Methodology v0.1
Durability, not popularity.
The core score asks whether different people found the same idea valuable across time. Discovery lenses answer different questions and never alter the core score.
Same good idea.
The 100-point durability score
Repeat traction
How often the link reaches the top 10% or 2% of its own year.
Independence
Distinct submitters, with Bayesian shrinkage for tiny samples.
Persistence
Active years and observed span, both log-scaled.
Generational reach
Strong appearances separated across genuinely different eras.
Discussion
Repeated substantial discussion, intentionally kept subordinate.
Why annual percentiles?
HN grew, voting behavior changed, and raw point totals drifted upward. Each submission is therefore compared with stories from the same calendar year. A historically exceptional 2008 post receives the same traction signal as a similarly exceptional 2026 post.
Confidence is separate
A link may score well on limited evidence. We publish an evidence-confidence percentage beside the durability score rather than secretly suppressing or over-rewarding it.
Discovery lenses
Forgotten classics, newly resurfaced, debate magnets, unusually comment-heavy, most persistent, and hidden gems are deterministic secondary scores. “Debate magnet” is explicitly a controversy proxy based on unusually high discussion relative to peer posts; HN does not expose downvotes or disagreement.
Identity and corrections
URLs are normalized automatically and redirects are recorded during survival checks. Ambiguous identities remain separate until a reviewed correction is available. Link survival is reported separately from cultural durability.
