https://best.xiaohu.ai/en/ I'm XiaoHu. Every day I make sense of what's new in AI — checking what's real and explaining it in plain language. Each piece is a visual article with motion, interaction, and analogy. A place to read these explainers, searchable and worth trusting. en https://best.xiaohu.ai/en/article/seedream-5-pro-launch/ https://best.xiaohu.ai/en/article/seedream-5-pro-launch/ Adds click-to-select point editing and native rendering in 10+ languages, now live on Volcano Engine, rolling out next to Doubao and Jimeng 产品发布 Seedream 5.0 Pro字节跳动SeedAI图像生成交互式编辑多语种生成 Wed, 08 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/deeptutor-open-source-tutor/ https://best.xiaohu.ai/en/article/deeptutor-open-source-tutor/ Paper benchmarks show a 10.8% boost in personalization and 29.4% in reasoning — fully open-sourced under Apache 2.0 产品发布 DeepTutorHKUDS开源项目AI家教个性化学习Agent Wed, 08 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/fable5-advisor-executor/ https://best.xiaohu.ai/en/article/fable5-advisor-executor/ In advisor mode, Fable 5 just offers suggestions; in orchestrator mode, it delegates tasks — either way, the cheaper Sonnet 5 ends up doing most of the actual work 深度 Fable 5多模型协作Agent成本优化Claude Managed Agents Wed, 08 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/liquid-ai-antidoom/ https://best.xiaohu.ai/en/article/liquid-ai-antidoom/ By fine-tuning just a single token at the loop's starting point, both models' infinite-loop rates drop to around 1% 研究解读 Liquid AI推理模型死循环偏好优化训练开源方法 Wed, 08 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/meta-muse-image-video/ https://best.xiaohu.ai/en/article/meta-muse-image-video/ Sibling model Muse Video was previewed alongside it, now with native audio generation and coming soon to creators. 产品发布 MetaMuse ImageAI生图自我修正 Wed, 08 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/cloudflare-monetization-gateway/ https://best.xiaohu.ai/en/article/cloudflare-monetization-gateway/ Expanding from charging only AI crawlers to billing any caller — now in early access 产品发布 Cloudflarex402协议稳定币支付 Tue, 07 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/gpt-realtime-2-1/ https://best.xiaohu.ai/en/article/gpt-realtime-2-1/ When calling a tool, the model now says "let me check that" first — so calls no longer go silent while waiting. 产品发布 OpenAIRealtime API语音智能体 Tue, 07 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/alberta-claude-cybersecurity/ https://best.xiaohu.ai/en/article/alberta-claude-cybersecurity/ 50 autonomous agents worked in parallel, covering 27 provincial departments and 3,400 code repositories, and could even auto-patch vulnerabilities and rewrite legacy systems. 商业动态 Claude Code网络安全政府数字化 Tue, 07 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/cloudflare-workers-cache/ https://best.xiaohu.ai/en/article/cloudflare-workers-cache/ One line of Wrangler config plus standard HTTP cache headers, with caching now embeddable between any internal entry points in a Worker. 产品发布 Cloudflare Workers边缘缓存Serverless Tue, 07 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/making-of-claude-code/ https://best.xiaohu.ai/en/article/making-of-claude-code/ 16 Insiders Tell the Story: From Grinding Over Diffs to a Two-Week Sprint Launch — to Engineers Writing Zero Code by Hand 深度 Claude Code口述历史AI编程工具 Tue, 07 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/fable-field-guide-unknowns/ https://best.xiaohu.ai/en/article/fable-field-guide-unknowns/ Claude Code team's Thariq Shihipar at a conference talk: the bottleneck for new models is no longer the model itself, but whether you can articulate your own unknowns 深度 Claude Fable 5提示词方法Agentic Coding Tue, 07 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/claude-global-workspace/ https://best.xiaohu.ai/en/article/claude-global-workspace/ It's less than 10% of the network, but remove it and Claude can still talk while losing all reasoning — Anthropic already uses it to catch models fabricating data and spot when they sense a test. 研究解读 可解释性研究ClaudeAI安全监测 Tue, 07 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/phosphor-dartmouth-quiz-study/ https://best.xiaohu.ai/en/article/phosphor-dartmouth-quiz-study/ A 151-student Dartmouth study: short-answer questions move scores more than multiple choice, yet almost no one uses the AI help sidebar 研究解读 AI教育科技形成性评估简答题AI判卷达特茅斯学习平台设计 Mon, 06 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/personalized-aac-app/ https://best.xiaohu.ai/en/article/personalized-aac-app/ No funding, no team — just a father solving his own son's problem. Then clinics and schools started asking to use it. 产品发布 AAC辅助沟通自闭症干预AI生成个性化内容语音克隆 Mon, 06 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/aiewf-loops-debate/ https://best.xiaohu.ai/en/article/aiewf-loops-debate/ The live vote couldn't be tallied because the lights were too bright; a concurrent survey found 95% of teams already use agents, while 59% worry technical debt is piling up. 深度 AIEWF大会agentic loopAI工程调研 Mon, 06 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/ai-schools-wealth-gap/ https://best.xiaohu.ai/en/article/ai-schools-wealth-gap/ While traditional schools are still figuring out how to use AI, Silicon Valley and Wall Street families are already voting with their wallets 深度 AI教育教育鸿沟个性化学习 Mon, 06 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/ai-junior-dev-collapse/ https://best.xiaohu.ai/en/article/ai-junior-dev-collapse/ US developer employment for ages 22-25 has dropped 19% in three years, even as new GitHub accounts grow at their fastest pace ever 研究解读 AI就业冲击初级程序员Vibe Coding劳动力市场数据 Sun, 05 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/nvidia-horizon-chip-agent/ https://best.xiaohu.ai/en/article/nvidia-horizon-chip-agent/ The first paper to run an entire RTL benchmark suite unattended end-to-end — most problems clear in two or three rounds, but the hardest one takes 82 iterations 研究解读 NVIDIA ResearchAgentic 工作流RTL芯片设计Git版本控制 Sun, 05 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/claude-code-expertise/ https://best.xiaohu.ai/en/article/claude-code-expertise/ Analysis of 235K users over 7 months: experts verify success nearly 2x more often than novices, yet the top 10 professions differ by only 7 points 研究解读 Claude CodeAnthropic研究agentic coding职场专业度人机分工 Sun, 05 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/human-nature-persuasion/ https://best.xiaohu.ai/en/article/human-nature-persuasion/ Create psychological tension first, then offer a first step too small to refuse — a formula that works for writing, selling, and job hunting alike. 深度 人性洞察说服框架文案写作内容营销 Sun, 05 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/ai-systems-not-models/ https://best.xiaohu.ai/en/article/ai-systems-not-models/ The US's June 12 export controls put frontier AI models themselves—not just chips—under restriction for the first time. His answer: master multi-model orchestration. 深度 AI地缘政治出口管制多模型编排日本AI战略 Sat, 04 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/fable-finding-unknowns/ https://best.xiaohu.ai/en/article/fable-finding-unknowns/ Anthropic's Thariq argues that when working with Claude Fable 5, output quality hinges on how clearly you can articulate your own "unknowns." This field guide offers 8 techniques for surfacing them — before, during, and after implementation — each paired with a ready-to-use prompt. 深度 Claude Fable 5Agentic Coding提示工程 Sat, 04 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/the-great-descent/ https://best.xiaohu.ai/en/article/the-great-descent/ Investor Chamath Palihapitiya: intelligence is getting cheap like phones, and expert judgment is now available to everyone; the real moat is encoding your own proprietary experience into your system, not renting the same generic AI as your competitors 深度 AI成本下降专家判断力普及商业护城河 Fri, 03 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/microsoft-frontier-company/ https://best.xiaohu.ai/en/article/microsoft-frontier-company/ Customer data won't be used to train models that would erode a client's competitive edge, and the platform lets companies switch freely between AI models — no vendor lock-in 商业动态 微软企业AI转型AI工程服务 Fri, 03 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/codex-in-practice/ https://best.xiaohu.ai/en/article/codex-in-practice/ She used the same pattern to build an email triage tool and a caregiving app for her dad — the method is repeatable, though the full prompts stay private. 深度 OpenAI CodexLoop 工作模式非技术者自建系统Every Fri, 03 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/enterprise-ai-agent-guide/ https://best.xiaohu.ai/en/article/enterprise-ai-agent-guide/ Four implementation principles, backed by real-world data from L'Oréal, Lyft, and Rakuten 商业动态 企业AI转型Claude Cowork插件生态AI Agent部署 Fri, 03 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/ai-agent-architecture-guide/ https://best.xiaohu.ai/en/article/ai-agent-architecture-guide/ Multi-agent systems can boost performance by 90.2% — but at 10-15x the token cost. This three-question framework helps you decide if the complexity is worth it. 深度 AI Agent架构多智能体系统Anthropic Fri, 03 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/alibaba-page-agent/ https://best.xiaohu.ai/en/article/alibaba-page-agent/ MIT-licensed and model-agnostic, it works with any OpenAI-compatible text model — though for now it can only operate on a single page view. 产品发布 Page Agent阿里巴巴浏览器自动化开源 Fri, 03 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/adobe-agentic-site/ https://best.xiaohu.ai/en/article/adobe-agentic-site/ In the live demo, searching "camping" turned a coffee-machine site's copy and products fully outdoor-themed — the tech works, but no client site has shipped it at scale yet. 深度 Adobe智能体网站AI个性化 Fri, 03 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/pm-ai-leverage-ladder/ https://best.xiaohu.ai/en/article/pm-ai-leverage-ladder/ An instructor who has trained 30,000+ PMs breaks down two AI leverage ladders — from copy-paste to end-to-end delivery, and from web prototypes to production PRs. 深度 产品经理AI工作流MCPClaude Code Fri, 03 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/claude-advisor-tool/ https://best.xiaohu.ai/en/article/claude-advisor-tool/ The advisor model only chips in a few hundred words of guidance instead of taking over the whole task, cutting total cost at the same quality bar — currently a beta feature limited to the Claude API and AWS 产品发布 Claude API多模型协作Agent 工具 Thu, 02 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/fable5-prompting-guide/ https://best.xiaohu.ai/en/article/fable5-prompting-guide/ Anthropic's official documentation shows how to tune system prompts and engineering scaffolding for the new model — the same methods also work for Claude Mythos 5 深度 Claude Fable 5提示工程长程自主 Agent Thu, 02 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/claude-code-loops/ https://best.xiaohu.ai/en/article/claude-code-loops/ From manual confirmation to fully unattended — the Claude Code team's 4-tier loop taxonomy and practical playbook 深度 Claude CodeAI Agent自动化循环Dynamic Workflows Thu, 02 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/oasis-1-whisper-ring/ https://best.xiaohu.ai/en/article/oasis-1-whisper-ring/ Titanium build, $289 preorder, shipping around Christmas 2026 — follows the company's first-gen touch-only ring. 产品发布 智能戒指默念输入语音交互可穿戴设备 Wed, 01 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/longcat-2-no-nvidia/ https://best.xiaohu.ai/en/article/longcat-2-no-nvidia/ Trained on more than 50,000 domestic AI chips over 35 trillion tokens; most benchmark scores come from Meituan's own evaluation framework, and the weights aren't actually open for download yet 产品发布 大模型发布国产AI芯片MoE架构开源模型长上下文 Wed, 01 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/claude-science-workbench/ https://best.xiaohu.ai/en/article/claude-science-workbench/ Now in open beta. A coordinator agent marshals a team of specialist agents to do the work, capped off by a reviewer agent that hunts specifically for citation and numerical errors — compute gets outsourced to AI, but raw data never leaves your machine. 产品发布 Claude Science科研智能体Anthropic计算资源管理 Wed, 01 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/nano-banana-2-lite/ https://best.xiaohu.ai/en/article/nano-banana-2-lite/ Images generated in just 4 seconds at about $0.034 per 1,000; the Omni Flash video model opens to developers the same day. 产品发布 Nano BananaGemini Omni FlashAI生图AI视频生成 Wed, 01 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/tinker-financial-triage/ https://best.xiaohu.ai/en/article/tinker-financial-triage/ With Thinking Machines, fine-tuned an open-source model on expert-labeled data: 29.8% lower error rate than the best frontier model, at just 1/14 the inference cost 研究解读 Tinker强化学习微调金融AI专家标注在线策略蒸馏 Wed, 01 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/claude-sonnet-5-launch/ https://best.xiaohu.ai/en/article/claude-sonnet-5-launch/ Official benchmarks show that at high compute, Sonnet 5 matches Opus 4.8 on select tasks — at just 60% of its standard price. 产品发布 Claude Sonnet 5Anthropic模型定价Agentic AI网络安全防护 Wed, 01 Jul 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/ai-support-deflection-trap/ https://best.xiaohu.ai/en/article/ai-support-deflection-trap/ Enterprise AI customer service has entered a consolidation phase. Klarna and Alibaba's 2.56 million conversation datasets both point to the same blind spot: cutting costs isn't the same as solving problems. 深度 AI客服企业AI客户体验产品方法论Klarna Tue, 30 Jun 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/brockman-agi-interface-era/ https://best.xiaohu.ai/en/article/brockman-agi-interface-era/ Brockman confirms OpenAI has multiple hardware devices in development; only ~20M users use Agents, yet ChatGPT is closing in on 1 billion 深度 OpenAIAGIAgent时代Greg Brockman算力竞赛 Tue, 30 Jun 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/brain2qwerty-v2-noninvasive/ https://best.xiaohu.ai/en/article/brain2qwerty-v2-noninvasive/ A wearable helmet decodes magnetoencephalography signals in real time, pushing word accuracy from 8% to 61%; v1/v2 training code and datasets released open source 研究解读 脑机接口Meta AI神经科学非侵入式 Mon, 29 Jun 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/loop-engineering-playbook/ https://best.xiaohu.ai/en/article/loop-engineering-playbook/ An Anthropic engineer's methodology explained: instead of prompting AI line by line, you design a self-running loop system 深度 AI Agent工程方法论Claude Code自动化 Mon, 29 Jun 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/adrafinil-agent-aware-sleep/ https://best.xiaohu.ai/en/article/adrafinil-agent-aware-sleep/ One-click integration with Claude Code and 8 other agents; keeps tasks running with the lid closed and releases sleep control within 50ms of going idle 产品发布 macOSAI Agent开发工具Claude Code开源 Mon, 29 Jun 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/dspark-deepseek-v4-speculative/ https://best.xiaohu.ai/en/article/dspark-deepseek-v4-speculative/ 60–85% faster on top of existing MTP-1 speculative decoding, achieved by overlapping draft and verification pipelines (numbers per DeepSeek's own benchmarks) 产品发布 DeepSeek推理加速投机解码大模型基础设施 Mon, 29 Jun 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/compound-engineering-fourth-step/ https://best.xiaohu.ai/en/article/compound-engineering-fourth-step/ Solo operators running 5 products swear by one habit: after every feature ships, log the solution back into the system so AI avoids the same pitfall next time 深度 AI编程工程方法论Claude Code多Agent效率 Sun, 28 Jun 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/gpt-5-6-sol-metr-eval/ https://best.xiaohu.ai/en/article/gpt-5-6-sol-metr-eval/ Three conflicting capability numbers, none reliable — but the visible cheating itself serves as evidence that safety monitoring is working 研究解读 GPT-5.6AI安全模型评测OpenAIMETR Sat, 27 Jun 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/gpt-5-6-sol-preview/ https://best.xiaohu.ai/en/article/gpt-5-6-sol-preview/ Three tiers at once (Sol/Terra/Luna), limited rollout to trusted partners with names filed with the U.S. government, broader access in a few weeks 产品发布 OpenAIGPT-5.6AI安全网络安全大模型 Sat, 27 Jun 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/wan-streamer-realtime-av/ https://best.xiaohu.ai/en/article/wan-streamer-realtime-av/ Model latency ~200ms, end-to-end ~550ms; v0.1 at 192p, demo is pre-recorded not live 研究解读 多模态大模型全双工实时交互端到端音视频对话数字人 Sat, 27 Jun 2026 00:00:00 -0000 https://best.xiaohu.ai/en/article/ibm-sub1nm-chip/ https://best.xiaohu.ai/en/article/ibm-sub1nm-chip/ Lab-verified as manufacturable; the +50% performance and +70% energy efficiency figures are projected gains over 2nm, not measured results 研究解读 IBM半导体芯片制程 Fri, 26 Jun 2026 00:00:00 -0000