Search Tips

It defaults to Exa for semantic search with highlights enabled, Firecrawl CLI for scraping with main content extraction, and Reddit MCP for discussion threads. The key insight is treating research as alternating cycles of searching and fetching, with specific guidance on when to break the defaults (like using Exa full text as a scraping fallback for blocked sites, or dropping the only-main-content flag when Future plc sites strip their own articles). Includes query reformulation tactics and content-type strategies across news, academic papers, code, and social sentiment. The reference files cover edge cases but the main doc gets you started without rigid rules.