Extract falsifiable hypotheses from a paper or research source. Use when the human says "distill this paper", "what are the key claims", or shares an academic paper/article for deep analysis.
Installation
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Extract falsifiable hypotheses from a paper or research source. Use when the human says "distill this paper", "what are the key claims", or shares an academic paper/article for deep analysis.
hypothesize
When to use
"distill this paper" or similar
Human shares a paper, preprint, or research article for analysis
Human asks for key claims, hypotheses, or falsifiable predictions from a source
Process
Read the full paper (URL, PDF, or pasted text)
Retrieve knowledge neighborhood — search existing knowledge for related concepts, claims, and patterns using 4-surface retrieval (lexical, structural, semantic, temporal). This primes cross-domain connections.
SUPER IMPORTANT: Assess ALL primary authors' work history and ALL primary references for relevance to the user's broader work. Identify ideas worth capturing.
Create the source note in superpaper/sources/<topic>/ with: title, authors, abstract summary, methodology notes. Give every key finding and quotable passage a ^block-id so it's independently embeddable (![[source#^finding-1]]).
Extract atoms aggressively (atomic-first). Every distinct finding, method, framework, and quotable claim from the paper deserves its own note — not just hypotheses. A key finding → its own note linked to the source. A novel method → its own note. A quote by an author → its own note linked to their people/ entry. The more atoms, the more potential cross-domain bridges.
Extract 3–5 falsifiable hypotheses from the paper's claims and broader ideas gathered that could address limitations of the work/area under consideration or fix gaps in the current paradigm.
For each hypothesis worth remembering, create a claim note in superpaper/concepts/:
type: claim, confidence: 0.4–0.7
Include predictions ("if true, expect…") and embed the evidence passage (![[source#^finding]])
Tags: #domain/... + #topic/...
Link to the source note and any related existing knowledge notes
Compose a hub note if the paper produced 4+ atoms — embed the atomic notes with thin connective prose explaining how the findings relate. This becomes the "processed paper" that reads like a document but traces every claim to its source.
Link everything: source note links to claim notes, claim notes link to each other and to related existing knowledge. Update 1–3 existing notes to link back (distributed write).
Outputs
1 source note in superpaper/sources/<topic>/
1+ evidence notes (key excerpts) in superpaper/.evidence/
2+ claim notes (hypotheses) in superpaper/concepts/
All cross-linked with distributed write
Done when
Hypotheses are testable and falsifiable
Claim notes have predictions ("if true, expect…")
Source note has proper source: frontmatter
All notes have 2+ wiki-links with distributed write (existing notes updated to link back)
Tags are specific (#domain/... + #topic/...), not broad