Lepton Search

Ask a question, get a written answer with numbered citations, the source cards behind it, and a set of follow-up questions — the shape everyone now recognises from Perplexity. The point of Lepton Search was that this shape is mostly product design over commodity parts: a search backend, an LLM, and a stream. The whole thing fits in under 500 lines.

I created it and it went out as an open-source project in January 2024. It picks up whichever search backend you have a key for — Bing, Google, SearchApi, Serper — runs the results through an LLM, and caches each answer so a result is a shareable URL rather than a session.

It reached over 8,000 stars. The repository is archived now: the pattern it demonstrated has long since become standard.