Hashgrid DNA: You search. DNA learns.

Hashgrid DNA

2 min read Original article ↗
Illustration: a double helix whose strands are your searches and your items. DNA proposes results between them; results your feedback confirms light up. Counters show top results found and the LLM calls saved compared with RAG plus a reranker.

A search engine that learns from your grades. The more you grade, the less you call your LLM.

Your judge. Our search.

Your items

Text, records, images, or your own embeddings.

DNA

Finds the best result. Learns from every grade.

Judge

An LLM, a small model, or your users.

Store, ask, grade.

Send your items, ask, and send back a grade when you have one. DNA embeds and trains on its side, so you don't host or tune a model.

Items can be tools, products, documents, people: anything you choose between.

Read the quick start

How many LLM calls?

7× fewer LLM calls than RAG + Reranker, on average.

100 calls

LLM checks every item

36 calls

RAG + Reranker

5 calls

DNA

one of the top 3 results not a top 3 result How many times an LLM is asked before the top 3 results among 100 items are found. Average across three test tasks: matching people, finding papers to cite, and picking tools.

Built for questions like these.

  • Which tool should my agent call?
  • Which agent should I delegate this task to?
  • Which product fits this customer?
  • Which document answers this question?
  • Which candidate fits this job?
  • Which ticket goes to which team?