The story’s score and claims

Comments on No. 26 of 123 · modem.ai · edition of Saturday 14 November 2026

Swapping pgvector for turbopuffer made our whole database faster and cheaper

Summary

Modem, a product that groups customer messages into topics, found its pgvector-based similarity search slowing sharply as data grew, with its smallest customers the slowest. Adding an HNSW index made results worse and altered some matches, so the team dropped it.

Comments

Ranked by a public rubric, not by votes: what a comment adds counts, not who wrote it. How comments are ranked · Comment rules