About Profoundd

What Profoundd is

Profoundd is an independent search engine and document archive, built and run by one person on self-hosted hardware in Oregon. It does two different jobs, and it is worth being clear about which is which.

  • Search over news and reference sources. We crawl RSS/Atom feeds from 303 vetted sources, index the articles, and link you to the publisher.
  • A hosted archive of primary documents. Here we are not a directory pointing elsewhere — we store the documents and serve them ourselves: the DOJ Epstein releases (over 1.3 million indexed documents), roughly 210,000 archive documents across 20 collections (court filings, FOIA releases, congressional testimony, agency records, historical newspapers), transcript archives, Webster's 1828 Dictionary, and the full text of the US Constitution.

If you are reading an archive document on this site, you are reading a copy we host and are responsible for — not a link we found.

What we hold, and what we don't

  • Some archives are text only. Several collections (commentary, medical talks, hearings) are transcripts produced by speech-to-text from publicly posted audio and video. We keep the transcript. We do not host the original audio or video, and we have no control over whether it stays online — so a transcript may outlive its source link.
  • Machine transcription and OCR make mistakes. Transcripts and scanned-document text are automated output, not certified records. Where a wording matters, check it against the original.
  • A few collections are not public. They are indexed so the site's subject Q&A can cite them, but they are not publicly browsable.
  • We honor redaction requests on archive material where a named person has a legitimate claim. A redacted document is removed from search, marked noindex, and its file stops being served.

Categories we cover

  • Prediction Markets - Live Odds & Forecasts from Polymarket
  • Polls - National & State Election Polls with Pollster Ratings
  • News - General & World News
  • Epstein Files - Jeffrey Epstein Case Documents & Coverage
  • Charlie Kirk - Charlie Kirk Coverage & Commentary
  • Medical - Health, Medicine & Research
  • Legal - Law, Courts & Regulations
  • Technology - Innovation, AI & Digital Frontier
  • Finance - Banking, Economy & Business
  • Science - Research, Space & Discovery
  • Education - Higher Ed, Research & Learning
  • Environment - Climate, Energy & Conservation
  • Politics - Policy, Government & Analysis
  • Markets - Stocks, Crypto, Precious Metals & Trading
  • Legislative Watch - Federal Bills, Hearings, Regulations & Executive Orders
  • YouTube - Video Content from YouTube Creators
  • Rumble - Video Content from Rumble Creators
  • State Watch - Oregon, Washington & California State Government

How ranking works

  1. Text relevance (BM25) - how well the document matches your query.
  2. Source credibility - each of the 303 crawled sources carries a 1–10 score that boosts or dampens its results. These scores are our own editorial judgment, set and adjustable in the admin panel. They are not awarded, audited, or certified by any outside body.
  3. Imported domain ratings - separately, we hold about 2,600 domain ratings imported from public datasets (the Iffy Index, CRED-1, Wikipedia's perennial-sources list, Media Bias/Fact Check). Where one exists it informs our score; it does not replace it, and we sometimes disagree with it.
  4. Recency - recent articles (last 24 hours to 7 days) get a boost on news surfaces.

About the AI answers

Some pages offer an AI-written summary or answer. Two things you should know:

  • A language model wrote it, over documents we retrieved. It can be wrong, and it is not a substitute for reading the source.
  • Every answer passes a deterministic verifier before you see it. The verifier strips any citation, URL, DOI, docket, case or study identifier that does not appear in the sources actually retrieved, and refuses the answer outright when what is left is mostly unsupported citation. Claims we could not tie to a retrieved source are marked ⟦unverified⟧.

The verifier exists because language models invent citations that look real. It catches that class of error; it cannot catch every wrong statement. Treat an AI answer as a reading aid and the document as the record.

What we don't do

  • No ads, no ad network, no third-party trackers, no data sales. See the privacy policy for exactly what is logged.
  • No Google properties in the index. Google-owned domains are excluded at crawl time, at serve time, and by a daily audit.
  • Profoundd is not open source. The site runs on code we have not published, and there is no public repository. (An earlier version of this page said otherwise; that was wrong.) What is open is the data — see below.

Open data

Every public archive collection is exported as JSONL, refreshed daily, with a manifest of document counts and checksums, published under CC-BY 4.0. Browse them at /datasets. Private collections are excluded from the export.

API Access

Profoundd provides a free JSON API:

GET /api/search?q=your+query&category=news&page=1

Returns JSON with articles, total count, and pagination info.