Why 30-year-old BM25 beats vector embeddings for agentic search, per search vet Jo Kristian Bergum
jobergum · x · 2026-10-30
In a free Maven lesson, HORNET.dev CEO Jo Kristian Bergum (ex-Vespa.ai chief scientist, Yahoo Distinguished Engineer) argues retrieval, not reasoning, is what holds most agents back — finding evidence in 1B tokens is the hard part. Key points: default settings and eval harnesses make lexical retrieval look weak; agents search like power users, and GPT-5 writes long, exact-phrase queries in 98% of sessions, which BM25 rewards; design the harness and measure end-to-end task success over single-query metrics. Analogy: grep is core to coding agents, BM25 plays the same role for search agents — often beating popular vector embedding approaches.
More from coding & agent
- Gym-Anything Adds Audit Agents That Try to Break Simulations With Legal Robot Actions — wellecks · 2026-10-30
- CMU's Gym-Anything Auto-Builds 200+ Agent Environments from Any Software, Selected by GDP — wellecks · 2026-10-30
- durable-object-migrator: zero-downtime exports for Cloudflare Durable Objects — samgoodwin89 · 2026-10-30
- This 5-question interview prompt makes Grok design your first two custom bots in minutes — nateliason · 2026-10-30
- Cursor and Harvey swap general models for task-specific brains, but most niches can't afford to — MartinGTobias · 2026-10-30
- Scenario open-sources 3D OS: 80 agent skills taking sketches to game-ready assets — Scobleizer · 2026-10-30