Discover the AI Tools

国内首部完全由AI生成的长剧《后西游记》于8月31日正式上线，芒果TV下午6点首播，湖南卫视黄金档随后跟进。这部剧改编自明清同名小说，首季共30集，每集40分钟。最大的亮点在于，它采用了“边审边播”的新模式，制作、审查、播出同步进行，大大缩短了创作周期，还能根据反馈灵活调整剧情。这会是影视行业的一次新探索吗？

ByteDance has postponed the launch of its Doubao large model 2.2, originally slated for August. The delay stems from a strategic decision to enhance the model's coding, tool calling, and agent intelligence through more thorough training. The company is doubling down on self-developed AI, with top executives rejecting reliance on AI distillation. This move signals a long-term commitment to foundational research, even if it means short-term setbacks. The team aims to make coding capabilities industry-leading by 2026.

Mihoyo's desktop companion app, 'BSide: Olivia Lin,' is shutting down on August 31, less than a month after its Steam early access launch. The developer has released an offline version for users to download before the server closes, retaining some features like local music playback and dynamic wallpapers. Users must back up their data before the deadline.

As AI tools become more prevalent, Danish high schools are seeing a rise in cheating during exams. In response, the government has introduced an emergency plan that includes oral defenses for thousands of students and monitoring computer screens during written tests. While some might see this as a crackdown, the approach is more about finding a balance—regulating AI use rather than banning it outright. Similar moves are happening in Australia and South Korea, signaling a global shift in how education adapts to generative AI.

Hugging Face's robotics arm, Pollen Robotics, has open-sourced Microduck, a 25cm-tall, 800g bipedal robot that costs under $400. It's not just a cute toy—it's a fully functional platform for learning reinforcement learning and embodied AI. With features like self-righting, object pickup, and even roller-skating, Microduck packs serious tech into a small, affordable package. The project includes open-source training code and a robust software architecture, making it a great entry point for developers and researchers.

A new Glassdoor report reveals a dramatic shift in how U.S. employees view AI. Positive mentions have plummeted from 81% in 2019 to just 43% by mid-2026, while negative comments now dominate. Frontline workers, especially software engineers and claims adjusters, are the most skeptical, citing job fears, clunky tools, and unrealistic expectations. The gap between executive optimism and worker reality is widening, raising questions about how companies can bridge the divide.

Sony and Warner have filed a lawsuit against AI company Anthropic, accusing it of using tens of thousands of copyrighted songs without permission to train its Claude models. The suit alleges widespread piracy, including downloading millions of pirated books and scraping lyrics illegally. If successful, the damages could reach billions, highlighting the growing legal risks for AI firms using unlicensed data.

8月31日，国内首部AIGC长剧《后西游记》正式上线，首季“花果山篇”共5集，每集40分钟，芒果TV与湖南卫视同步播出。该剧改编自明清同名小说，讲述孙悟空成佛1080年后，小石猴孙小圣等“西游二代”重走取经路的故事。值得一提的是，该剧是广电总局“21条”发布后首部采用“边审边播”模式的剧集，制作、审核、播出同步进行，开创行业先河。

Anthropic is adjusting usage limits for its AI coding assistant, Claude Code. Starting September 14, basic weekly quotas for paid plans will permanently rise by 25%, but because a temporary 50% boost ends that day, users will actually see a 17% cut from current levels. The company also rolled out performance upgrades and an automated feedback tool, signaling a shift from rapid expansion to precise resource management in AI coding services.

华为与上海交大医学院附属瑞金医院合作，正式发布了新一代病理AI模型RuiPath 2.0，并已在90多家医院投入使用。该模型在识别罕见癌症亚型方面表现突出，有望减少漏诊和误诊。团队还在探索多模态生成模型在淋巴瘤等复杂病例中的应用，为医生提供更精准的决策支持。

腾讯混元大模型Hy4preview上线即遭抢购，Agent能力大幅提升引发体验热潮。由于并发需求过高，WorkBuddy平台出现排队现象，腾讯已紧急扩容推理集群，并延长Hy3免费使用期限至9月底。这一现象折射出AI Agent在实际生产场景中的潜力正在加速释放。

中国信通院最新测试显示，上海原点星光的StartLux-V1.0-27B-Preview在MCP专项测试中综合排名第二，超越DeepSeek-V4-Flash。这款仅27B参数的模型，凭借技术创新，性能比肩万亿参数模型，并已能在消费级PC上运行。这标志着AI竞赛从参数规模转向实际应用，边缘AI时代正在到来。

Apple's leadership is changing hands. John Trenor, the hardware engineering chief, becomes CEO on September 1, with a clear mandate to accelerate the company's AI efforts. The new Siri is expected to launch soon, and a foldable iPhone might be on the horizon. Can Trenor leverage Apple's hardware strengths to catch up in the AI race?

微软最近开始严管员工使用AI的Token消耗，因为一份内部数据泄露显示，有员工在28天内花费了约2.8万美元的AI费用。为了控制成本，微软不仅加强了监控，还计划将内部默认工作负载切换到GPT-5.6 Sol，并鼓励员工使用自家AI工具，减少对第三方AI的依赖。

Apple's new CEO, John Ternus, officially takes office on September 1st, with AI as his top priority. The outgoing Tim Cook will become Executive Chairman. Ternus, a hardware veteran, faces his first big test at the September 9th launch event, where the iPhone 18 Pro and a foldable phone are expected. Will he deliver on AI?

iFLYTEK is making waves in the AI world with its upcoming Spark X2.5 series. On September 1st, two edge-side models (4B and 1.7B) will be open-sourced, boasting a 1M token context window and improved reasoning. Then, on September 7th, the massive 293B base model drops, promising even stronger code generation and agent collaboration. Plus, a new fully domestic-compute model is teased for the 1024 Developer Festival. Exciting times for AI enthusiasts!

Anthropic has introduced the Model Hardware Standard (MHS), a new protocol that lets AI directly control lab equipment and robots. Early tests show it can slash experiment times by two-thirds. This move signals a shift from chatbots to AI that can actually do things in the real world.

Anthropic has launched MHS, a new hardware standard that lets AI agents control physical devices. Developed with HHMI's Janelia Research Campus, MHS cuts hardware integration time from weeks to minutes. Early tests show Claude behaving like a curious scientist, even recovering from failures on its own. While still in preview, MHS marks a big step toward AI's real-world impact.

腾讯混元于8月28日正式发布开源旗舰模型Hy4preview，总参数达770B，激活参数49B，支持1M上下文长度，面向真实生产力场景。该模型已在HuggingFace、GitHub、ModelScope等平台开源，并同步上线腾讯云TokenHub和OpenRouter。用户可通过WorkBuddy、CodeBuddy、元宝、ima等产品直接体验。Hy4preview在软件工程、办公、游戏开发、科学研究等领域展现出显著能力提升，内部盲测表现优于GLM-5.3和Kimi K3，并在三维Blaschke–Lebesgue问题等科研任务中取得突破。

蚂蚁集团与中金公司合作，推出了首个金融增强大模型Ling-3.0-flash-Fin，该模型在信息检索等核心能力上表现突出，并计划开源。同时，双方还联合发布了金融搜索基准FinFIRST，以促进生态发展。