Find the passage. Keep the page number.
Drop in papers, reports and slides. Thread extracts their text on your device, page by page — so a search or a question lands on the exact page, right beside your own notes on it.
- PDF text extracted page by page, on your device
- Scanned pages and figures read with local OCR
- Answers cite the paper and the page
Sleep & memory · Literature
In this Space · 38 items
Hall & Ortiz 2021 — retention.pdf
PDF · 24 pages
Spacing meta-review.pdf
PDF · 61 pages
Library scan — ch. 6.pdf
Scanned PDF · OCR
Lab meeting — March.pptx
Slides · 18
My reading notes: spacing
Note
one-week delay
Hall & Ortiz 2021
PDF · page 7
…after a one-week delay, the spaced group recalled 21% more items…
My reading notes: spacing
Note
Careful: only a one-week delay — check longer intervals.
Library scan — ch. 6
Scan · page 143
…retention at a one-week and one-month delay…
You read it somewhere. Which paper, which page?
Research piles up across downloads folders, reference lists, slide decks and half-finished notes. The finding you need is in there — but finding it again means reopening twenty PDFs.
Thread makes every page of every file searchable, keeps your notes next to the sources, and shows where each answer came from.
How it works, step by step.
Step 1
Collect everything for a topic in one Space
Supported files and extractionDrop PDFs, Word files, slide decks, spreadsheets and links into a Space. Originals stay unchanged; Thread extracts searchable text in the background — page numbers for PDFs, slide numbers for decks.
Just added
- Text extracted by page
Spacing meta-review.pdf
PDF · 61 pages
- Searchable by slide
Lab meeting — March.pptx
Slides · 18
- OCR on this device
Figure 2 — photo.jpg
Image
- Saved
journal.example.org/article/…
Link
Step 2
Write notes in your own words — and mark what’s still a hunch
How typed statements workLink concepts with
[[Spacing effect]]and people with@Name. Start a line with Hypothesis: and Thread will never present it as fact — answers label it “Unconfirmed idea”.Read [[Spacing effect]] section of Hall & Ortiz again. Ask @Mara Lind whether the lab replicated it.
Unconfirmed ideaHypothesis [Spacing]: Benefits shrink for procedural skills.
Fact [Hall 2021]: n = 212, one-week retention test.
Step 3
Search every page you’ve ever saved
How search worksPress / and search across titles, notes and the text inside every PDF, scan and slide. It runs on your device, tolerates typos and works offline — on a train or in the archive.
interleaving
Spacing meta-review.pdf
PDF · page 34
…interleaving produced larger effects for category learning…
Lab meeting — March.pptx
Slide 11
Next study: interleaving vs. blocked practice
Ideas for chapter 3
Note
Could interleaving explain the gap in Table 2?
Step 4
Check the source before you quote it
Ask & Why? explainedWhen you ask a question, each statement in the answer carries a numbered citation. Open it to read the passage in context — the page, the slide, the note — before it goes into your draft.
Source · Hall & Ortiz 2021 — page 7
Participants in the spaced condition outperformed the massed condition; after a one-week delay, the spaced group recalled 21% more items, although the difference narrowed for…
Open PDF at page 7Page 7 of 24
Ask across your whole literature.
Ask in plain language. Thread finds the relevant pages and your notes, answers with citations, and lets you open each one. Click the numbers.
Ask · Sleep & memory
Two of your sources report better long-term retention with spaced sessions. Your own note flags that one study only tested a one-week delay, so longer intervals are not covered.
Source 1 of 3
Hall & Ortiz 2021
PDF · page 7
Participants in the spaced condition outperformed the massed condition; after a one-week delay, the spaced group recalled 21% more items.
Open PDF at page 7
Example content. Written answers need a local model or your own API key; without AI, Ask assembles an answer from your decisions and matching passages.
What makes it work.
- Page-level PDF extractionText is extracted per page, so search results and citations point to the page.
- Local OCR for scansScanned PDFs, photos of pages and figures become searchable without an OCR service.
- Hypotheses stay hypothesesTyped hypotheses are labelled “Unconfirmed idea” and never presented as facts.
- Keep notes currentMark a note outdated or set a review date; answers tell you when they rely on one.
- Work with Claude or CodexConnect an AI tool via MCP to search your Thread and save notes — with a connection you control.
- Export everythingTake your notes and original files with you at any time, with missing files reported.
Unpublished work stays where you put it.
Extraction, OCR and search run on your device. Mark a Space “Keep on this device only” and it is never uploaded or synced; mark items “Don’t use with AI” and they stay searchable but are never sent to a model.
See exactly how data movesPractical questions.
Which files can I search?
PDF (including scanned PDFs via local OCR), DOCX, PPTX, XLSX, ODT/ODP/ODS, RTF, CSV, JSON, HTML, Markdown and text files, images, and transcribed audio. A single file can be up to 200 MB.
Does Thread connect to my reference manager?
No. Thread doesn’t sync with reference managers. Drop the PDFs you’re working with into a Space; you can import Notion pages you choose and export everything at any time.
Does Thread summarise papers?
With a local model or your own API key, Thread can suggest titles, tags and summaries. They are suggestions and never replace what you wrote. Extraction, search and citations work without AI.
Can I trust the answers?
Answers are checked against the retrieved passages; unsupported statements are removed or flagged. Every sourced statement has a citation you can open — and for anything you’ll publish, you should.
Other ways people use Thread
Every source, one search away.
Try every feature for 30 days. No credit card; your library stays readable and exportable whether you subscribe or not.