Core Insights - AI is creating a new paradigm for information retrieval, moving beyond traditional search engines like Google to provide high-quality answers directly to users through applications like Perplexity and ChatGPT [1] - Companies like Exa and You.com are leading this shift by developing search engines specifically designed for AI, aiming to enhance the speed and accuracy of information retrieval [4][5] Company Summaries Exa - Exa has raised $85 million in Series B funding, led by Benchmark, with a valuation of $700 million [3] - The company utilizes a proprietary neural network search architecture, focusing on predicting the next link rather than traditional keyword matching, and offers a search API designed for AI [7][11] - Exa's Research API achieved a high score of 94.9% in SimpleQA benchmarks, and the company emphasizes a pay-per-use model without ads, prioritizing quality [11] You.com - You.com has secured $100 million in Series C funding, led by Cox Enterprises, with a valuation of $1.5 billion [3] - Initially launched as a traditional search engine, You.com has evolved to focus on AI-driven search capabilities, offering features like YouCode and YouWrite for generating content based on queries [14][15] - The company's core technology is a model-agnostic AI operating system that enhances the accuracy and reliability of any large language model [15][18] Industry Trends - The search query volume for AI is expected to surpass human queries in the coming years, indicating a significant shift in how information is accessed [3] - The emergence of AI-native search engines presents disruptive entrepreneurial opportunities in the search domain, with potential for both human-targeted and AI-targeted applications [20][21] - Companies are encouraged to either focus on high-performance tools like Exa or develop integrated solutions like You.com that bind closely with customer workflows [21]

Core Insights - Exa, an AI search company co-founded by Jeffrey Wang and Will Bryk, has raised $85 million in Series B funding, valuing the company at $700 million [1][3] - The company aims to create a search engine that surpasses Google, positioning itself as the AI equivalent of Google [3][4] Company Overview - Exa was founded in 2021, prior to the emergence of ChatGPT, with the belief that AI requires a dedicated search engine [3][4] - The company has developed a large-scale indexing system and various new search technologies to provide users with advanced search capabilities [4][5] Product Features - Exa's search engine is designed specifically for AI, featuring six proprietary characteristics: 1. High-quality knowledge acquisition, optimizing for quality over SEO content [5][6] 2. Comprehensive content access, providing complete page information for AI processing [6][9] 3. Speed, with a goal to create the fastest search API, achieving latency below 450 milliseconds [7][9] 4. High computational capability, allowing extensive searches for various information [9][10] 5. Customization, enabling tailored searches for specific AI applications [10][13] 6. Zero data retention, ensuring sensitive query data is not stored [10][13] Future Aspirations - Exa plans to expand its indexing and processing capabilities significantly, aiming to collect a vast majority of global information and build a GPU cluster five times larger than its current one [13][18] - The ultimate goal is to surpass Google in search capabilities [13][18] Team Background - Jeffrey Wang, a Harvard graduate, co-founded Exa and previously worked at Plaid, focusing on data and network infrastructure [15][18] - The company includes several skilled team members from prestigious universities, enhancing its technological expertise [17][18]

作者：yongxin 编辑：Siqi Agentic AI 的 3 要素是：tool use，memory 和 context，围绕这三个场景会出现 agent-native Infra 的 机会。 01 . 为什么 Search API 很重要 按照场景和信息需求类型，搜索行为大致可以被分为四类： • 第一类，高频快速查询， 指的是一两步内就能完成的查询。Google 大部分的 query 还是以几个单词为 主，用户得到答案后马上离开，不会进行深入的查询。对于这类查询 Google、Bing 还是最好的应用，新 玩家几乎没有挑战的机会。 • 第二类，研究性质的深入查询， 用户可以和搜索工具反复交流，获取知识。这一类搜索是 LLM 和 LRM 带来的新场景，对应的代表性产品形态分别是 Chatbot 和 Deep research。 Agent 所获取到的信息质量是 agent 推理的起点，虽然 LLM 带来了 perplexity 为代表的 AI answer engine，提供了完全不同于传统搜索引擎的体验，但这些产品仍旧面向的是人类用户，产品逻辑是围 绕人类行为设计的。 在我们 MCP 的研究中发现， ...