Discover the AI Tools

Neuralink just pulled off its most ambitious upgrade yet: connecting Elon Musk's AI model Grok directly to a human brain. In a 72-hour coding marathon at a patient's home, engineers built eight new features, including a system that lets you 'speak' with your thoughts—no buttons, no clicks. For ALS patient Kenneth, it's a game-changer. But what does it mean for the rest of us?

Anthropic just gave Claude a major mobile makeover. The standout? iPhone users can now chat with Claude through CarPlay while driving. Beyond the car, the update brings faster voice, 24 new languages, multi-account switching, and smarter layouts for big screens. It's a clear push to make Claude a constant companion—at home, at work, and on the road.

Mistral just dropped Mistral Large4, a trillion-parameter open-source model that doesn't just chase benchmarks—it saves tokens. With 47% fewer tokens per coding task and top-tier scores, it's a cost-effective powerhouse. But can it really balance performance and price? We break down the specs, the safety nets, and the license that keeps MaaS providers on a leash.

英伟达开源了Nemotron 3.5 Transcribe语音转写模型，采用MoE架构，总参数340亿，单次可处理1小时音频。它支持40种语言转写和31种语言与英语互译，还能区分最多8个说话人。在Open ASR和Artificial Analysis测试中，词错误率分别低至5.33%和2.8%，表现亮眼。

字节跳动旗下AI助手豆包又出大招了！10月30日起，新老用户只要每天签到，就能轻松攒积分，最高能白嫖8800积分，新用户还能免费试用30天付费套餐。这波羊毛不薅白不薅，赶紧打开桌面端看看吧！

Google Research just dropped a paper called 'Language Models Need Sleep'—and no, it's not about giving your chatbot a nap. The idea: let AI models consolidate new knowledge during a sleep phase, so they don't wipe out old skills. Early tests show near-zero forgetting on continual learning tasks. Could this finally kill the dreaded 'knowledge cutoff'?

Baidu just made AI search cheaper and smarter. The company announced a major upgrade to its AI search and Search AI API, slashing the API price to just 4.9 yuan per thousand calls. Meanwhile, the Baidu App gets four new AI features, including Wenxin Assistant that now delivers tasks directly and AI Lens that understands nine dialects. It's a clear push to make AI search more accessible and useful for everyone.

Spotify is betting big on AI music. By early 2027, paid subscribers might be able to remix or cover licensed songs for an extra $4–$5 a month. Deals are inked with Universal, Merlin, and Believe, but Sony and Warner are still haggling. Artists get a say—and a cut. Will fans pay to play producer?

Google just made Gemini a whole lot more useful. Starting October 30, you can edit PDFs, spreadsheets, and code files right inside the chat window—no more downloading and reopening in another app. The update covers a wide range of formats and aims to close the loop between generating and editing documents, giving tools like Claude a run for their money.

Meta quietly launched Style Sketch, a free app that scans your Facebook and Instagram photos to build a virtual wardrobe. It then dresses a model with your body type and skin tone, suggests outfits based on weather and travel plans, and lets you share looks with friends. The app is currently testing in Southeast Asia, but privacy concerns linger over which photos are accessed and whether they can be truly deleted.

Anthropic's upcoming Fable 5.5 model isn't just smarter—it comes with a secret safety mechanism called Self-Assessment. After each response, Claude quietly reviews the conversation for risks, catching 78% of violations with minimal compute. But does this invisible audit cross a privacy line? Here's what we know.

Daxiao Robotics just released ACE-Brain-0.6, an API that helps machines understand where things are in 3D space. Trained on 1.5 million spatial anchoring examples—up from 1.1 million—the model now scores 75.41 across seven public benchmarks, edging out OpenAI's GPT-6 Astra. For developers building embodied AI, that means robots can finally see clearly and move to the right spots.

GitHub just made choosing an AI model one less thing to worry about. At Universe 2026, it launched Copilot Fusion—a smart router that automatically picks and combines models from OpenAI, Anthropic, Google, and others. The result? Nearly top-tier coding performance at about half the cost, with tasks finishing 40% faster. It's available now across all Copilot plans.

Anthropic quietly rolled out an environment variable in Claude Code v2.1.311 that gives AI agents permission to delete files without approval. The name? CLAUDE_CODE_EVEN_MORE_DANGEROUSLY_SKIP_PERMISSIONS. An engineer confirmed it's not a joke. Developers must opt in, but it raises fresh questions about how much autonomy we really want to hand over.

Wu Tian, Baidu VP and deputy director of Baidu Research, has left the company. But Baidu isn't sitting still: just two days before the news broke, it open-sourced Wenxin Unlimited OCR2, which can parse documents of any length—like all seven Harry Potter books in one go. New Wenxin model updates are expected next week.

10月29日，芯擎科技甩出王炸——X7 Ultra智驾芯片，700 TOPS算力，99.99%的实测对齐度。这家公司从底层需求倒推，重新设计了计算架构、内存方案和数据流，还留足了未来模型升级的空间。大模型定义芯片，这次玩真的。

谁能想到，造火箭的 SpaceX 如今靠卖算力赚得盆满钵满？最新文件显示，AI 新贵 Anthropic 贡献了其超三分之一收入，长期合同潜在价值高达 1205 亿美元。从今年 12 月起，Anthropic 每月要付约 31 亿美元，但留了提前解约的后门。这场合作让 SpaceX 意外成为 AI 淘金热里的隐形赢家。

Meta quietly launched Style Sketch, an AI fashion app that builds a virtual wardrobe from your Instagram photos and suggests outfits using a virtual model that matches your body type and skin tone. Currently available in Indonesia, Singapore, and Thailand, it's part of Meta's broader push into standalone AI apps.

你听说过AI工作手机吗？深兰科技和华为刚刚联手推出了全球首款大模型AI智能体工作手机Wowoker。这款手机不只是通讯工具，它能自动获客、智能回复、甚至用AI克隆你的声音打推销电话。目标客户是政府、金融等企业用户，主打数据安全和效率提升。

企业采购大模型，以往得跟不同厂商分别谈，费时费力。现在，字节跳动的豆包工作直接集成了DeepSeek、Kimi、GLM三款旗舰模型，企业用户可以在一个工作台里自由切换，按需选用。更省心的是，费用自动从现有额度扣除，无需额外审批。管理员还能精细控制访问权限，确保数据安全。这波操作，让企业AI工具采购变得前所未有的简单。