Intel / article

Fast Agentic Search

Source: relace.ai
Author: Relace
Date: Published Dec 8, 2025

Context

Relace's Fast Agentic Search delegates codebase exploration to a specialist model. It reads and searches several file paths in parallel, then returns relevant code to the main coding . Relace's evaluation reports roughly fourfold faster search, but the whole coding task improved less.

On 500 problems, adding the search specialist reduced median completion time from 7 minutes 44 seconds to 7 minutes 1 second. The share solved was 71.4%, compared with 72.0% for the original agent. Faster search therefore produced a modest overall speed gain in this test, where problem descriptions often already identified relevant files.

Terms in this piece · Glossary

AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
eval — A repeatable test for AI quality — a set of tasks plus scoring — used the way software teams use test suites, because model output is too variable to judge by eyeballing.
SWE-bench — The standard benchmark for AI coding agents: real GitHub issues from real repositories, scored by whether the agent's patch passes the project's own tests.
multi-agent — Using several AI agents on one problem — splitting work in parallel, checking each other, or filling different roles like planner and reviewer.

Read the source relace.ai