Show HN: I analyzed 297k HN posts – your fate is sealed in 2 hours

I built an archiver that tracked every HN item for 27 days (Dec 3-30, 2025) and captured 72k temporal snapshots to see how posts gain and lose attention.

Key findings:

- Early velocity (points/hour in first 2 hours) predicts final score with ρ=0.82. A simple classifier achieves 98.4% precision for viral prediction.

- No Matthew effect: high-scoring posts don't get more upvotes per hour than low-scoring ones (ρ=-0.04). HN's gravity penalty actually works.

- Yet extreme inequality persists: Gini=0.89, meaning bottom 80% of posts get <10% of total upvotes. Inequality without cumulative advantage.

- Attention decays as a power law (α=0.52), slower than exponential; quality content has longer tails.

Paper (open access): https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5910263

Code + data: github.com/philippdubach/hn-archiver