Scry is a programmable internet-research platform that exposes billions of source-native records as queryable relations and lets agents run bounded, Turing-complete programs over them. Agents issue single read-only queries subject to deadlines, memory ceilings and row caps, operate on documented relations (reddit.posts, reddit.comments, academic.papers, etc.), use vector helpers and embeddings, and retain source-native identifiers and provenance. Common workflows include graph traversals (e.g., rebuilding entire comment threads to a fixpoint), embedding-based concept minting and ranking (averaging multiple concept vectors to find related posts while excluding tokens), and standing queries that watch records without running separate crawlers. Every relation’s schema, indexes, extent and known holes are discoverable via the schema endpoint.
Scry publishes coverage and freshness metrics (≈158.7 billion queryable rows across 43 sources, 413.3 billion rows held, ~37.7 billion rows landed in the last 24 hours; Hacker News and LessWrong arrive in under 15 minutes) and reports a DeepSearchQA result where an agent using Scry scored 71.8% exact-set correctness versus 66.1% for a Google entry. Integration is via an MCP server (https://mcp.scry.io) for ChatGPT, Claude and other clients or via an HTTP API with API keys. Pricing floats with congestion so heavy queries cost more under load. The platform positions itself as public-benefit infrastructure, subsidized for noncommercial research, with traceable public evidence and reviewed access available for additional sources.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.