Source Subscriptions
Continuous ingestion from individual feeds—stay current without manual uploads
The information landscape moves faster than humans can process. Documentation updates frequently. Best practices evolve. Your AI models answer from training data that's months old, missing recent changes.
Source subscriptions monitor individual feeds for continuous ingestion. When new content publishes—documentation updates, blog posts, changelogs—Recurse processes it automatically. Frames get extracted, relationships form, and through temporal versioning, your knowledge stays current as information evolves.
How This Works
Consider asking about recent changes to a technical specification. Without subscriptions, your model answers from months-old training data. You manually check the documentation, realize it changed, copy-paste context. With a subscription monitoring that URL, the system has already ingested updates. Your query retrieves current frames—30 seconds instead of manual assembly.
The key difference: when content changes, Recurse doesn't just add the new version—it recognizes updates and revises existing frames. A source claimed "Method X achieves 95% accuracy." Months later, new information shows production accuracy is 73%. The system recognizes the contradiction and updates the frame, preserving version history. Your knowledge evolves without manual reconciliation.
What gets subscribed:
Track changes for URLs (documentation sites, blogs, specific pages) or subscribe to RSS feeds (newsletters, podcasts, publication feeds). Additional source types like Twitter/Bluesky feeds coming soon.
The pattern: identify relevant sources, subscribe once, benefit continuously. You curate information flows rather than processing individual documents.