Sources
4DA reads from 20+ source adapters, all running locally in the background. Each one fetches public content from an endpoint you configured — there is no 4DA-operated relay in between.
What's covered
| Category | Sources |
|---|---|
| Community | Hacker News, Reddit, Lobsters, DEV.to, Stack Overflow |
| Code & releases | GitHub, crates.io, npm, PyPI, Go modules |
| Research | arXiv, Papers with Code, Hugging Face |
| Security | CVE / OSV vulnerability databases |
| Social & video | Twitter/X, Bluesky, YouTube |
| Launches | Product Hunt |
| Anything else | Custom RSS feeds |
How adapters run
Adapters run on a background interval you control. Each pull is scored immediately against your context, so the database only accumulates items that matter — not a raw mirror of everything published.
Because the scoring signal comes from your filesystem, adding more sources doesn't add more noise. A firehose source and a niche RSS feed are held to the same gate: an item still needs 2+ independent axes to survive.
Custom feeds
Point 4DA at any RSS/Atom feed to fold a newsletter, a changelog, or a personal blog into the same scoring pipeline. It's treated like every other source — no special-casing, same confirmation gate.
Next: The 5 axes or Privacy & BYOK.