How Exa is Building the Perfect Search Engine | Deep Dives with a16z

a16z Deep Dives · 2026-06-04 · 49 分鐘
https://www.youtube.com/watch?v=k6OkMkDdbUY影片總結
Exa CEO Will Bryk says agent search needs comprehensive results, and retrieval can cut customer token costs by up to 20x.
Exa cofounder and CEO Will Bryk tells a16z’s Sarah Wang that he began pursuing better search in college, then saw transformers make a new approach possible in 2021. Exa launched its search engine in November 2022, two weeks before ChatGPT, and later found demand from AI products. Bryk says agents need complex, controllable queries and comprehensive results—sometimes 1,000 or 10,000, not just 10—while businesses need search reliable enough for hiring, sales, and investment decisions. He argues retrieval lets smaller models work more effectively and says Exa can reduce customers’ costs by up to 20x. He also describes research using reinforcement learning on Exa search, predicts agentic search could exceed Google Search’s business by the 2030s, and says infrastructure and unrecorded data will become major bottlenecks. He closes on Exa’s culture: fun, ambitious work and hiring people with visible passion.
章節
- 0:00Exa Origin Story: Will Bryk’s 2021 mission to build better-than-Google search
- 2:35Beating Google With LLMs: Exa’s November 2022 launch and GPT-3 search vision
- 9:00Search Built For Agents: Exa’s semantic queries, chemical-formula keywords, and filters
- 11:27Search Built For Agents: Exa’s 10,000-result retrieval and LLM reranking
- 15:47Search Built For Agents: Agent search’s 99.9999% reliability bar
- 19:03Search Built For Agents: Exa’s go-to-market intelligence and search beyond the web
- 22:30Search Built for Agents: Exa Improves Devin’s Coding Retrieval
- 26:01Tokenpocalypse and Retrieval: Exa Says Retrieval Can Cut Costs 20x
- 29:21RL for Better Search: Exa’s Tool Used Fewer Calls with Higher Performance
- 32:23Benchmarks and Agentic TAM: Bryk Sees Agentic Search Exceeding Google Ads in 2030
- 36:52Benchmarks and Agentic TAM: 1,000x Search Growth Brings Infrastructure and Data Bottlenecks
- 40:19Benchmarks and Agentic TAM: Exa’s Goat, Mission-Driven Naming, and 10¹⁸ Meaning
- 44:20Culture Hiring and Wrap: Exa’s project freedom, search mission, and passion-led hiring
完整分析還有什麼
13 章 · 72 個重點 · 13 段段落總結 · 心智圖
第一章預覽 · Exa Origin Story: Will Bryk’s 2021 mission to build better-than-Google search
其餘 69 個重點、段落總結與心智圖在完整分析裡。 分析這支影片,看完整內容 →
這支影片 50 分鐘,免費方案可分析 30 分鐘以內的影片;Plus 方案起可分析 2 小時以內。
這是 Tier 1 公開摘要
每章重點、段落總結、心智圖由分享者控制是否公開。想看完整分析?自己提交一支。
同頻道的其他分析
To Regulate AI Effectively, Focus on How It’s Useda16z Deep DivesMartin Casado argues AI laws should target harmful uses, warning regulatory uncertainty is pushing U.S. startups toward Chinese open-source models.
AI Copilots Are a Dead End. Here's What Actually Works | Kavak CEOa16z Deep DivesKavak CEO Carlos García Ottati says AI agents now handle 90–95% of customer interactions after a year of flat growth.
Are Tariffs Fueling the AI Race? | Deep Dives with a16za16z Deep DivesJacob Helberg says winning the AI race requires leading models, global adoption, secure supply chains, and rebuilding U.S. manufacturing.
Why Modern War Needs Intelligent Power Systems | Chariot Defense CEO on a16za16z Deep DivesChariot Defense’s M424 hybrid system delivered 36 hours of low-signature power as Army leaders push faster procurement and U.S. battery production.
Submarines and the Future of Defense Manufacturing | Hadrian CEO on a16za16z Deep DivesHadrian opened its 2.25-million-square-foot Alabama factory as the Navy faces 70 million submarine-production hours and a severe workforce gap.
相關主題的分析
Chelsea Finn: This is the State of the Art in RoboticsY CombinatorChelsea Finn says Physical Intelligence’s robots ran a latte task for 13 hours and RL doubled throughput.
AI, Infrastructure, and the Next Investment Cyclea16za16z’s David George, Sarah Wang, Alex Immerman, and Santiago Rodriguez examine $780B hyperscaler CapEx and AI adoption’s 2% measurable impact.
Can Open Models Keep Up With OpenAI and Anthropic? | Nathan Lamberta16zNathan Lambert says China matches U.S. AI talent but lacks compute, while open models trail frontier labs on different timelines.