件折旧可能就是几十万人民币。 如果某节点出问题，或通信带宽被浪费了，损失巨大。 ARGUS 解决的核心问题：当集群出问题，能不能在几分钟内找到原因。 论文发现：在万卡规模下，超过 70% 的训练中断，是由网络通信问题导致的，而不是 GPU 本身坏了。 网络拓扑、路由策略、甚至网线质量，都会成为瓶颈。 ARGUS 的做法是：采集每个 GPU 的实时数据，包括计算负载、显存使用、网络带宽、通信延迟，然后自动做关联分析。 一旦发现异常，直接定位到具体是哪块 GPU、哪条链路出了问题。 原始论文见评论区

译腾讯团队开源ARGUS方案，用于管理和监控超10,000块GPU的集群。大模型训练成本极高，万卡集群每天电费和折旧达数十万元。ARGUS解决的核心问题是在集群出问题时几分钟内定位原因。论文发现：万卡规模下，超70%训练中断由网络通信问题导致，而非GPU故障。ARGUS通过采集每GPU实时数据（计算负载、显存、网络带宽、通信延迟）进行关联分析，精准定位故障GPU或链路。

Xiaomi MiMo@XiaomiMiMo · 2天前54

Open-weights are taking over — proud to see more developers building on MiMo. 👏 @cline

译我们对 GLM-5.2 印象深刻，因此推出每月 $9.99 的订阅，让你能以 2-5 倍折扣访问它以及其他开源权重模型，如 DeepSeek、Kimi、MiniMax、Mimo、Qwen。可通过 Cline CLI 和 IDE 使用，通过 `npm i -g cline` 注册可享受 $1.99 特价促销。

小互@xiaohu · 2天前62

OpenClaw 推出了自己的手机客户端 • 通过二维码或设置码与你的 小龙虾 配对 • 在手机上和 A小龙虾聊天 • 支持实时和后台语音对话模式 • Agent执行操作前，先在手机上向你确认审批 • 直接把文字、链接、图片从其他 App 分享进来 • 可授权摄像头、定位、照片、通讯录、日历、提醒事项等设备权限 • 接收推送通知和节点状态更新

译OpenClaw 推出手机客户端，可通过二维码或设置码与 AI 助手“小龙虾”配对。支持在手机端实时及后台语音对话；Agent 执行操作前需在手机上确认审批；可跨 App 分享文字、链接、图片；授权摄像头、定位、照片、通讯录、日历等设备权限；接收推送通知与节点状态更新。

歸藏(guizang.ai)@op7418 · 3天前75

Twitter 终于有 MCP 了，这个非常有用！ 如果你像我一样依赖 Twitter 信息的话，就可以让 AI 帮你自动整理、总结和对你的数据进行分析 不过依然需要你注册 Twitter 的 API，然后按量付费。 我今天早上配置了一下，它针对个人信息（就是你自己的信息）其实有一个比较大的优惠： 调用一次是 0.01 美元，也就是 1 美元可以调用 1000 次。我拉取了一下最近三天的书签，只花了 0.1 美元。 我大致做了几张图，简单写一下如何配置： 1. 创建 APP 与充值：去 Twitter 开发者后台创建一个 APP。注意，充钱是必须的。 2. 获取配置 ID (a) 创建完 APP 后，从侧边栏点进你创建的这个 APP。 (b) 在设置里填写并修改相关内容。 (c) 修改完成后，系统会弹窗显示两个 ID，这两个 ID 是需要在 MCP 里填写的，一定要记好。 3. 辅助配置 (a) 我的那张图片发给 Codex 或者 Claude，让他们帮你配置。 (b) 配置完后，让他们打开配置文件，你把那两个 ID 替换掉就可以了。 4. 授权启动 在启动时，第一次需要打开网页进行授权。如果你的 AI 没有自动打开网页，可以提醒它一下。

译X（Twitter）官方推出 hosted X MCP，AI 智能体可通过 MCP 协议直接调用 X API 获取实时信息，支持 Grok、Cursor 等工具。用户需注册 X API 并按量付费，个人优惠价每次调用 0.01 美元（1 美元 1000 次）。有用户实测拉取近三天书签仅花 0.1 美元。配置步骤：创建 APP 并充值、获取配置 ID、辅助配置（可交给 Codex/Claude）、授权启动。

AYi@AYi_AInotes · 2天前66

Hermes独霸移动端的日子要结束了，OpenClaw直接把私有云AI助手搬进了你的手机，数据完全隔离不进训练集。 这款工具现已原生支持iOS与安卓双端，所有运行在私有云容器中，文件存储在个人云工作空间，全程不会将你的数据用于LLM训练。 免费套餐每天提供20条由Gemini驱动的AI消息，每月20美元的订阅即可解锁无限使用权限。 过去几个月Hermes占据了移动端Agent的大部分注意力，主打隐私隔离的OpenClaw，相当于给用户多了一个完全不同取向的选择。 你选AI助手时，会把数据隐私放在第一位吗。

译OpenClaw推出原生iOS和Android应用，将私有云AI助手搬进手机。所有Agent运行在私有云容器中，文件存储于个人云工作空间，数据完全隔离不进LLM训练集。免费套餐每天提供20条由Gemini驱动的AI消息，每月20美元订阅可解锁无限使用权限。与移动端Agent Hermes相比，OpenClaw提供隐私优先的差异化选择。

Rohan Paul@rohanpaul_ai · 3天前40

Satya Nadella just talked about a new AI Dev Box (RTX Spark Dev Box) with 20 CPU cores, 128GB unified memory, and petaflop of local compute for trillion-parameter-scale model runs. ---- From "Stanford Online" YouTube channel, (link in comment)

译Satya Nadella 刚刚谈到了一款新的 AI Dev Box（RTX Spark Dev Box），配备 20 个 CPU 核心、128GB 统一内存和千万亿次本地计算能力，用于运行万亿参数规模的模型。

歸藏(guizang.ai)@op7418 · 3天前65

Cline 推出了一个每月 9.9 美元的 token plan, 支持调用 DeepSeek、Kimi、MiniMax、MIMO 和千问，还有 GLM5.2 现在好像还有 1.99 美元的优惠。 海外这种聚合的 token plan 也越来越多了，而且他们如果有渠道的话，更方便去推理服务的供应商也不缺卡

歸藏(guizang.ai)@op7418 · 3天前51

Codex 因为消耗速度过快的问题又重置了

译OpenAI 的 Codex 因使用量消耗异常再次重置所有用户使用限制，并额外补偿一次重置（未来 24 小时可用）。调查发现是多个小问题叠加：Auto-review 过度主动、变更导致更多子智能体工作、后台建议重复运行或失败重试过于频繁。已回滚相关变更并修复调度与重试行为，减少不必要 token 消耗。此外，Auto-review 此前被错误显示为 GPT-5.4 的使用量，失败与限流请求也被计入 turn 图表；现在 Auto-review 归入独立类别，仅成功请求计入。修复已通过 CLI、桌面应用和用量后端热更新部署，历史图表中旧数据仍显示为 GPT-5.4。

歸藏(guizang.ai)@op7418 · 3天前30

Open AI 要出 Codex 定制的 Vibecoding 键盘了 跟 Work Louder 合作，型号应该是 Creator Micro 2，做了些外观的变动。 键盘包含 13 个机械按键和一个旋钮、一个摇杆、触摸传感器，支持RGB 背光 + RGB 底部灯光。 推动摇杆即可弹出屏幕上的圆形快捷菜单，完全可自定义。6 层映射，每层包含 13 个按键 + 7 个径向菜单槽位 + 旋钮映射，总共可配置大量快捷方式。 Creator Micro 2 现在的售价是 144 美元起，不知道 OpenAI 这个会卖多少钱

译OpenAI 与 Work Louder 合作推出基于 Creator Micro 2 外观定制的 Codex 键盘，配备 13 键、旋钮、摇杆、触摸传感器，摇杆弹出圆形快捷菜单，支持 6 层映射。标准版售价 144 美元起，定制版价格待公布。7 月 15 日 Codex 快捷键将升级。

meng shao@shao__meng · 3天前69

Codex 团队 @Dimillian 也是一位知名 iOS/Swift 开发者，把他的 Codex Skill Manager 开源了！ 这是一个用 SwiftUI + SwiftPM 原生构建的 macOS 应用，作为 Codex Skills 的图形化管理器和商店，把 Codex、Claude Code、OpenCode 和 Copilot 的四套 Skills 目录抽象成 SkillPlatform 枚举，让一个 GUI 统一管理跨平台的 Skills。 https://github.com/Dimillian/CodexSkillManager

PixVerse@PixVerse_ · 3天前11

Bushido is beauty and resolve and it's all in this single strike！

译武士道是美与决心，凝聚在这一击之中！

歸藏(guizang.ai)@op7418 · 3天前61

Cursor 的 iOS 版上线了，完成度很高啊 锁屏的时候，还会展示当前的这个进度，然后也完成以后，会把界面的视频和图片给你发过来，让你审核 当前，针对付费用户的 beta 测试已经开放

Berryxia.AI@berryxia · 3天前76

Cursor iOS 版本正式来了，这下可以随时随地Vibe Coding了啊！ 还有75折的优惠使用Composer 2.5 模型！

Tibo@thsottiaux · 3天前43

Codex usage limits will be fully reset again in the next hour and we will credit one additional reset into your bank for your own usage over the next 24 hours. We investigated reports that Codex usage was being consumed faster than expected. There wasn't one central issue, but a few smaller problems compounded for some users. Here's what we found and changed: - Actual usage: Auto-review had become more proactive, another change was triggering more subagent work, and background suggestions could run twice or retry too frequently after failures. We reverted the changes and fixed suggestion scheduling, duplicate generation, and retry behavior. This should reduce unnecessary background token consumption while preserving the work users explicitly request. - Usage reporting: Auto-review was incorrectly appearing as GPT‑5.4 usage, and failed or rate-limited requests were still shown as turns. Auto-review now appears as its own category, and only successful requests count toward the turn graphs. Rate-limited requests were never charged, but they were being displayed incorrectly. - Immediate relief: We reset usage limits while rolling out the fixes, then shipped hotfixes across the CLI, desktop app, and usage backend. - What to expect: New usage data should be clearer and actual consumption should be lower. Historical charts may still show auto-review under GPT‑5.4 because older turn data was not relabeled. Features that intentionally perform more work; such as /goal, subagents, and higher reasoning levels will still naturally use more capacity. All fixes are now deployed, and we've added more detailed monitoring so we can detect background-usage regressions sooner. We'll continue watching the results closely. Thank you for building and doing all sorts of things with Codex.

译Codex 使用限制将在下一小时完全重置，额外一次未来24小时重置。消耗过快并非单一原因，而是自动审查更主动、某些改动触发更多子Agent、后台建议重复生成或失败后重试等小问题叠加。现已回滚相关改动，修复调度、重复生成和重试行为，减少不必要的后台token消耗。使用报告方面，自动审查此前被错误归类为GPT-5.4使用量，已单独分类；仅成功请求计入轮次图表。所有修复已部署，并增加更详细监控。团队周日紧急排查日志。

Runway@runwayml · 3天前56

Seed Audio 1.0 is now available on Runway for all paid plans. Generate up to 120 seconds of dynamic speech, sound design and music with simple text prompts. Try it now at the link below.

译Seed Audio 1.0 现已在 Runway 上对所有付费方案开放。只需简单文本提示即可生成最长 120 秒的动态语音、音效设计和音乐。立即通过下方链接尝试。

Tibo@thsottiaux · 3天前24

Do you think it has a reset button?

译你最喜爱的 Codex 快捷键即将升级，7 月 15 日推出。主推文问道：“你觉得它有重置按钮吗？”

🚨 AI News | TestingCatalog@testingcatalog · 3天前62

OpenClaw released its own mobile apps for iOS and Android. Seems like it would cover loads of features and follow the UI design of the web version. Gotta try that 👀

译OpenClaw 正式发布 iOS 与 Android 原生移动应用，支持智能体（Agents）、频道、任务和回复等核心功能，采用与网页版一致的 UI 设计。用户可在手机上随时随地运行 AI 智能体。

OpenClaw🦞@openclaw · 3天前71

OpenClaw is now on iOS + Android 🦞 📱 Native mobile apps, finally 💬 Agents in your pocket 🔔 Channels, tasks, replies on the go Run agents from wherever your thumbs are. iOS: https://apps.apple.com/us/app/openclaw-ai-that-does-things/id6780396132 Android: https://play.google.com/store/apps/details?id=ai.openclaw.app

译OpenClaw 现已登陆 iOS 和 Android 🦞 📱 原生移动应用，终于来了 💬 智能体装进口袋 🔔 频道、任务、回复，随时处理 用你的大拇指，在任何地方运行智能体。 iOS: https://apps.apple.com/us/app/openclaw-ai-that-does-things/id6780396132 Android: https://play.google.com/store/apps/details?id=ai.openclaw.app

Emad@EMostaque · 3天前70

We have seen multi model harnesses for cheaper & faster tasks What about for the hardest challenges? What about open source? Proud to share the latest update our Zenith harness, taking models you can use today above Fable on tasks that take hours or days

译ii 团队推出开源 Zenith harness，通过自适应自我改进（adaptive self improvement）将基础模型推向 FrontierSWE 基准榜首，在需数小时或数天的复杂任务（如蛋白质预测模型训练、编译器优化）上超越 Fable。同时预告 GLM 5.2 即将到来。

Chubby♨️@kimmonismus · 3天前37

Really excited for the next version of Claude Code. Keep it coming Boris!

译下个版本的 Claude Code 中，子智能体将默认在后台运行，让你在子智能体工作时仍可继续与 Claude 对话。若想让智能体前台运行，只需告诉 Claude。Kim 对此表示非常期待，并喊话 Boris 继续发力。

Chubby♨️@kimmonismus · 3天前67

This is the first "AI company" product I've seen that doesn't feel like pure cosplay. Two interesting points: Matrix treats the company idea seriously. You are not just creating agents and hoping they coordinate. Matrix beat both Codex and Claude Code on GDPval-Bench, with 95.45% against 84.9% and 80.3% respectively. That gap seems to matter most on longer tasks, where planning and coordination actually decide the outcome rather than raw model capability. Which is maybe the point. A lot of "AI companies" are really just prompt orchestrators with a nice UI. Matrix looks like it's building something closer to an actual operating layer. Whether that holds up beyond benchmarks, I don't know yet. But it really makes me want to find out.

译Matrix 被 Kim 称为首个「不像 cosplay」的 AI 公司产品。它在 GDPval-Bench 上以 95.45% 的得分击败 Codex (84.9%) 和 Claude Code (80.3%)，长任务差距说明规划和协调比原始模型能力更关键。Matrix 定位为运行「零员工公司」的运行时，而非简单提示编排器。上周有限 beta 期间用户已创建数万个零员工公司并开展真实业务，即日起向所有人开放公测。

Rohan Paul@rohanpaul_ai · 3天前46

This is really cool. AI is now coming for the most broken search problem on the internet: finding the right person. @CLODOAI is moving the intelligence back to the search query: Describe the person in plain English, search across 1.2 billion indexed profiles, qualify the match, enrich the contact, and surface a way to reach them. People discovery is finding the right humans for whatever you're working on. But most current people search tools sell you access to a snapshot. A frozen moment in time with a CSV at the end of it. Clodo searches the live web, reads the signals, qualifies the fit, and tells you why this specific person, right now. @sixtensor @vikchuppala 👏

译AI人物搜索工具CLODO上线，支持用自然语言描述目标人物，在12亿（1.2B）索引档案中实时搜索、匹配、丰富联系人信息并生成触达方式。与输出静态CSV的传统搜索不同，CLODO检索实时网络信号，判断匹配度并解释推荐原因。该工具已获得Mercor、US News、Corgi以及美国情报界等机构的信任，并获得Y Combinator、CRV等投资。

elvis@omarsar0 · 3天前18

Finally, a people-search tool that actually works. Most people-search tools sell you a frozen list. @CLODOAI searches the live web, reads the signals, and tells you why this specific person, right now. Good stuff, @sixtensor @vikchuppala 👏

译DAIR.AI 的 Elvis Saravia 称赞 CLODO 是真正有效的人搜索工具：不同于传统工具提供固化列表，CLODO 搜索实时网络、读取信号并解释为何找到此人。CLODO 由 @sixtensor 推出，用户描述目标即可每次准确找到，覆盖 12 亿个人资料，并提供资料丰富及主动联系服务。已被 Mercor、US News、UseCorgi、美国情报界等采用，获得 Y Combinator、CRV、Dalton Caldwell、Paul Thiel 等投资。

Andrew Milich@milichab · 3天前56

Try SpaceXAI Voice models in the Vercel AI Gateway

译xAI Grok 的实时语音模型现已登陆 Vercel AI Gateway，支持实时对话、语音合成与语音转写。欢迎试用。

Claude@claudeai · 3天前55

Claude in Microsoft Foundry is now generally available, hosted on Azure. Azure customers get Claude Opus 4.8 and Claude Haiku 4.5, with Azure authentication, billing, and commitment retirement.

译Claude 现已在 Microsoft Foundry 中正式可用，托管于 Azure。 Azure 客户可获得 Claude Opus 4.8 和 Claude Haiku 4.5，并支持 Azure 身份验证、计费和承诺预留。

🚨 AI News | TestingCatalog@testingcatalog · 3天前67

Update: Cursor iOS app is now available in EU as well!

译更新：Cursor iOS 应用现已在欧盟地区可用！

Google Gemini@GeminiApp · 3天前54

More users in the U.S. can now create personalized images for free. By choosing to connect your Google apps to Gemini in settings, Gemini is able to better understand your preferences, so it can automatically fill in the blanks and ground image creation in the things you care about most. You can choose to connect or disconnect your Google apps at any time, so you’re in control. Try it out today and share what you create in the replies 👇

译Gemini 宣布更多美国用户可免费创建个性化图像。用户通过设置连接 Google 应用后，Gemini 的“Personal Intelligence”功能可理解用户偏好，自动填充空白，基于用户关心的内容生成图像。用户可随时连接或断开 Google 应用，完全掌控隐私。

AYi@AYi_AInotes · 3天前68

Cursor iOS端上线了， 中国的iOS用户今晚可以甜甜入睡了😴

译Cursor for iOS 正式发布，支持启动云端智能体，或从 App 远程控制电脑上的智能体。Composer 2.5 即日起至 7 月 5 日享 75% 折扣。 中国的iOS用户今晚可以甜甜入睡了😴

Chubby♨️@kimmonismus · 3天前25

Try typing "the founder who wrote one great paper in 2021, left to build something, and would actually reply to a cold email" into LinkedIn search. You get a dropdown for job title and company. That's it. Everything that actually matters gets ignored. Clodo (@clodo_ai) takes that whole sentence, searches 1.3 billion profiles across the live web, finds who fits, and gives you back verified emails and the signals it found.Kind of scary how well it works. Also really useful.

译传统LinkedIn搜索只能匹配职位和公司，忽略自然语言中的关键信号。Clodo（CLODO）则能直接理解一段完整的描述，在13亿个实时网上个人资料中搜索，找到匹配对象，并返回验证过的邮箱及匹配证据链。该工具已获@mercor_ai、@usnews、@UseCorgi及美国情报界等机构使用，由Y Combinator、CRV等投资方支持。