AI Agent News
实时追踪 AI Agent 赛道的重大事件、融资动向、模型发布和技术突破
最新行业资讯
实时追踪 AI Agent 赛道的重大事件、融资动向、模型发布和技术突破
重大事件时间线
OpenClaw GitHub 爆发
OpenClaw 10 天冲上 GitHub 全球 Top 10，超越 Linux 内核 Star 增速
Meta 20亿收购 Manus
Meta 以 20 亿美元收购 Manus AI，通用 Agent 赛道正式被巨头锁定
DeepSeek-V3 开源
性价比之王，成本仅 GPT-4 的 5%
Manus 一夜爆火
全球首款通用 AI Agent 在国内社交平台引发空前关注
OpenAI Deep Research
OpenAI 推出深度研究 Agent，一键生成专业研究报告
MCP Server 破 500
MCP 生态爆发，3 个月构建 500+ Server
DeepSeek-R1 震惊全球
开源推理模型，成本仅 OpenAI 的 3%，引发全球 AI 格局震动
MCP 协议诞生
Anthropic 发布 Model Context Protocol，成为 Agent 接口事实标准
Claude Computer Use
Anthropic 让 AI 首次直接操控电脑屏幕，开创计算机使用新范式
Replit Agent 全栈自动化
自然语言到上线产品，面向非工程师
Cursor ARR 破亿
史上增长最快 SaaS，AI 编程工具新王者
Claude 3.5 登顶 SWE-bench
最强编程 AI，Bug 修复能力达到初级工程师水平
Devin 发布
全球首个自主 AI 软件工程师，能独立完成完整编程任务
Perplexity AI Launches Enterprise Plan with Private Deployment
Perplexity AI launched Perplexity Enterprise with private model deployment, company knowledge base integration (Confluence, Notion, SharePoint), SSO, and admin controls—addressing the data privacy concerns that had held back corporate adoption. The enterprise tier targets research-intensive workflows: competitive intelligence, market research, and meeting preparation. Priced at $40/user/month, the launch is central to Perplexity's path to $50M ARR projected by end of 2025. The company recently raised additional funding at a valuation exceeding $1B.
Cursor Raises $100M Series B at $1.25B Valuation as AI Code Editor Wars Heat Up
Cursor, the AI-first code editor, raised a $100M Series B led by Andreessen Horowitz at $1.25B valuation. The company reports 1M monthly active users (up from 200K in 6 months) and $30M ARR, driven almost entirely by word-of-mouth from developers experiencing productivity gains. Unlike AI code completion add-ons, Cursor is built from the ground up with AI as first-class collaborator: full codebase context, multi-file editing, and natural language-to-code with understanding of existing patterns. Enterprise features including team collaboration and compliance controls launching later in 2025.
Microsoft Phi-4 Mini: GPT-4 Level Performance in 3.8B Parameter Model
Microsoft Research released Phi-4 Mini (3.8B parameters), demonstrating that careful training data curation enables small models to achieve quality comparable to GPT-4 on reasoning-heavy tasks. Phi-4 Mini achieves 72.3% on MMLU and 75.4% on GPQA, exceeding much larger models. The model runs on consumer hardware: 4GB RAM for CPU inference, runs at 30+ tokens/second on Apple M-series chips. Key applications: edge AI (phones, laptops, IoT without cloud), cost-efficient inference at scale, specialized fine-tuning for domain-specific tasks. Available on Hugging Face under MIT license.
GitHub Copilot Workspace Goes GA: AI-Powered Software Development Environment
GitHub announced general availability of Copilot Workspace, a cloud-based AI development environment where developers describe what they want to build and AI handles the implementation planning, code generation, and testing. The system creates a complete development plan, generates all necessary code changes across multiple files, runs tests, and iterates until tests pass—with human oversight throughout. Early access customers report completing multi-file features in 30 minutes that previously took half a day. Available to GitHub Enterprise customers with Copilot Enterprise license.
Harvey AI and Ironclad Lead Legaltech AI Wave Reshaping Law Firm Economics
The legal industry's AI adoption is accelerating dramatically. Harvey AI (backed by OpenAI and GV) has signed agreements with 8 of the top 20 law firms and is processing 500,000+ legal documents daily. Ironclad's AI contract management system is managing 2M+ contracts for Fortune 500 legal departments. The economic impact: associates at AI-using firms report completing document review 4-5x faster, with senior partners redirecting to higher-value work. A recent survey found 67% of Big Law partners expect AI to reduce associate headcount over 5 years, while 78% expect to expand the business's overall capacity using current headcount.
Linear AI Transforms Engineering Project Management with Autonomous Issue Handling
Linear, the project management tool popular with engineering teams, released Linear AI—a suite of AI features that handle routine project management autonomously. New capabilities: automatic issue creation from Slack conversations and customer feedback, smart priority scoring based on team velocity and business impact, sprint planning recommendations, automated bug report triage, and writing issue descriptions from rough notes. Linear reports teams using AI features complete 20% more issues per sprint with same team size. Available at no additional cost to existing Linear plans.
AI Startup Valuations Reset: Investors Demand Real Metrics in 2025
The AI startup funding environment has rationalized in 2025. Average pre-revenue AI valuations declined 40% from 2024 peaks as investors demand demonstrated revenue and genuine differentiation. "GPT wrapper" startups struggle to raise while AI companies with data moats, deep workflow integration, and regulated-industry specialization command 15-25x ARR multiples. Analysis reveals the bifurcation: real AI businesses flourish; hype-driven startups face down rounds. Pre-revenue caps dropped from $20-50M in 2024 to $8-15M in 2025.
Vapi Enables AI Phone Agents with Human-Level Conversation Quality
Vapi, the voice AI platform, has emerged as the leading infrastructure for AI phone agents, powering 50M+ calls per month. Its conversational AI engine achieves sub-300ms response latency—fast enough to eliminate awkward pauses in phone conversations. Customers include healthcare providers (appointment scheduling, patient follow-up), real estate firms (property inquiry handling), and e-commerce companies (order support). The platform handles interruptions naturally, remembers context across the conversation, and escalates to humans smoothly. At $0.05/minute, AI phone agents cost 70% less than human agents for routine calls.
AI Search Is Disrupting Google: 15% of Search Queries Now Go to AI-First Tools
New research from Similarweb and SparkToro shows that 15% of informational search queries that would previously have gone to Google are now handled by AI tools: ChatGPT (8%), Perplexity (4%), Claude (2%), Gemini (1%). The shift is most pronounced for research questions, how-to queries, and complex multi-part questions where AI provides synthesized answers without requiring link-clicking. Google's response—Gemini integration across search—has mixed user reception. SEO professionals are adapting to "Answer Engine Optimization" (AEO) strategies. Publishers report 20-30% decline in traffic from informational queries.