I built a privacy-focused search engine from scratch (and learned why search is hard)

Short summary

A developer shares lessons from building Slick, a privacy-focused search engine, discovering that crawling, indexing, ranking, and infrastructure are far harder than expected. The project uses Elasticsearch for indexing, BM25 plus semantic embeddings for ranking, and defaults to DuckDuckGo's index while its own matures. The piece is more of a project overview than a technical deep dive.

• •Slick is a privacy-focused search engine combining BM25 keyword matching with semantic embeddings and reranking
• •Crawling, indexing, storage, and ranking require far more infrastructure than anticipated
• •Project is early-stage, defaults to DuckDuckGo's index, and is actively soliciting feedback

Generated with AI, which can make mistakes.