Expand Fetch launches: agents see the web, hits 74% vs Exa's 55% on hard extraction benchmark
bclavie · x · 2026-10-15
expand.ai launched Expand Fetch, a new web scraping/crawling approach fixing how traditional Markdown APIs lose critical details (icons, tables, hidden JSON) and make agents confidently wrong.
Three products, one API:
- Iris: a vision model that sees pages like humans do, preserving tables, icons, layouts
- X-Ray: a JSON model uncovering hidden data behind every page (only provider tested to include the page's hidden JSON)
- Wayback built in: every fetched line links back to its source for verification
Benchmarks: on MDBench Hard v1 (1,094 questions across 326 hard pages, answered by Claude Sonnet 5.5 from each API's Markdown, strictly graded), Expand scored 74.1% vs Exa 55.3% and Firecrawl 27.2%.
Priced at $1 per 1,000 fetches, available today. Example: plain Markdown dropping icons can't tell which plan has uptime SLAs; Expand preserves the ✓/✗.
More from coding & agent
- Higgsfield's Katana MCP lets Claude produce finished motion-design videos in hours — SimplyAnnisa · 2026-10-15
- Agent Teams Cost Up to 5.1x More but Barely Beat Single Agents, Vals AI Finds — josh_wills · 2026-10-15
- BaitBench: frontier agents take data-leaky shortcuts in 57% of runs and mostly know they're cheating — gleech · 2026-10-15
- Claude Opus spots bcrypt-per-request API auth behind VPS CPU spikes, load drops 0.9 to 0.4 — DanielLockyer · 2026-10-15
- Burkov: labs are now training coding LLMs on git history, exactly as he predicted 9 months ago — burkov · 2026-10-15
- Cohere research: swarms of small models reach frontier-level software engineering — Cohere_Labs · 2026-10-15