Back to browse
GitHub Repository

A new search paradigm where documents have gravity, queries converge into basins, and multi-signal scoring uses interference instead of linear fusion.

36 starsRust

I rebuilt search using physics instead of statistics. +18.5% NDCG 10

by KendallCBooker·Apr 4, 2026·2 points·0 comments

AI Analysis

●●●BangerBig BrainWizardryZero to One

Gravity-based search with interference scoring beats linear fusion by 18.5% NDCG.

Strengths
  • Genuinely novel paradigm replacing nearest-neighbor search with gradient descent convergence
  • Interference terms (sqrt(V*T)) capture signal agreement instead of losing information
  • Pure Rust implementation with sub-millisecond latency and formal proofs included
Weaknesses
  • Physics metaphor may confuse teams expecting standard vector search APIs
  • Adoption depends on convincing engineers to abandon familiar BM25/HNSW patterns
Category
Target Audience

Search engineers, ML researchers building custom retrieval systems

Similar Projects

Health●●Solid

Health optimization as agent-guiding gradient descent

Research-as-loss-function lets stale-knowledge agents guide health optimization via gradients.

Big BrainNiche Gem
dingmuti
103mo ago