Pinned26 October 2026 · Essay
What is SuperSearch?
The last mile of knowledge for superintelligence, and the first product built toward it.
Read the essayComing next
More essays are on the way: on the receipt, on what leaves your machine when an agent searches, and on how a search becomes a unit of work on a chain. One at a time, with the evidence.
Follow on X for the next onePinned26 October 2026 · Essay
The last mile of knowledge for superintelligence, and the first product built toward it.
Read the essay10 November 2026 · Comparison
Features, observability and agent integration, side by side, from each product's own documentation.
Read the comparison10 November 2026 · Research
A router picks one search engine per query. SuperSearch asks every engine and merges, with a receipt for each. Jevlin does both. Measured on 485 questions.
Read the numbers10 November 2026 · Guide
What a multi-search-engine API is, one real call, what each engine adds, what it costs, and when one engine is enough.
Read the numbers10 November 2026 · Research
The same 4,365 queries on Exa, Brave, Parallel and Perplexity: two engines share about two of ten pages, and a page all four return is ten times as likely to be the right one.
Read the data3 November 2026 · Live test
On US midterm election night every race ran through Jevlin: one call, six search engines. No single engine was first on every race. The results, race by race.
See the results27 October 2026 · Benchmark
485 developer questions, five indexes on their own and the router, three phrasings, three trials, every call published.
Read the results