Turning the World into Your Context Window: Rebuilding Web Search to Make AI Reliable in Production

An intelligent system needs access to the world, and the web is the largest live knowledge base ever created — but it was designed for humans, not agents. Legacy SERP-based search breaks in agent workflows, and the search built into models isn't enough for production systems. Sacha argues that web search has to be rebuilt from the ground up for AI: low latency, rich context, custom indexing, and deep research capabilities. Web search is becoming a core infrastructure layer for production AI, and this talk explains what that layer actually has to look like.