Discover the AI Tools

A recent survey reveals that 90% of executives admit AI hasn't boosted productivity, despite widespread layoffs and heavy investment. The real efficiency gains came from remote work, not AI. Employees are skeptical, and companies are facing a 'boomerang effect' as they rehire laid-off workers. The cost of AI-driven layoffs may be far higher than anticipated.

Google has officially unveiled the Pixel 11 series, focusing on deep Gemini AI integration rather than hardware changes. The new phones offer agent-like assistance, improved camera features, and a faster Tensor G6 chip. With prices starting at $899, the Pixel 11 aims to make daily tasks easier through smart automation.

European AI coding startup Lovable just raised $40 million in Series C funding, pushing its valuation to a staggering $13.3 billion. With annual recurring revenue already past $500 million and a platform buzzing with 60 million projects, Lovable is doubling down on growth and even investing in local AI talent. What's next for this rising star?

8月13日，支付宝鸿蒙版迎来重大更新，AI原生助手“阿宝”正式面向鸿蒙用户开放公测。用户只需在华为应用市场更新至最新版，即可通过右滑或点击左下角按钮体验。这标志着“阿宝”已覆盖iOS、Android和鸿蒙三大移动操作系统。自6月推出以来，“阿宝”以自然语言交互颠覆了传统菜单点击模式，用户一句话即可调用餐饮、出行、充值缴费、家政、政务等数千种服务。随着鸿蒙设备数突破7000万，支付宝正推动移动互联网服务从“人找服务”向“AI按需调度服务”转变。

Google DeepMind has unveiled SL2T, a multilingual sign language-to-text model now integrated into Pixel 11 smartphones. This marks the first time sign language AI has reached mainstream consumer devices, allowing users to communicate via gestures through the front camera. Trained on over 100,000 hours of data from 50+ sign languages, SL2T sets new benchmarks and prioritizes privacy by processing on-device. The feature initially supports American Sign Language to English, with plans for expansion.

韩国人工智能实验室Upstage近日发布了专为复杂代理任务设计的商用大模型Solar Pro 4，支持512K上下文窗口和128K输出长度，定价极具竞争力。该模型在多个国际评测平台上亮相，成为首个登上Agent Arena等榜单的韩国模型，其性能表现令人期待。

CodeBuddy今日宣布，其旗下WorkBuddy及全系列IDE、插件和CLI产品已全面升级至DeepSeek-V4-Pro正式版，为开发者带来最新模型能力。同时，官方预告DeepSeek V4系列即将调价，建议用户提前规划。此次升级覆盖多终端，将影响AI编程与智能开发场景的应用成本。

约克大学和卡尔加里大学的研究团队分析了Reddit上6000多条关于AI编程助手的评论，发现Cursor等工具频繁因权限过大引发事故，如覆盖文件、删除数据甚至生成恶意代码。研究指出，AI助手权限越大，潜在破坏力越强，开发者需警惕过度授权，加强人工审查。

Anthropic is reportedly planning an IPO in October, with investors eyeing a valuation that could exceed $2 trillion—surpassing SpaceX's record. The company's revenue is soaring, projected to hit $100-$120 billion by 2026, driven by strong demand for its Claude models. However, challenges like competition and regulation loom.

Before stepping down as CEO of Google DeepMind, Demis Hassabis quietly discussed plans for an independent AI safety organization with U.S. officials, including Treasury Secretary Scott Bessent. His vision? A body akin to the International Atomic Energy Agency, but for artificial intelligence. With Google's recent leadership shuffle, Hassabis now focuses on shaping AGI's future—and possibly, the rules that govern it.

Blacksmith, a startup specializing in AI code testing and verification, has seen its valuation jump from $60 million to $550 million in under a year. The company just closed a $4.5 million Series B round, led by PeakXV Partners, with participation from GV and Y Combinator. As AI coding tools flood the market, Blacksmith is tackling the new bottleneck: ensuring all that generated code actually works. With over 5,000 customers and a lean team of 30, they're proving that verification is the next big thing in software development.

WorkBuddy's latest update introduces an AI-native knowledge base that supports collaborative editing of HTML and Markdown, plus a light app builder. This shift transforms AI-generated content from disposable outputs into reusable, evolving assets, addressing common pain points in AI office work.

xAI's Grok 4.6 has hit the scene, matching GPT-5.6's intelligence score while undercutting its price by over 60%. With major gains in agent tasks and a focus on cost-efficiency, this model is shaking up the AI landscape. Discover how it stacks up against the competition and what it means for developers and businesses.

Google's next-gen Gemini model is reportedly delayed by two months due to internal disagreements over priorities and resource allocation, compounded by tight compute supply. Despite owning TPU chips, the company faces a crunch from multiple demands. Co-founder Sergey Brin has stepped into core training, while leadership changes aim to streamline focus. This internal shake-up reflects the high-stakes race to lead in AI.

Alibaba's Qwen team has open-sourced its flagship Max-level model, Qwen3.8-2.4T-A95B, with 2.4 trillion total parameters and 95 billion activated per token. It supports a 260k-token context, expandable to 1M. The model rivals top international systems in benchmarks like PaperBench and OSWorld-Verified, and can autonomously handle multi-day programming tasks. This move marks a significant step in democratizing advanced AI.

Anthropic quietly added invisible watermarks to Claude's output to comply with EU regulations, but Reddit users are divided. Critics call it a 'digital tattoo' that unfairly targets ordinary users, while supporters argue it's necessary for transparency. The debate highlights tensions between AI transparency and user freedom.

Apple is in talks with publishers to license news content for its AI-powered Siri, but with a twist: it wants to pay only when the content is actually used. This usage-based model could reach hundreds of millions of dollars and marks a shift from traditional licensing deals. As Siri gets a major upgrade, Apple aims to fix its real-time information gap while avoiding the pitfalls of AI-generated news summaries that previously sparked controversy.

SpaceX AI has officially launched Grok 4.6, its latest flagship model, designed to tackle complex, long-term tasks. In benchmark tests, it now stands shoulder to shoulder with OpenAI's GPT-5.6 Sol, trailing only Claude Opus 5 and Fable 5 Max. With competitive pricing and high speed, it's already available on Cursor and Grok Build, with API access for developers.

On Wednesday, a malicious edit to Sam Altman's Wikipedia page led Google to briefly display a false death notice. The incident, which involved multiple edits over several hours, highlights the fragility of search results in the AI era. Google has since corrected the error, but questions remain about the reliability of public information sources.

西湖机器人作为西湖大学在AI和机器人领域的首个重大成果转化项目，半年内完成四轮融资，总额达5亿元。最新A轮融资由海源资本领投，募资近亿元，将用于人形机器人统一大模型研发和人才培养基地建设。公司创始人王东林教授带领团队，坚持全栈自研，已推出自研GAE“阿凡达”系统，并与阿里、英特尔等企业合作，订单近亿元。未来将拓展至高危工业巡检等场景。