Take the collection with you.

Search it your way. Give it to your agent. Follow a thread we missed. The same data that powers this site, without an account.

The complete dataset

All five files below, together in one TAR.GZ archive. No PDFs.

Download the complete dataset

Or choose a file.

Read the data notes

Published dataset files and individual downloads
File & contents	Download
papers.jsonl 1,237 papers One paper per line: metadata, topics, score, and recurring mentions.
mentions.jsonl 7,820 excerpts Up to 20 attributed comments per paper, with original HN identifiers and rule-based stance.
topics.json 21 topics Topic names and their paper identifiers.
DATA.md Method & attribution Field definitions, scoring, limitations, provenance, and rights.
manifest.json Build receipt Snapshot dates, thresholds, counts, and SHA-256 checksums.

What you’re downloading.

These files contain the entire published catalog. The per-paper mention count covers all qualifying comments; the exported excerpts are capped at 20. PDF files are never downloaded or hosted here.

Attribution travels with the data.

HN archive data is attributed to open-index under its declared ODC Attribution license. Original comments and paper abstracts retain their respective authors’ rights; the database license does not grant a new license to those texts. Read DATA.md before reuse.

Stance and topics use transparent rules. Human review: pending. Snapshot through August 21, 2026. Read the limitations.

Visit the source archive