Euro Vibe Coding Standout Lovable Now Valued at $13.3B
The fast-growing vendor has attracted significant international funding.
AI BusinessOriginal ↗
Agent 行业需要一个 Android。 作者｜张勇毅 编辑｜靖宇 北京时间 8 月 13 日晚上八点半，DeepSeek 正式公布了它成立以来的第一个 Agent 产品，之前预热很久，大家期待值很高的 Deepseek Harness。 截至发稿，它的 GitHub 仓库已经涨到超过 5 万个 star—— 而仓库公开刚刚过了 12 个小时。 Deepseek Harness GitHub 主页截止发稿时已突破 5 万 star｜图片来源：GitHub DeepSeek 官方给它的定义其实和别家也差不多：模型加上 Harness，才等于 Agent。模型是脑子，Harness 是手脚；聊天机器人交付的是一段话，Agent 交付的是一件做完的事。 发布当晚，我们把它装进了自己的电脑，让它干了两件活：照着 The Verge 的风格重构极客公园官网，再调 GitHub 的接口，画出它自己的涨星曲线。两件活都交付了，全程花掉的 token 不到 3 块钱。 但它也确实还是个毛坯：界面对不写代码的人算不上友好，目前上传的开发者预览版的毛边随处可见。12 个小时用下来，我们的印象是——Dee…
蚂蚁百灵与 ASystem 团队合作，用 Ling-3.0-tiny 和 AReno 在 DGX Spark 上跑通单机 Agentic RL 后训练闭环。以井字棋为最小验证任务，用 GSPO 算法训练 400 步后，rollout/rewards_mean 从约 -0.5 升至 0.4，response_len 降至约 850 tokens，工具调用与动作选择趋于稳定。 🔗 阅读原文 via AIHOT · https://aihot.virxact.com/items/cmssf79uf05rwrod09r9ocb2w
对标 Claude Cowork：DeepSeek Harness 公测，同步开放插件生态 8 月 13 日消息，DeepSeek-V4-Pro-0813 正式开源，同时还推出了其开源（MIT 协议）的代码智能体框架 Harness 的 v0.1 开发者预览版（基于 Cordis），并同步开放了配套的插件生态系统。相关 GitHub 仓库也已公开。 DeepSeek 正式开源 DeepSeek‑V4‑Pro‑0813 模型，同步推出基于 Cordis、采用 MIT 协议的代码智能体框架 Harness v0.1 开发者预览版，并开放配套插件生态，GitHub 仓库现已公开，开发者可通过 npx 命令或源码两种方式启动 WebUI，该产品定位对标 Claude Cowork 与 OpenAI Codex，面向编程办公 AI 生产力场景。 Harness 贯彻「一切皆插件」的设计理念，依托 Cordis 元框架，模型、工具、UI、沙箱等全部 Agent 能力都以插件形式存在，无需修改源码就可以自由替换扩展。框架提供 dsh 命令行工具，拥有标准、PTC 程序化工具调用、极简、创造四种运行模…
The fast-growing vendor has attracted significant international funding.
SpaceXAI's new model focuses on long-running tasks, while remaining competitively priced compared to other frontier models.
GeForce NOW is giving cloud gaming an extra-credit upgrade just in time for back-to-school season. The native Linux app for GeForce NOW is officially out of beta. GeForce NOW is also delivering new cloud optimizations that make Frame Generation feel even more responsive while streaming. On top of that, Performance members will see higher frame […]
Learn how startups use GPT-5.6 to build faster, more cost-efficient AI agents with smarter model selection and new Responses API capabilities.
Preview Ultrafast, a new OpenAI API service tier that runs GPT-5.6 Sol up to 14× faster. Powered by Cerebras, it delivers up to 750 output tokens per second.
OpenAI appoints Dali Rajic as Chief Revenue Officer to lead its global revenue organization and help businesses realize the full value of AI.
作者｜桦林舞王 编辑｜靖宇 没有预告，没有倒计时，甚至连更新日志都没来得及写好。DeepSeek V4 Pro 的正式版，就这么悄悄的来了。 8 月 13 日凌晨，DeepSeek 官网悄悄刷新了 API 文档。模型名还是那个 deepseek-v4-pro，但 fingerprint 已经变成了 fp_v4pro_20260812。打开社区一看，消息已经炸开—— V4 Pro 正式版，来了。 从 4 月 24 日预览版上线算起，整整 111 天。 这期间 Kimi K3 高调发布抢走了开源头条，V4 Flash 正式版先行一步在 7 月 31 日上线，API 涨价的预告更是悬在所有开发者头顶。 所有人都在问同一个问题：Pro 的正式版到底什么时候来？ 答案是一个周三的深夜。 01 架构没变，能力暴涨 先说硬参数。V4 Pro 正式版的模型架构和预览版完全一致——1.6T 总参数、49B 激活参数的 MoE 结构，支持 1M 上下文、384K 最大输出。 变的是后训练，而且变化大到像换了个模型。 最夸张的数字来自 Agent 相关评测。DeepSWE（DeepSeek 自家的软件工程基…
Pixel 11 不是一台想赢的手机，它是 Gemini 的身体。 作者｜张勇毅 编辑｜靖宇 纽约时间 8 月 12 日下午，谷歌在布鲁克林开完了今年的 Made by Google 发布会。 Pixel 11 全系四款手机、Pixel Watch 5，加上防丢器 Pixel Tag，一口气全发了，当天就开预购。 热闹是足够热闹。但两个小时看下来，最能概括这场发布会的，是 Pixel 11 背面新增的那一小圈灯。 本次发布的全系 Pixel 新品｜图片来源：Google 全系背面都多了这圈环形 LED，在 Pro 机型上，它顶替的正是上一代温度传感器的位置。官方名字叫 HiLight。 它会呼吸、会变色：聆听、思考、回答各有一种灯效——那是 Gemini 的三种状态。它会呼吸、会变色，聆听、思考、回答各 HiLight 灯效，对应 Gemini 的三种状态｜图片来源：Google 除此之外，它只剩一个用途：特定联系人来电时，亮一种你指定的颜色。但现在它不支持短信，不支持第三方通知，也没有给开发者留接口；隔壁 Nothing 靠背面灯效做出了半个品牌身份，而谷歌的这个灯，目前还只认 Ge…
DeepSeek V4 Pro 正式版 API 更新上线，多项测试性能接近 Fable 5 8 月 13 日消息，DeepSeek V4 Pro 正式版正式发布，已更新至 API，调用模型名不变。 新版本增强了 Agent 能力，支持 Responses API 和 Codex 接入。 从官方群放出的评测对比表可以看到，DeepSeek V4 Pro 正式版（DeepSeek-V4-Pro-0813）在多项测试中接近 Fable 5 水平，相比之前的预览版能力大幅提升。 定价如下： 百万 tokens 输入（缓存命中）：0.025 元 百万 tokens 输入（缓存未命中）：3 元 百万 tokens 输出：6 元 （来源：IT 之家） 腾讯发布 2026 年 Q2 财报，第二季度营收 2048 亿元，计划近期发布 Hy4 腾讯控股发布 2026 年第二季度财报。财报显示，腾讯第二季度营收 2,047.9 亿元人民币，预估 2,028.4 亿元人民币；第二季度销售费用 118.7 亿元人民币，预估 119.6 亿元人民币；第二季度增值服务业务收入 984.1 亿元人民币。 腾讯表示，混…
The Mission Robotic Vehicle is making the first attempt to attach a new thruster to an aging satellite.
The Paris-based vendor continues to build European AI infrastructure.
The cloud provider said the upgraded model is now better at completing tasks more quickly and using fewer tokens.
Alibaba released the open weights for Qwen3.8-2.4T-A95B (Qwen3.8-Max), its largest open-weight model, bringing near-frontier capabilities to the open... Alibaba released the open weights for Qwen3.8-2.4T-A95B (Qwen3.8-Max), its largest open-weight model, bringing near-frontier capabilities to the open ecosystem. It has 2.4T total parameters with 95B activated per token. It’s a fine-grained mixture of experts (MoE) architecture with a hybrid of full and linear attention, a context window of up to…
In this guide, we show how to build a defect detection and visual inspection system with computer vision using Roboflow.
AI infrastructure spans multiple layers, from compute and networking to storage, orchestration, and applications. When performance degrades, identifying the... AI infrastructure spans multiple layers, from compute and networking to storage, orchestration, and applications. When performance degrades, identifying the source can be difficult because a symptom observed at one layer may originate elsewhere in the stack. A full-stack observability strategy connects telemetry across these layers, helpi…
The ambitious plan comes as autonomous truck development lags behind that of autonomous cars.
The partnership with Together AI comes amid booming demand for open models.
The model is small enough for enterprises to run on local devices and is targeted toward specific tasks.
A survey of over 700 professionals examines how visual and physical AI teams build systems, why models fail, and where data work drives production. Download this free whitepaper now!
Introducing sign-language-to-text (SL2T), our breakthrough model powering new sign language features for Deaf and hard of hearing users.
NVIDIA founder and CEO Jensen Huang is ranked No. 1 on Glassdoor’s Best CEOs list for 2026. In the just-released ranking, recognition is earned directly from the people who know their leadership the best — employees. Huang topped the list, with 99% of employees approving of the job he does. “As AI and shifting expectations […]
Judges around the world have made headlines for illicitly using generative AI in their work. But in Pakistan, a large-scale trial of a specially designed AI tool for judges found the technology—together with appropriate training–boosted the number of cases resolved by 6.3 percent with no obvious drop in the quality of judgments. With a backlog of 2.26 million cases and fewer than two judges per 100,000 people—compared to 22 in the EU and eight in Brazil—Pakistan’s judiciary was in sore need of h…
