Expand 发布 Expand Fetch，让 AI 智能体读取网页时保留图标与表格
This looks like a big step for agents that read the web.
AI 初创公司 Expand 发布 Expand Fetch，让 AI 智能体读取网页时不再丢失图标、表格和隐藏数据。在 Expand 自有的 1094 道题、326 个网页的基准测试中，使用该工具的智能体答对 74%，次优 API 为 55%。它用自研视觉模型 Iris 像人一样看页面，每次抓取还会生成快照，Markdown 每一行都能链接回来源，便于开发者追溯错误答案。该工具今日上线，定价为每 1000 次抓取 1 美元。
查看原帖Introducing Expand Fetch, a new way to scrape and crawl the web. Your agent has never seen the web. No wonder it gets basic questions wrong. Traditional Markdown APIs lose critical details. So we trained two System one models to change that: Iris, a vision model that sees pages like you do, preserving tables, icons, and layouts. X-Ray, a JSON model that…
展开完整原文
Introducing Expand Fetch, a new way to scrape and crawl the web. Your agent has never seen the web. No wonder it gets basic questions wrong. Traditional Markdown APIs lose critical details. So we trained two System one models to change that: Iris, a vision model that sees pages like you do, preserving tables, icons, and layouts. X-Ray, a JSON model that uncovers the hidden data behind every page. Every fetch also comes with a Wayback Machine. Every line links back to its source, so you can verify exactly what your agent saw. Three products. One API. On our benchmark of 1,094 questions across 326 challenging web pages, agents using Expand got 74% right. The next best API got 55%. Available today at $1 per 1,000 fetches. Try it at https://t.co/x4SDYsUX25
正文
当前显示原文，中文全文译文尚未就绪。
This looks like a big step for agents that read the web.
Expand, a new AI startup, released Expand Fetch, a tool that lets AI agents read web pages without losing icons, tables and hidden data along the way. On Expand’s own private test, agents using it answered 74% of hard web questions correctly, against 55% for the next best tool.
When AI agents read a website today, the page is usually converted to plain text first, and details like a ✗ next to a feature get thrown away, so the agent confidently gives wrong answers.
Expand Fetch, uses its own vision model, Iris, to look at the page the way a person does and keep those details intact.
Every fetch is also stored as a snapshot whose Markdown lines link back to their source, letting developers trace a wrong answer to what the agent saw.