The Hard Part with Evan McCann Simon Hørup Eskildsen from turbopuffer
15 snips
Aug 25, 2025 
Simon Hørup Eskildsen, Co-founder and CEO of turbopuffer, shares insights on revolutionizing data search. He discusses the blend of vector and full-text search in his low-cost engine, highlighting real-world applications with companies like Cursor, Notion, and Linear. Simon explains the intricacies of caching mechanisms, scaling challenges, and how TurboPuffer's S3-first architecture optimizes performance. He also navigates the startup landscape, emphasizing founder-led sales, transparent pricing, and the importance of emotional resilience in entrepreneurship.
AI Snips
Chapters
Transcript
Episode notes
From Shopify To TurboPuffer Launch
- Simon built TurboPuffer after nearly a decade working on databases at Shopify and launched it solo in summer 2023.
- Cursor became an early customer on day one and cut their search costs by 95% after migrating to TurboPuffer.
Why Vector Search Changed Everything
- Vector embeddings map unstructured data into high-dimensional coordinates where proximity implies semantic similarity.
- Embedding-based search gives superior results but dramatically increases storage per item compared to raw text.
Store On S3, Cache For Speed
- Storing primary data in object storage (S3) drops per-GB costs dramatically versus memory or disk.
- TurboPuffer's architecture uses S3 as the source of truth and layered caching to reclaim performance at far lower cost.