RLMs May Compose Search as a Tool Using Grep, Regex, and ColBERT

CShorten30 · x · 2026-09-01

Search may not disappear in a world of long-context models; instead, it may become something a Retrieval Language Model (RLM) learns to compose. Alex Zhang explains how an RLM can combine grep, regex, and ColBERT-style token-level retrieval as callable tools to analyze messy, evolving agent logs via programmatic decomposition on the Weaviate Podcast.

Original post →

More from coding & agent

coding & agent channel →